DeepSeek Disrupts AI Landscape: Challenging Big Tech's Dominance

DeepSeek and Big AI: Good ideas can beat $billions

You can't make monopoly money without a monopoly, but you sure can lose it Opinion It would take a heart of stone not to explode with joy at the massive infusion of schadenfreude provided in recent days by the DeepSeek AIpocalypse. Trillion dollar markdown in tech stocks, slack-jawed panic at tech companies previously "too big to care" suddenly caring a whole lot, and the mainstream media unable to talk about anything else. All through a single app from an unknown Chinese company with the on-the-nose logo of a whale in its death throes. If that's not enough, the sound of the cosmically misnamed OpenAI grunting in indignation over abuse of intellectual property in training data by an actually open AI is funnier than any three Monty Python sketches combined. "What's yours is mine, what's mine is mine too" makes a great state motto for Silicon Valley, but "information wants to be free" might be even better. Some have argued that this is a Sputnik moment, when a disdained rival suddenly leapfrogs a complacent establishment figure. That certainly fits the New Cold War narrative, but there's a better analogy closer to home - the rise of the PC architecture at the expense of the mainframe. IBM was the establishment, and it created the PC to keep things that way. It didn't bother to burden the design with too much intellectual property in either OS or hardware, reckoning that since it owned so much of the IT market anyway, it couldn't lose. Then Compaq overwhelmed the tiny legal defense of the copyright BIOS and the barbarians poured in through the gates. (As we explained here, back in the early '80s, when Big Blue started building the first IBM PC, the only part of the computer it had copyright control of was the BIOS ROM chip, and before long Compaq had figured a way to reverse-engineer it so that it could sell IBM-compatible systems for less than Big Blue was charging.) DeepSeek has pulled off a similar trick, using a variety of existing AI ideas to not only replicate the magic beans that the giants have built their beanstalk business models on, but to start giving it away. Oh, and it (apparently) costs much less in chips and watts to train. Of the three underlying assumptions that have passed for genuine rationales in the AI bubble - LLMs will drive innovation, they will create untold billions in added value, and only the very largest tech companies can play - all but the first has just been disproved. The one that's left, the assumption that LLMs will change the world in wonderful ways, is going to be much easier to test if DeepSeek's claims about training costs are true. AI works best when it is reined into specific tasks instead of generalities, and many more companies, institutions, and researchers will be able to experiment here. That's badly needed, with Apple, Google, and Microsoft desperately trying to force-feed the bad stuff down our necks. It's also bad news for the parody of an industrial policy recently adopted by the US and UK governments, that simply pouring billions into giant ocean-boiler datacenters will compensate for lack of thought or investment in infrastructure that will stay relevant for more than five years, which is good news for those of us left in the reality zone. But then, if you hadn't twigged what it means that the initial investors in Stargate are Oracle and Softbank, that may not be a zone you have a visa for. There is a notion at large that none of this matters, as taking the ideas, indeed the code, of DeepSeek and decanting that into giant ocean-boiler datacenters will allow Big Tech to regain the advantage. Microsoft seems to have reached this conclusion remarkably quickly. Yet this presupposes that throwing money instead of smarts at LLM development will always work. When you have more money than smarts, that's a compelling argument, even though it's just been shown not to be true. All this is saying nothing about DeepSeek's nature as a product of China. The usual caveats apply: don't run the app unless you trust the Chinese state to behave itself, do get stuck into running it locally if you're interested, and don't be surprised if restricting what high tech a country can buy results in uncomfortable consequences over time. Best jam a plan B in there. It is in the nature of technology to constantly evolve, just like living systems, and just as hard to predict what small furry mammal is going to eat your dinosaur lunch. And, again like biological evolution, each new burst of novelty acts as scaffolding for the next. In this case, we've filled the planet with networked pocket computers of incredible power. On top of that, a massive gene pool of open source components can spawn novel organisms at tremendous speed. This can give a new idea universal impact virtually overnight. It took 24 years for IBM to be driven out of the PC market it created. 24 hours may soon be closer to the mark. ®

DeepSeek: what you need to know about the Chinese firm disrupting the AI landscape

Before January 27 2025, it's fair to say that Chinese tech company DeepSeek was flying under the radar. And then it came dramatically into view. Suddenly, everyone was talking about it - not least the shareholders and executives at US tech firms like Nvidia, Microsoft and Google, which all saw their company values tumble thanks to the success of this AI startup research lab. Founded by a successful Chinese hedge fund manager, the lab has taken a different approach to artificial intelligence. One of the major differences is cost. The development costs for Open AI's ChatGPT-4 were said to be in excess of US$100 million (£81 million). DeepSeek's R1 model - which is used to generate content, solve logic problems and create computer code - was reportedly made using much fewer, less powerful computer chips than the likes of GPT-4, resulting in costs claimed (but unverified) to be as low as US$6 million. This has both financial and geopolitical effects. China is subject to US sanctions on importing the most advanced computer chips. But the fact that a Chinese startup has been able to build such an advanced model raises questions about the effectiveness of these sanctions, and whether Chinese innovators can work around them. The timing of DeepSeek's new release on January 20, as Donald Trump was being sworn in as president, signalled a challenge to US dominance in AI. Trump responded by describing the moment as a "wake-up call". From a financial point of view, the most noticeable effect may be on consumers. Unlike rivals such as OpenAI, which recently began charging US$200 per month for access to their premium models, DeepSeek's comparable tools are currently free. They are also "open source", allowing anyone to poke around in the code and reconfigure things as they wish. Low costs of development and efficient use of hardware seem to have afforded DeepSeek this cost advantage, and have already forced some Chinese rivals to lower their prices. Consumers should anticipate lower costs from other AI services too. Artificial investment Longer term - which, in the AI industry, can still be remarkably soon - the success of DeepSeek could have a big impact on AI investment. This is because so far, almost all of the big AI companies - OpenAI, Meta, Google - have been struggling to commercialise their models and be profitable. Until now, this was not necessarily a problem. Companies like Twitter and Uber went years without making profits, prioritising a commanding market share (lots of users) instead. And companies like OpenAI have been doing the same. In exchange for continuous investment from hedge funds and other organisations, they promise to build even more powerful models. These models, the business pitch probably goes, will massively boost productivity and then profitability for businesses, which will end up happy to pay for AI products. In the mean time, all the tech companies need to do is collect more data, buy more powerful chips (and more of them), and develop their models for longer. But this costs a lot of money. Nvidia's Blackwell chip - the world's most powerful AI chip to date - costs around US$40,000 per unit, and AI companies often need tens of thousands of them. But up to now, AI companies haven't really struggled to attract the necessary investment, even if the sums are huge. DeepSeek might change all this. By demonstrating that innovations with existing (and perhaps less advanced) hardware can achieve similar performance, it has given a warning that throwing money at AI is not guaranteed to pay off. For example, prior to January 20, it may have been assumed that the most advanced AI models require massive data centres and other infrastructure. This meant the likes of Google, Microsoft and OpenAI would face limited competition because of the high barriers (the vast expense) to enter this industry. Money worries But if those barriers to entry are much lower than everyone thinks - as DeepSeek's success suggests - then many massive AI investments suddenly look a lot riskier. Hence the abrupt effect on big tech share prices. Shares in chipmaker Nvidia fell by around 17% and ASML, which creates the machines needed to manufacture advanced chips, also saw its share price fall. (While there has been a slight bounceback in Nvidia's stock price, it appears to have settled below its previous highs, reflecting a new market reality.) Nvidia and ASML are "pick-and-shovel" companies that make the tools necessary to create a product, rather than the product itself. (The term comes from the idea that in a goldrush, the only person guaranteed to make money is the one selling the picks and shovels.) The "shovels" they sell are chips and chip-making equipment. The fall in their share prices came from the sense that if DeepSeek's much cheaper approach works, the billions of dollars of future sales that investors have priced into these companies may not materialise. For the likes of Microsoft, Google and Meta (OpenAI is not publicly traded), the cost of building advanced AI may now have fallen, meaning these firms will have to spend less to remain competitive. That, for them, could be a good thing. But there is now doubt as to whether these companies can successfully monetise their AI programmes. US stocks make up a historically large percentage of global investment right now, and technology companies make up a historically large percentage of the value of the US stock market. Losses in this industry might force investors to sell off other investments to cover their losses in tech, leading to a whole-market downturn. And it shouldn't have come as a surprise. In 2023, a leaked Google memo warned that the AI industry was exposed to outsider disruption. The memo argued that AI companies "had no moat" - no protection - against rival models. DeepSeek's success may be the proof that this is true.

DeepSeek gives Europe's tech firms a chance to catch up in global AI race

DeepSeek's emergence is changing the landscape for AI, offering companies access to the technology at a fraction of the cost, according to interviews with more than a dozen startup executives and investors. It also has the potential to push other AI companies to improve their models and bring down prices.Hemanth Mandapati, boss of German startup Novo AI, was an early adopter of DeepSeek chatbots when he switched to the Chinese AI model from OpenAI's ChatGPT two weeks ago. "If you have built your application using OpenAI, you can easily migrate to the other ones ... it took us minutes to switch," he said in an interview on the sidelines of the GoWest conference for venture capitalists in Gothenburg, Sweden. DeepSeek's emergence is changing the landscape for AI, offering companies access to the technology at a fraction of the cost, according to interviews with more than a dozen startup executives and investors. It also has the potential to push other AI companies to improve their models and bring down prices. "There was an offer from DeepSeek which was five times lower than their actual prices," said Mandapati. "I am saving a lot of money and users don't see any kind of a difference." Europe's tech startups had struggled to adopt the new technology at the same rate as U.S. rivals, which have easier access to funding. But executives say DeepSeek could be a game changer. "It marks a significant step forward in democratising AI and levelling the playing field with Big Tech," said Seena Rejal, chief commercial officer of British firm NetMind.AI, another early adopter of DeepSeek. Analysts at Bernstein estimate that DeepSeek's pricing is 20 to 40 times cheaper than equivalent models from OpenAI. OpenAI charges $2.5 for 1 million input tokens, or units of data processed by the AI model, while DeepSeek is currently charging $0.014 for the same number of tokens. Concerns have been raised by regulators about whether DeepSeek is copying OpenAI data or censoring answers that could portray China in a bad light. It is currently being investigated in different European countries. "While the future of DeepSeek as a business is difficult to predict, the structural impact seems quite pervasive," said Sanjot Malhi, partner at venture capital firm Northzone. Wake-up call Nearly $100 billion was invested by venture capitalists in AI companies in 2024 in the U.S. compared with about $15.8 billion in Europe, according to data from PitchBook. Just on Jan. 22, U.S. President Donald Trump unveiled a $500 billion AI project called Stargate, a joint venture backed by OpenAI, SoftBank and Oracle. Investment in Europe has been more modest. Only France's Mistral features among the list of top foundational models dominated by the likes of OpenAI, Meta , Anthropic and Google. China's DeepSeek attracted attention after writing in a paper last month that the training of DeepSeek-V3 required less than $6 million worth of computing power from Nvidia H800 chips. It has since overtaken ChatGPT to become the top-rated productivity application available on Apple's App Store. "This is a wake-up call that bigger isn't always better," said Fabrizio Del Maffeo, CEO of Axelera AI. "By making models more attainable to everyone, the total cost of ownership and barriers to building innovative tech are lowered which can be a catalyst for the whole industry." While some analysts doubt that DeepSeek's training cost is as low as the company says, they agree it is lower than comparable American models. "I see DeepSeek as a tremendous opportunity for companies like ours," said Ulrik R-T, CEO of Denmark's Empatik AI. "It showed that we do not need huge budgets to be able to achieve our vision." Cost vs safety The price war may have already started. Microsoft last week released OpenAI's o1 reasoning model to all Copilot users for free, instead of the usual subscription fee of $20 per month. "AI prices are going down, so future usage is probably going where there is transparency, which is usually open source, even though it's in China," said Scale Capital's Joachim Schelde. Bigger companies, ranging from Finland's Nokia to Germany's SAP, are more cautious about switching. "Cost is just one factor," said Alexandru Voica, Head of Corporate at Britain's Synthesia, which was last valued at $2.1 billion. "Other factors are: 'do you have all the security certifications, the frameworks, the software ecosystem that allows companies to build and integrate with your platform?'"

How DeepSeek Could Really Disrupt Big Tech

The Chinese chatbot has already hit the chipmaker giant Nvidia's share price, but its true potential could upend the whole AI business model. Only rarely does a single company's new product provoke a major market sell-off. But that's exactly what happened on Monday, when a large language model from a Chinese company named DeepSeek drove the entire Nasdaq index of tech companies down more than 3 percent and shaved more than 17 percent off the market capitalization of the chipmaker Nvidia -- which, until that moment, had been the most valuable company in the world. The panicked selling of Nvidia had a surface logic. The company provides almost all of the computer chips (called GPUs) that companies such as Alphabet, OpenAI, Microsoft, and Meta rely on to train their LLMs. (The Atlantic entered into a corporate partnership with OpenAI in 2024.) Consequently, it has been the biggest beneficiary of the huge boom in corporate spending on AI that we've seen over the past few years. (Nvidia's annual revenue has quadrupled since 2022.) Although DeepSeek also used Nvidia chips to train its model, the company said that they were an older type of GPU -- U.S. export controls imposed by the Biden administration have prevented Chinese companies from buying cutting-edge chips. DeepSeek's disclosure raises the possibility that future progress in training LLMs could be made with fewer, simpler chips, and at a lower cost than previously anticipated. That would obviously put a big dent in Nvidia's profits. So investors dumped its stock. Read: The DeepSeek wake-up call If investors are very concerned about how DeepSeek might hurt chipmakers, they seem surprisingly unconcerned about how it might affect big AI software companies. Meta's stock price, for instance, actually rose on Monday, and although the stocks of Alphabet and Microsoft did take a hit, they bounced back over the next couple of days. Some of that is because the underlying business of these companies, independent of AI, remains enormously profitable. But it also suggests that investors aren't paying enough attention to the way DeepSeek's success could disrupt the AI market, and in doing so threaten the future profits of the tech companies that are currently spending many billions of dollars every year on their LLMs. Tech investors have historically profited by spotting the new new thing. But at the moment, they seem implicitly to assume that all of the fundamental change in the LLM business has already happened and that its future will look much like its present, with the companies that currently dominate the space -- many of which are not simply competitors but also financial partners -- continuing to do so indefinitely. What happened over the past week is a reminder that these assumptions may not be so solid. The large language model that caused such a stir on Monday, DeepSeek-R1, is clearly comparable with LLMs such as ChatGPT o1-mini and Claude 3.5. Measured by industry benchmarks that rate subject knowledge, reasoning, and accuracy, the DeepSeek model seems to deliver similar performance while costing much less to develop -- though just how much less remains a matter of debate. Beyond dispute is that it's cheaper to use: Consumers can get access to DeepSeek's core functions for free, and third-party developers are being charged a fraction of the cost of a product such as ChatGPT. DeepSeek also uses open-source technology, meaning that, in theory, you could download the program and run your own AI on your desktop if you had a powerful-enough computer. The fact that the LLM offers reasonable performance -- results that, even a year ago, would have seemed startlingly good -- at a significantly lower cost means that it has to be taken seriously as a competitor. From one angle, in fact, DeepSeek looks like what the business-school professor Clayton Christensen, in his book The Innovator's Dilemma, dubbed a "disruptive technology": a product that's less powerful than the products at the top of the market but also much cheaper, and that has the possibility of improving in quality over time to the point where it offers a superior combination of price and performance for most customers. In this regard, the rapid uptake of DeepSeek by users around the world has been striking. The LLM still has miles to go in market share to catch ChatGPT, which has more than 300 million weekly users, but since its release on January 20, its mobile-app version has been downloaded more than 3 million times from Google Play and Apple, making it the most popular app on both stores. That suggests that the cost of switching from one AI tool to another is very low, and that the moats big AI companies are building around their business may be much shallower than they'd hoped. Read: China's DeepSeek surprise The underlying wager that these companies have made is that the big money they're investing will result in radically better performance, which in turn will enable them to charge hefty sums to businesses and, to a lesser extent, consumers. (OpenAI, for instance, is reportedly targeting $100 billion in revenue by 2029.) And these companies remain committed to that bet. This week, the CEOs of both Microsoft and Meta said that enormous spending is essential to staying competitive in the market. Dario Amodei, a co-founder and the CEO of Anthropic (in which both Amazon and Google have invested heavily), wrote in a blog post that companies are going to continue to "spend more and more on training powerful AI models, even as ... the cost of training a given level of model intelligence declines rapidly," because "the economic value of training more and more intelligent models is so great." In the long run, such investment may well result in the kind of performance improvement that a company like DeepSeek (which can't even get access to the most powerful GPUs) -- or the many other low-cost LLM developers that are sure to try to emulate it -- cannot keep up with. When you look at ordinary users' embrace of DeepSeek, though, you can also see an alternative future. In this one, AI performance improves so much that most customers are happy with cheap, good-enough LLMs, and AI models end up as essentially interchangeable, commoditized products, with the small profits that always follow that type of commercial diffusion. We're going to find out whether the great authors of the disruptive technology that's transforming the business world might themselves get disrupted.

How DeepSeek is upending AI innovation and investment

The tech world experienced a costly and highly consequential wake-up call this week with the revelation that Chinese newcomer DeepSeek had developed an advanced AI model requiring just millions -- rather than billions -- of dollars in development costs. Despite concerns about DeepSeek security and that it possibly copied rival ChatGPT, the news sent US AI leaders reeling, causing them to lose more than $1 trillion in total market value -- including nearly $600 billion from chip king Nvidia alone. The DeepSeek news also raised the possibility that the existing model of AI investment and development might soon be ready for a rethink. Indeed, rather than a bastion of small-scale disrupters and entrepreneurs, the vast majority of AI funding comes from tech giants such as Microsoft, Alphabet and Meta. It's the "tech-industrial complex" former President Joe Biden warned in his farewell address last week, lorded over by an emerging oligarchy that he believes has grown so powerful it threatens basic institutions and perhaps even democracy itself. Today's Internet is dominated by giant technology companies like Google, Facebook, Apple, Microsoft, and Amazon. Taken together, these digital conglomerates are worth $12 trillion, lording over the US stock market, and earning the moniker "The Magnificent Seven." And even as DeepSeek threatens to upend their reign, "The Magnificent Seven" has staked much of their fortunes' future on AI. Should we be worried? "If Big Tech continues to dominate the AI era, we risk [cultivating a culture] where users are products and their data is the most valuable commodity," says Tom Serres, co-founder of Nautilus Asset Management, referring to Big Tech's insatiable appetite for data, which it uses to target ads at users. AI may appear to be at the forefront of innovation and opportunity, but it is mostly funded by nation-scale investments from legacy tech giants. Microsoft recently announced that it will spend $80 billion on AI data centers this year, more than the United Kingdom's entire defense budget. It has also invested billions in OpenAI, maker of ChatGPT. Over at Google, Demis Hassabis, CEO of Google's AI-focused DeepMind division said the company will spend more than $100 billion to develop AI technology. Amazon is developing its own AI chips and has already invested $8 billion into ChatGPT competitor Anthropic while funneling billions into its own data center buildout. Facebook owner Meta recently projected that it would invest $35-40 billion in AI and its metaverse arm Reality Labs, including billions on chips made by Nvidia. Apple's spending spree saw it acquire DarwinAI, WaveOne, and dozens of other companies in the past few years. Elon Musk, CEO of Tesla, said that any company that doesn't spend a minimum of $10 billion a year on AI like Tesla won't be able to compete. Big Tech even got a boost from the new Trump administration, which announced a $500 billion AI "moonshot" initiative backed by several of these companies. Only governments can come close to matching Big Tech's largesse. To wit, sovereign wealth funds from Saudi Arabia, and the Gulf Countries have joined the party, plowing billions into high-profile AI deals. If these companies can buy up all the key hardware, vacuum up the best people, and leverage their market position in search, social media, e-commerce, and robotics to cross-sell their AI products, can anyone else compete? A splashy Beijing-backed upstart like DeepSeek, perhaps, but there are only so many splashy upstarts -- and so many governments with the resources and political motivations like the one in China. Perhaps Western governments should enact new regulations to tame the tech behemoths. But government intervention alone will not solve the problem of tech dominance, and may even make it worse. By codifying new laws under the guise of "AI safety," for instance, Washington may increase legal and regulatory burdens and make it harder for smaller companies to compete. AI needs those smaller companies to evolve into larger companies to fuel ongoing innovation cycles. Don't forget: "The big players we think of as having massive advantages in the age of AI were themselves upstarts not so long ago that took on the incumbents of their day," says Douglas Heintzman, chief catalyst of the BRI, a think-tank. By investing so heavily in AI today, companies like Meta and Microsoft are actually lowering the barrier for others to compete and spurring future innovation. Big Tech so heavily dominates AI investment because the raw computing power needed to build AI models like ChatGPT was traditionally thought of as scarce and expensive, pricing out smaller players, says Serres, of Nautilus Asset Management. But that seeming lead is illusory. DeepSeek, after all, was reportedly trained and built with only a $6 million investment, a far cry from the billions of dollars many assumed necessary to achieve an AI model that can match or even exceed ChatGPT from OpenAI. The DeepSeek model is open source, meaning anyone can audit the code and build on top of it. Tech entrepreneurs and venture capitalists are applauding DeepSeek while cautioning that it reveals China is far more advanced than we thought in AI. Legendary venture capitalist Marc Andreessen described the arrival of DeepSeek as AI's "Sputnik Moment." The launch is causing tremors across Big Tech; DeepSeek's debut caused the stock of AI chipmaker Nvidia to crash more than 15% in a single trading day, and now analysts are questioning whether Big Tech was just overspending on AI -- throwing money at a problem without understanding it at a deeper level. Joseph Geraci, founder and both chief strategy and technology officer at NetraMark, adds that the AI that "currently dominates cannot be trained on consumer-level computers," instead requiring hundreds of supercomputers known as "GPUs" that can cost $40,000 a pop. DeepSeek's showstopper announcement proves that deep AI work can be done with far less equipment and financial outlay. Darwin.AI CEO Sheldon Fernandez says "AI can be used as a substitute for human creativity and logical reasoning," posing both a risk and opportunity to many professions and opening new doors for start-up founders. Mark Zuckerberg recently admitted as much, saying AI can perform like a "mid-level" software engineer. How much longer until they can perform better than the best? Indeed, recent reports surfaced that OpenAI was working on an AI "agent" with PhD level intelligence. Sam Altman, CEO of OpenAI, once said AI would enable the first "billion dollar one person company" with AI handling much of the workload such as finance, marketing and logistics. What's more, thanks to AI, entrepreneurs may not even need to know how to code. Your AI "programmer" can do it for you as the industry increasingly "lower[s] the barrier to value creation," says Heintzman. All of this will require vast sums of investment, though now with the arrival of DeepSeek, the industry's financing model may be ready for a rethink. Startups can also succeed by specializing, leveraging the tech built by others, and creating new applications. Fernandez of Darwin.AI says the best new startups will "train and augment core models in highly specific and technical ways to achieve their goals." The result will be AI startups for law, engineering, construction, and countless other industries. This raises a separate issue of platform risk, where building on someone else's technology, whether their AI model or their cloud, makes you beholden to them. It is the technology behind Bitcoin, called blockchain, however, that promises to shake up Big Tech's platform investment economics once and for all. Indeed, it is often the combination, or convergence, of two or more technologies that have the biggest impact -- and bang for the investment buck. Consider how smartphones combined with wireless networks and GPS led to mobile apps, location-based services, and more. Today it may be AI and crypto. This plays out in a few ways. Most importantly, in crypto, users can pool their computers together and make them available to the public as "decentralized clouds," that can compete with centralized systems, such as Microsoft Azure or Amazon Web Services. This will increase access and lower costs for developers who want to train and run AI models. Newer and cheaper AI models will increase the number of AI agents, who, since they can't open a bank account, must use crypto to do transactions. The technology industry is in a constant state of reinvention. Clunky mainframe computers gave way to the PC, which ushered in the Internet era, and the mobile web. Today, AI is shaking the windows and rattling the walls of incumbent technology companies. It remains to be seen whether DeepSeek will be the company that upends the status quo with its low cost development and investment model. After all, DeepSeek is small, but it has China at its back -- which is both a boon and a blessing. As financial columnist Charles Gasparino noted this week, "I'm skeptical about the DeepSeek threat. I'm not saying it's a deep-fake, but I just don't trust anything that comes out of China." The fight for the AI future will be competitive, but one thing we can count on is change.

AI is not just powerful. What's really worrying is that DeepSeek has made it cheap, too | John Naughton

The AI startup has upended the industry by developing a model that costs much less to produce - and is available free to a universe of tinkerers Nothing cheers up a tech columnist more than the sight of $600bn being wiped off the market cap of an overvalued tech giant in a single day. And yet last Monday that's what happened to Nvidia, the leading maker of electronic picks and shovels for the AI gold rush. It was the biggest one-day slump for any company in history, and it was not alone - shares of companies in semiconductor, power and infrastructure industries exposed to AI collectively shed more than $1tn in value on the same day. The proximate cause of this chaos was the news that a Chinese tech startup of whom few had hitherto heard had released DeepSeek R1, a powerful AI assistant that was much cheaper to train and operate than the dominant models of the US tech giants - and yet was comparable in competence to OpenAI's o1 "reasoning" model. Just to illustrate the difference: R1 was said to have cost only $5.58m to build, which is small change compared with the billions that OpenAI and co have spent on their models; and R1 is about 15 times more efficient (in terms of resource use) than anything comparable made by Meta. The DeepSeek app immediately zoomed to the top of the Apple app store, where it attracted huge numbers of users who were clearly unfazed by the fact that the terms and conditions and the privacy policy they needed to accept were in Chinese. And it clearly energised the Silicon Valley crowd. "DeepSeek R1," boomed venture capitalist Marc Andreessen, one of the loudest mouths in California, "is AI's Sputnik moment". He also called it "one of the most amazing and impressive breakthroughs I've ever seen - and as open source, a profound gift to the world". Donald Trump, who does not believe in giving gifts to the world, described R1 as a "wake-up call" for American tech firms. Historical resonances were rife. Andreessen was referring to the seminal moment in 1957 when the Soviet Union launched the first Earth satellite, thereby displaying technological superiority over the US - a shock that triggered the creation of Nasa and, ultimately, the internet. Other people were reminded of the advent of the "personal computer" and the ridicule heaped upon it by the then giants of the computing world, led by IBM and other purveyors of huge mainframe computers. Suddenly, people are beginning to wonder if DeepSeek and its offspring will do to the trillion-dollar AI behemoths of Google, Microsoft, OpenAI et al what the PC did to IBM and its ilk. And of course there are the conspiracy theorists wondering whether DeepSeek is really just a disruptive stunt dreamed up by Xi Jinping to unhinge the US tech industry. Is the model really that cheap to train? Can we believe the numbers in the technical reports published by its makers? And so on. Standing back, there are four things to take away from the arrival of DeepSeek. The first is that China has caught up with the leading US AI labs, despite the widespread (and hubristic) western assumption that the Chinese are not as good at software as we are. Even a cursory examination of some of the technical details of R1 and the V3 model that lay behind it evinces formidable technical ingenuity and creativity. Second, the low training and inference costs of R1 will turbocharge American anxiety that the emergence of powerful - and cheap - Chinese AI could upend the economics of the industry, much as the advent of the PC transformed the computing marketplace in the 1980s and 90s. What the advent of DeepSeek indicates is that this technology - like all digital technology - will eventually be commoditised. R1 runs on my laptop without any interaction with the cloud, for example, and soon models like it will run on our phones. Third, DeepSeek pulled this off despite the ferocious technology bans imposed by the first Trump administration and then by Biden's. The company's technical report shows that it possesses a cluster of 2,048 Nvidia H800 GPUs - technology officially banned by the US government for sale to China. And last, but by no means least, R1 seems to be a genuinely open source model. It's distributed under the permissive MIT licence, which allows anyone to use, modify, and commercialise the model without restrictions. As I write this, my hunch is that geeks across the world are already tinkering with, and adapting, R1 for their own particular needs and purposes, in the process creating applications that even the makers of the model couldn't have envisaged. It goes without saying that this has its upsides and downsides, but it's happening. The AI genie is now really out of the bottle. When Trump meets tech A really sobering analysis by William Cullerne Bown of what the new regime in Washington means for the UK and Europe. A dystopia like Philip K Dick's An essay explaining why Henry Farrell thinks that our future might be like something written by the great author.

DeepSeek Shock Could Spur AI Space Race

Last week America's tech billionaires were literally centre stage at Donald Trump's presidential inauguration. The message was clear: the US government wants to parade them in front of the world as the face of modern America. It also shows they have unparalleled influence over the incoming government. These are some of the richest people in the world: Musk, Zuckerberg, Bezos and Apple CEO Tim Cook were all visible. Other tech leaders had background meetings with the president in the run up to the inauguration. The day after his inauguration, President Trump announced a US$500 billion private sector joint venture to build a new generation of AI data centres called Stargate. He said the project would keep the US ahead of China in the global race for supremacy in artificial intelligence. Fast forward a week, and America's place in that race was in question. The share prices of many US tech giants plummeted. Nvidia alone lost more shareholder value than any firm in history. It dropped 17 percent which is around a trillion New Zealand dollars. The cause of the panic is DeepSeek, a Chinese AI chatbot. DeepSeek upset the market because it cost a tiny fraction of the amount of money the US companies spent developing their AI systems. Around 5 percent -- or one-twentieth -- of ChatGPT's development cost. The two are not comparable, but it cost $6 million compared with the $500 billion investment announced by the president. Only one of those two made front page news around the world. There has been some questioning of DeepSeek's investment, but niggling over details changes little. The Chinese build an AI that is on a par with US AI, but cost less and works with fewer resources. Deepseek is a free download and it was the world's most downloaded app last week. It is also Open Source, which undermines OpenAI which, despite its name, is not Open Source. There is another reason DeepSeek took the US tech sector by surprise. America cut off the supply of the latest AI chips to China as a way of keeping that competitive advantage the president refers to. It's an oversimplification, but many in the US assumed Chinese developers were lagging far behind their American counterparts. America see AI as one of the few remaining areas where that nation leads the world. At the same time, US companies have a particular way of looking at AI development. We could call it a brute force approach to AI or we could think of it as a 'bigger is better' philosophy. This approach requires building ever-larger data centres with ever more processing power, which, in turn, demands more energy to power the processing. It is why new generating capacity is being built and old power plants are being fired up once more. You need a lot of electricity to draw your AI cat pictures or to make up the incorrect facts that AI often serves up. You also need it to do the real, worthwhile work that AI is sometimes capable of. This is a little like the story of the tortoise and the hare. The apparent plodder may not have got to the finish line first, but it is close behind. In effect the Chinese developers squeezed the older generation chips it could access harder in order to get more out of them. Almost overnight they proved they are almost on a par with the US. DeepSeek isn't as powerful as the leading US AI systems, but it comes close and has the potential to close the gap. There's one more angle we need to talk about. The US tech giants could struggle to make their AI investments pay off. Customers are not buying as much of it as the tech companies hoped. Recently Microsoft secretly hiked the consumer price of a subscription to its Office software to include its lame AI offering. The company could not get customers to pay separately for it. Why would you, Co-Pilot performs poorly by the standards of its rivals. Customers can opt out, but must dive deep into the settings to do so. Tech giants have already spent tens of billions, they are building data centres, including the new data centres in New Zealand, to ride what they see as a profitable AI future, but DeepSeek suggests they may find it even harder to make their AI investments pay off in the short or medium term. "DeepSeek could potentially burst the AI bubble and challenge the US's dominance in tech innovation." Venture capitalist Marc Andreessen said "DeepSeek is AI's Sputnik moment". In the 1950s the Russians shocked America by putting a satellite into orbit before the US. That sparked the space race. This is why we should celebrate the arrival of DeepSeek despite its shortcomings. The censorship is worrying, but all AIs come with a degree of editing. History tells us that technology advances best and delivers the most benefits to people when competition spurs innovation. The competition between US-based tech giants is one thing, but a competitor from China working from a different philosophical base, taking an alternative approach has to be a welcome addition. Taking the initiative out of the hands of one nation, or one group of tech giants will be refreshing. DeepSeek could provide the impetus for bright entrepreneurs elsewhere in the world to seek out separate fresh AI paths. What's more, DeepSeek makes better uses of energy and other resources. Competition in this area is also welcome. It could slow or even stop moves to reopen coal mines or drill more oil to generate power. We need to be wary of DeepSeek in particular and AI in general, there are many risks with both, but by showing there is another way, DeepSeek has done everyone a favour.

DeepSeek gives Europe's tech firms a chance to catch up in global AI race

GOTHENBURG (Reuters) - Hemanth Mandapati, boss of German startup Novo AI, was an early adopter of DeepSeek chatbots when he switched to the Chinese AI model from OpenAI's ChatGPT two weeks ago. "If you have built your application using OpenAI, you can easily migrate to the other ones ... it took us minutes to switch," he said in an interview on the sidelines of the GoWest conference for venture capitalists in Gothenburg, Sweden. DeepSeek's emergence is changing the landscape for AI, offering companies access to the technology at a fraction of the cost, according to interviews with more than a dozen startup executives and investors. It also has the potential to push other AI companies to improve their models and bring down prices. "There was an offer from DeepSeek which was five times lower than their actual prices," said Mandapati. "I am saving a lot of money and users don't see any kind of a difference." Europe's tech startups had struggled to adopt the new technology at the same rate as U.S. rivals, which have easier access to funding. But executives say DeepSeek could be a game changer. "It marks a significant step forward in democratising AI and levelling the playing field with Big Tech," said Seena Rejal, chief commercial officer of British firm NetMind.AI, another early adopter of DeepSeek. Analysts at Bernstein estimate that DeepSeek's pricing is 20 to 40 times cheaper than equivalent models from OpenAI. OpenAI charges $2.5 for 1 million input tokens, or units of data processed by the AI model, while DeepSeek is currently charging $0.014 for the same number of tokens. Concerns have been raised by regulators about whether DeepSeek is copying OpenAI data or censoring answers that could portray China in a bad light. It is currently being investigated in different European countries. "While the future of DeepSeek as a business is difficult to predict, the structural impact seems quite pervasive," said Sanjot Malhi, partner at venture capital firm Northzone. WAKE-UP CALL Nearly $100 billion was invested by venture capitalists in AI companies in 2024 in the U.S. compared with about $15.8 billion in Europe, according to data from PitchBook. Just on Jan. 22, U.S. President Donald Trump unveiled a $500 billion AI project called Stargate, a joint venture backed by OpenAI, SoftBank and Oracle. Investment in Europe has been more modest. Only France's Mistral features among the list of top foundational models dominated by the likes of OpenAI, Meta, Anthropic and Google. China's DeepSeek attracted attention after writing in a paper last month that the training of DeepSeek-V3 required less than $6 million worth of computing power from Nvidia H800 chips. It has since overtaken ChatGPT to become the top-rated productivity application available on Apple's App Store. "This is a wake-up call that bigger isn't always better," said Fabrizio Del Maffeo, CEO of Axelera AI. "By making models more attainable to everyone, the total cost of ownership and barriers to building innovative tech are lowered which can be a catalyst for the whole industry." While some analysts doubt that DeepSeek's training cost is as low as the company says, they agree it is lower than comparable American models. "I see DeepSeek as a tremendous opportunity for companies like ours," said Ulrik R-T, CEO of Denmark's Empatik AI. "It showed that we do not need huge budgets to be able to achieve our vision." Microsoft last week released OpenAI's o1 reasoning model to all Copilot users for free, instead of the usual subscription fee of $20 per month. "AI prices are going down, so future usage is probably going where there is transparency, which is usually open source, even though it's in China," said Scale Capital's Joachim Schelde. Bigger companies, ranging from Finland's Nokia to Germany's SAP, are more cautious about switching. "Cost is just one factor," said Alexandru Voica, Head of Corporate at Britain's Synthesia, which was last valued at $2.1 billion. "Other factors are: 'do you have all the security certifications, the frameworks, the software ecosystem that allows companies to build and integrate with your platform?'" (Reporting by Supantha Mukherjee in Stockholm and Toby Sterling in Copenhagen; Editing by Matt Scuffham and Louise Heavens)

DeepSeek means a rethink on AI investment

But Chinese startup shakeup doesn't herald 'drastic drop' in need for infrastructure buildout, say analysts Analysis The shockwave following the release of competitive AI models from Chinese startup DeepSeek has led many to question the assumption that throwing ever more money at costly large-scale GPU-based infrastructure delivers the best results. As The Register reported earlier, shares of some of the largest American tech brands in the AI boom tumbled following the release of the DeepSeek R1 model, which is said to perform favorably against those from OpenAI and Meta and has been trained using fewer Nvidia GPUs. The China-based company's claims that DeepSeek's performance is on par with the best existing models and that it cost less than $6 million to train are also unverified The move called into question the assumption that spending billions on infrastructure in a race to build larger and more complex models is the way forward if China can do this with limited supplies of older hardware. Nvidia, which has enjoyed record profits from its GPU accelerators for AI, lost almost $600 billion off its market valuation in response. This comes on top of a growing unease that more investment is being funneled into AI development and the infrastructure to support it, with little return to be seen so far. But the panic may have been misplaced as the freefall in US tech shares soon halted, and experts pointed out that DeepSeek appears to have used output from existing models developed by Anthropic and OpenAI in its training. The China-based company's claims that DeepSeek's performance is on par with the best existing models and that it cost less than $6 million to train are also unverified. "I believe the concerns regarding DeepSeek's innovations are highly overblown," Omdia's Principal Analyst for Datacenter IT, Manoj Sukumaran, told The Register. "There is no doubt that there are some ingenious innovations in DeepSeek's model pre-training, like the use of reinforcement learning as a core training methodology, moving away from the reliance on large labeled datasets, sparse activation of model parameters, and adaptive routing to select the expert models to work," he said. But these innovations are essential to make GenAI accessible to more users, Sukumaran added, and will instead hasten user adoption of this technology. As far as the infrastructure to power all this goes, Sukumaran says that massive AI buildouts are likely to continue. "The AI inference market is just unfolding and it will grow significantly over the next several years. Omdia has estimated that the number of servers shipped each year for AI inference will increase at a 17 percent CAGR out to 2028," he added. Nevertheless, Taiwan-based research operation TrendForce says that it expects to see organizations conduct more rigorous evaluations of AI infrastructure investments in future, and focus on adopting more efficient models to reduce reliance on hardware such as GPUs. The firm also says it envisages growth in the adoption of infrastructure using custom ASICs (application-specific integrated circuits) to lower deployment costs, and that demand for GPU-based products could see "notable changes" from 2025 onward. "Historically, the AI industry has relied on scaling models, increasing data volume, and enhancing hardware performance for growth. However, escalating costs and efficiency challenges have prompted a shift in strategy," TrendForce states. "DeepSeek has adopted model distillation techniques to compress large models, improve inference speed, and reduce hardware dependencies." Meanwhile, IBM CEO Arvind Krishna saw in DeepSeek some validation for his own company's approach to AI. "We have been very vocal for about a year that smaller models and more reasonable training times are going to be essential for enterprise deployment of large language models. We have been down that journey ourselves for more than a year," Krishna claimed during the company's recent earnings call. "We see as much as 30 times reduction in inference costs using these approaches. As other people begin to follow that route, we think that this is incredibly good for our enterprise clients. And we will certainly take advantage of that in our business, but I believe that others will also follow that route." In a note on the implications of DeepSeek issued by Gartner, the analyst biz said that efficient scaling of AI will in future be more important than how much compute can be assembled to build it. "DeepSeek-engineered systems combine models, frameworks, and underlying infrastructures to more effectively utilize infrastructure resources. This results in lower costs while delivering efficiencies," Gartner observed. However, the firm said that the Chinese AI doesn't set a new state of the art for model performance as it often matches but doesn't surpass existing models. As far as infrastructure goes, Gartner says that "it's not proof that scaling models via additional compute and data doesn't matter, but that it pays off to scale a more efficient model." The takeaway is that DeepSeek isn't going to suddenly lead to a drastic drop in demand for AI infrastructure, so Nvidia investors and those pumping money into datacenters can rest a bit easier. Neither is it the harbinger of the AI bubble bursting that many keep expecting. Instead, it serves as a reminder that things can always be done better, and that just throwing money and resources at a problem is not always the best way to solve it. "DeepSeek's superior price-to-performance ratio serves as a reality check for the AI industry, particularly US companies and their venture capital backers," said Neil Roseman, CEO of security firm Invicti. "While companies make massive bets on AI, current results don't justify these investments. Success will come from efficient, focused development addressing genuine needs." ®

The Chinese AI companies that could match DeepSeek's impact

DeepSeek roiled global markets and shocked Silicon Valley, but it's just one of a raft of Chinese AI companies that could challenge US dominance DeepSeek's release of an artificial intelligence model that could replicate the performance of OpenAI's o1 at a fraction of the cost has stunned investors and analysts. Markets reeled as Nvidia, a microchip and AI firm, shed more than $500bn in market value in a record one-day loss for any company on Wall Street. Investors feared that DeepSeek challenged the dominance of US AI leaders. Donald Trump described DeepSeek as a "wake-up call". In China, DeepSeek's founder, Liang Wenfeng, has been hailed as a national hero and was invited to attend a symposium chaired by China's premier, Li Qiang. The pace at which China has been able to catch up with frontier AI research in the US is accelerating. But DeepSeek is not the only Chinese company to have innovated despite the embargo on advanced US technology. Matt Sheehan, a fellow at the Carnegie Endowment for International Peace and an expert on Chinese AI, said: "If the US government thinks all we need to do is crush DeepSeek and then we'll be OK, then we're in for a rude surprise." In recent weeks, other Chinese technology companies have rushed to publish their latest AI models, which they claim are on a par with those developed by DeepSeek and OpenAI. But what are the Chinese AI companies that could match DeepSeek's impact? On 29 January, the first day of the lunar new year holiday, leading Chinese technology company Alibaba Cloud, a subsidiary of Alibaba, released an updated version of its Qwen 2.5 AI model, called Qwen 2.5-Max. According to Alibaba Cloud, Qwen 2.5-Max outperforms DeepSeek V3 and Meta's Llama 3.1 across 11 benchmarks. The company said that it was "full of confidence in the next version of Qwen 2.5-Max". Some analysts said that the fact that Alibaba Cloud chose to release Qwen 2.5-Max just as businesses in China closed for the holidays reflected the pressure that DeepSeek has placed on the domestic market. But Sheehan said it may also have been an attempt to ride on the wave of publicity for Chinese models generated by DeepSeek's surprise. Zhipu is a Beijing-based start-up that is backed by Alibaba. Known as one of China's "AI tigers", it was in the headlines recently not for its AI achievements but for the fact that it was blacklisted by the US government. On 15 January, Zhipu was one of more than two dozen Chinese entities added to a US restricted trade list. Zhipu in particular was added for allegedly aiding China's military advancement with its AI development. Zhipu condemned the decision and said it lacked a factual basis. Allegations about military uplift aside, it is clear that Zhipu's progress in the AI space is rapid. Its most recent product is AutoGLM, an AI assistant app released in October, which helps users to operate their smartphones with complex voice commands. On the same day that DeepSeek released its R1 model, 20 January, another Chinese start-up released an LLM that it claimed could also challenge OpenAI's o1 on mathematics and reasoning. Moonshot AI is another Alibaba-backed AI start-up, based in Beijing and valued at $3.3bn. Unlike Alibaba, a behemoth that was founded in 1999, Moonshot AI is a relative newcomer. Like DeepSeek, it was founded in 2023. Its offering, Kimi k1.5, is the upgraded version of Kimi, which was launched in October 2023. It attracted attention for being the first AI assistant that could process 200,000 Chinese characters in a single prompt. Moonshot AI later said Kimi's capability had been upgraded to be able to handle 2m Chinese characters. Moonshot AI "is in the top echelons of Chinese start-ups", Sheehan said. "It wouldn't surprise me at all if Moonshot or Zhipu has a model that equals or comes close to DeepSeek in performance within the next weeks or months." Another lunar new year release came from ByteDance, TikTok's parent company. On 29 January it unveiled Doubao-1.5-pro, an upgrade to its flagship AI model, which it said could outperform OpenAI's o1 in certain tests. As well as performance, Chinese companies are challenging their US competitors on price. Doubao's most powerful version is priced at 9 yuan per million tokens, which is nearly half the price of DeepSeek's offering for DeepSeek-R1. For comparison, OpenAI's o1 costs the equivalent of 438 yuan for the same usage. Primarily known for gaming and WeChat, the ubiquitous messaging app, Tencent has also made strides in AI. Its flagship model is a text-to-video generator called Hunyuan, which Tencent said can perform as well as Meta's Llama 3.1. Some estimates suggest that Hunyuan required around a tenth of the computing power used by Meta to train Llama.

China's AI dark horse DeepSeek puts Western tech giants on the back foot

Predictions for what the low-cost Chinese LLM could mean for the market The news that an affordable artificial intelligence by DeepSeek, a Chinese AI startup, has outperformed ChatGPT developer OpenAI on several benchmarks has upended the global AI industry. Tech analysts in the US are closely watching the emergence of a new contender in the field while noting that DeepSeek's achievements epitomize the normal tendency for costs to fall with technological advancement. It has generally been assumed that AI development was a cash-intensive endeavor that required investments numbering in the tens of billions of dollars. But now it seems the industry has reached an inflection point. Dario Amodei, the co-founder and CEO of Anthropic, argued that DeepSeek, which exhibited an AI model on a level similar to those of Big Tech firms a step behind its competitors, can't be viewed as true innovation. In a blog post published on Wednesday, Amodel wrote: "Sonnet's training was conducted 9-12 months ago, and DeepSeek's model was trained in November/December, while Sonnet remains notably ahead in many internal and external evals." Amodei is referring to Claude 3.5 Sonnet, a mid-sized model from Anthropic. Amodei argued that according to historical trends, DeepSeek is actually in alignment with a normal cost-curve reduction, which is normally a reduction by a factor of four in one year. Amodei argues that DeepSeek is "worse than US frontier models" by a factor of two but that its training cost was reduced by a factor of eight compared to US models developed a year ago. This would equate to cost-curve reduction by a factor of four, which, as Amodei points out, is "on-trend at best and probably not even that." The training cost for the V3 iteration of DeepSeek, a large-language model (LLM) publicly revealed last month, was reportedly only US$5.57 million, a figure that made waves. Amodei explains, however, that Anthropic's model, which came out six months prior, cost US$10 million to train. Considering the six-month difference, the reduction in initial cost is a completely expected outcome. Big Tech firms create war rooms to respond Big tech firms, while struggling to maintain their poker faces amid the emergence of such unexpected competition, have been spurred into action after being shaken out of complacency by DeepSeek. Sam Altman, the CEO of OpenAI, posted on X (formerly Twitter) on Tuesday: "deepseek's r1 is an impressive model, particularly around what they're able to deliver for the price." Satya Nadella, the CEO of Microsoft, posted: "As AI gets more efficient and accessible, we will see its use skyrocket, turning it into a commodity we just can't get enough of." Despite their praise, both OpenAI and Microsoft are raising suspicions about DeepSeek developers collecting OpenAI data without authorization, and have set out to investigate. Reuters reported that American AI experts view DeepSeek as a "distillation," which is essentially when "one model learns from another model." Through distillation, these American experts argue, DeepSeek copied OpenAI's model while extracting its data without authorization to develop its own model. Meta, the parent company of Facebook, which has offered an open-source AI model similar to DeepSeek, reportedly set up a "war room" to draft a strategy for responding to the competitive threat posed by DeepSeek. Yet during an earnings call on Wednesday, Meta CEO Mark Zuckerberg was reportedly untroubled by the emergence of DeepSeek, despite others in the industry voicing concerns. Impact on AI semiconductor stocks Some people expect the Trump administration to implement export controls on semiconductors to prevent Chinese advances in AI technology. In addition to export controls on high-spec semiconductors, there are even considerations of restricting exports of low-spec semiconductors, such as Nvidia's H20 chip, to China. DeepSeek has revealed that it utilized around 2,000 H800 Nvidia chips in its development of DeepSeek V3, which was released last month. Amid reports that DeepSeek utilized low-spec semiconductors to develop AI technology on the same level as Big Tech firms, the stocks of firms that manufacture AI semiconductors, such as Nvidia and Broadcom, plummeted on Monday and have been volatile ever since. Shares of Nvidia, which controls around 70% of the AI chip market, made a recovery of 8.93% on Jan. 28 but fell 4.1% on the following day. By Sun Dam-eun, staff reporter; Jeong Nam-ku, staff reporter

DeepSeek: Everything you need to know about the AI chatbot app | TechCrunch

Chinese AI lab DeepSeek broke into the mainstream consciousness this week after its chatbot app rose to the top of the Apple App Store charts. DeepSeek's AI models, which were trained using compute-efficient techniques, have led Wall Street analysts -- and technologists -- to question whether the U.S. can maintain its lead in the AI race and whether the demand for AI chips will sustain. But where did DeepSeek come from, and how did it rise to international fame so quickly? DeepSeek is backed by High-Flyer Capital Management, a Chinese quantitative hedge fund that uses AI to inform its trading decisions. AI enthusiast Liang Wenfeng co-founded High-Flyer in 2015. Wenfeng, who reportedly began dabbling in trading while a student at Zhejiang University, launched High-Flyer Capital Management as a hedge fund in 2019 focused on developing and deploying AI algorithms. In 2023, High-Flyer started DeepSeek as a lab dedicated to researching AI tools separate from its financial business. With High-Flyer as one of its investors, the lab spun off into its own company, also called DeepSeek. From day one, DeepSeek built its own datacenter clusters for model training. But like other AI companies in China, DeepSeek has been affected by U.S. export bans on hardware. To train one of its more recent models, the company was forced to use Nvidia H800 chips, a less-powerful version of a chip, the H100, available to U.S. companies. DeepSeek's technical team is said to skew young. The company reportedly aggressively recruits doctorate AI researchers from top Chinese universities. DeepSeek also hires people without any computer science background to help its tech better understand a wide range of subjects, per The New York Times. DeepSeek unveiled its first set of models -- DeepSeek Coder, DeepSeek LLM, and DeepSeek Chat -- in November 2023. But it wasn't until last spring, when the startup released its next-gen DeepSeek-V2 family of models, that the AI industry started to take notice. DeepSeek-V2, a general-purpose text- and image-analyzing system, performed well in various AI benchmarks -- and was far cheaper to run than comparable models at the time. It forced DeepSeek's domestic competition, including ByteDance and Alibaba, to cut the usage prices for some of their models, and make others completely free. DeepSeek-V3, launched in December 2024, only added to DeepSeek's notoriety. According to DeepSeek's internal benchmark testing, DeepSeek V3 outperforms both downloadable, openly available models like Meta's Llama and "closed" models that can only be accessed through an API, like OpenAI's GPT-4o. Equally impressive is DeepSeek's R1 "reasoning" model. Released in January, DeepSeek claims R1 performs as well as OpenAI's o1 model on key benchmarks. Being a reasoning model, R1 effectively fact-checks itself, which helps it to avoid some of the pitfalls that normally trip up models. Reasoning models take a little longer -- usually seconds to minutes longer -- to arrive at solutions compared to a typical non-reasoning model. The upside is that they tend to be more reliable in domains such as physics, science, and math. There is a downside to R1, DeepSeek V3, and DeepSeek's other models, however. Being Chinese-developed AI, they're subject to benchmarking by China's internet regulator to ensure that its responses "embody core socialist values." In DeepSeek's chatbot app, for example, R1 won't answer questions about Tiananmen Square or Taiwan's autonomy. If DeepSeek has a business model, it's not clear what that model is, exactly. The company prices its products and services well below market value -- and gives others away for free. The way DeepSeek tells it, efficiency breakthroughs have enabled it to maintain extreme cost competitiveness. Some experts dispute the figures the company has supplied, however. Whatever the case may be, developers have taken to DeepSeek's models, which aren't open source as the phrase is commonly understood but are available under permissive licenses that allow for commercial use. According to Clem Delangue, the CEO of Hugging Face, one of the platforms hosting DeepSeek's models, developers on Hugging Face have created over 500 "derivative" models of R1 that have racked up 2.5 million downloads combined. DeepSeek's success against larger and more established rivals has been described as "upending AI" and ushering in "a new era of AI brinkmanship." The company's success was at least in part responsible for causing Nvidia's stock price to drop by 18% on Monday, and for eliciting a public response from OpenAI CEO Sam Altman. As for what DeepSeek's future might hold, it's not clear. Improved models are a given. But the U.S. government appears to be growing wary of what it perceives as harmful foreign influence.

Was this the week DeepSeek started the slow unwinding of the AI bet?

The cheap Chinese chatbot has stunned tech giants - and opened up the possibility that other countries, not just China, could now afford to enter the AI race At 2.16pm California time last Sunday, the US billionaire tech investor Marc Andreessen called it. "DeepSeek R1 is AI's Sputnik moment," he posted on X. A Chinese startup, operating since 2023 and helmed by a millennial mathematician, had unveiled a new chatbot that seemed to equal the performance of America's leading models at a fraction of the cost. Never mind that its answers on everything from the status of Taiwan to the 1989 Tiananmen Square massacre were curbed by Chinese Communist party (CCP) censors. To Andreessen, a veteran of decades of technology booms and busts, it was like the Soviet Union getting the first satellite into orbit in 1957 and shocking America. The next day, shares in several of the world's biggest companies plunged - including the biggest fall in US market history for microchip maker Nvidia, which lost nearly $600bn. Investors believed DeepSeek's achievement meant China would no longer need so many American chips; that US supremacy in AI was under threat or already over; and that the Silicon Valley giants, who had only a week earlier announced a $500bn AI investment plan, were spending much more money than they needed. The Chinese AI lab said the training cost for one of its base models had been just $5.6m. In the biggest week for AI since the launch of ChatGPT in November 2022, DeepSeek's app, with its jaunty blue whale logo, became the most downloaded free app on Apple's app stores in the US and UK as people rushed to find out what it was about. But was the world's largest autocratic nation about to leapfrog the west in AI? What might it mean for control of a technology that many fear could be pressed into malicious use in cyber-attacks, the production of biological weapons and thought control? And given AI is widely considered to now be one of the main playing fields of geopolitical competition, where did this leave US hopes of maintaining supremacy by suppressing China's progress with export bans on microchips that are key to progress? Tremors had been rumbling out of DeepSeek's laboratory in Hangzhou, outside Shanghai, for a while. Some experts had been quietly impressed by the developments overseen by DeepSeek's boss, Liang Wenfeng, a 40-year-old hedge fund entrepreneur. But it wasn't until last Wednesday that a proper earthquake hit. The firm published a 22-page paper unveiling the DeepSeek R1 model, boasting of "powerful and intriguing reasoning behaviours" and saying it is comparable to Open AI's 01 model, and even better in some areas. While Google, Meta and OpenAI typically swaddle their new releases in marketing hype, DeepSeek's matter-of-fact approach was clear from the soporific title of its announcement: "Incentivizing Reasoning Capability in LLMs via Reinforcement Learning". The model was free to use and it seemed pioneering in the way it was engineered to be more efficient than ChatGPT-o1, OpenAI's $20-a-month reasoning model. It used less computing power as it had been engineered only to activate the relevant part of the system to answer the query. Performance that cost other companies billions seemed to be available for millions. In response, OpenAI announced the launch of a new reasoning model, o3-Mini, on Friday that will be made available to all users, including people on ChatGPT's free tier. Liang was said to be on holiday for lunar new year as his team's creation upended not just markets, but also the geopolitical calculus between the US and China as they vie for supremacy in AI with all its economic, political and military potentials. Around the world, experts tried to make sense of how the Chinese had made necessity the mother of invention and found a way around a shortage of chips. Jimmy Goodrich, an adviser on technology to the Rand Corporation, told Reuters: "It's been long known that DeepSeek has a really good team, and if they had access to even more compute, God knows how capable they would be." "I confess I hadn't heard of them," said Michael Wooldridge, a professor of the foundations of AI at the University of Oxford. "[They] appear to have built something which is as capable as a GPT class model, not necessarily better, with something like a hundredth of the resources." He says the development "pulls the rug out from under Nvidia", meaning a far greater number of developers can build AI models, making it a "much more accessible technology". Mike Gualtieri, a principal analyst at Forrester Research, says that accessibility will widen the number of startups that can create their own AI models. But also, the bigger US tech players, with their considerable data processing firepower, could accelerate their own development. "The companies that already have a lot of chips, or access to them - the OpenAIs and the Googles - once they apply these [DeepSeek] techniques, they can experiment more rapidly," he said. In London, hopes and fears were in conflict. The technology secretary, Peter Kyle, said he would not download the Chinese app, surely aware that anything he typed in or uploaded would be stored in China and that all Chinese firms are obliged under the national intelligence law to "support, assist and cooperate" with intelligence efforts. But, as a minister tasked with using AI to deliver economic growth, he was "really excited" by the breakthrough. It seemed to show that skills, rather than brute-force computing power funded by hundreds of billions of dollars, were more important than previously thought in making significant AI breakthroughs - good news for the research-heavy UK tech economy. By midweek, DeepSeek had disappeared from app stores for Google and Apple devices in Italy after the data protection regulator demanded reassurances about what personal data is collected. The Dublin Data Protection Commission also demanded from DeepSeek explanations about its "data processing conducted in relation to data subjects in Ireland". In the US, where Donald Trump signed an executive order to "solidify [the US] position as the leader in AI", the arrival of DeepSeek was like a needle scratching across a record. Trump called it a "wake-up call for our industries that we need to be laser-focused on competing to win". Or as one X user parsed his message: "Get back in the code mines." It didn't take long for suspicions to take hold. David Sacks, the White House AI adviser, said: "There's substantial evidence that what DeepSeek did here is they distilled knowledge out of OpenAI models, and I don't think OpenAI is very happy about this." OpenAI's founder, Sam Altman, said he thought it was "legit invigorating to have a new competitor". But then, a day later, his company said it was "reviewing indications that DeepSeek may have inappropriately distilled our models". It also became apparent that DeepSeek would censor itself in real time when its answers might be politically embarrassing or challenging for the CCP. In Brazil, one user showed how DeepSeek began thinking about a question about free speech in China by wondering whether to include issues like Beijing's crackdown on protests in Hong Kong; the "persecution of human rights lawyers"; the "censorship of discussions on Xianjiang re-education camps"; and China's "social credit system punishing dissenters". Then, when it ruminated on how "in China, the primary threat is the state itself which actively suppresses dissent", the whole screed of "thinking" was deleted and DeepSeek apologetically asked the user if he wouldn't mind talking about maths or logic problems instead. Users could see what the chatbot really thought and the effect of the CCP on free speech; to see it all in action was unintentionally subversive. It was another week in which the strange world of AI got stranger and the stakes rose higher.

DeepSeek's success will undermine the US-China tech war

The writer is a founding partner of Hong Kong-based VC firm IN. Capital DeepSeek has forever altered the trajectory of global rivalry in tech. In China, founder Liang Wenfeng has become a local champion. For a country where overseas degrees -- especially those in the US -- are still perceived as more prestigious than their domestic equivalent, students and parents have been stunned to discover that his artificial intelligence start-up's research team were all educated domestically. Beijing is more confident than ever in its pursuit of the technology. DeepSeek's success undermines the barriers that have been created in the US-China tech war. The Hangzhou-based company's decision to release a low-cost, open-sourced AI model, alongside detailed disclosure of its training methods, means that everyone, from researchers in São Paulo to start-ups in Stockholm and doctors in Nairobi, can access state-of-the-art AI at little to no cost. Within the Chinese start-up sector a chain reaction is taking place. New AI applications are being created. Competition is going to become more fierce. Risk appetite from early-stage venture investment is increasing. DeepSeek's decision to pursue an open-source AI model is inspiring and putting pressure on others to do the same. The first to react was Alibaba's Qwen team, which released Qwen2.5 as open source last month on the eve of Chinese new year. This is a remarkable change. After the US start-up OpenAI released its generative AI model ChatGPT in late 2022, the global digital economy was edging towards control by a handful of tech giants. These players chase scale over efficiency -- building ever-larger models that demand staggering compute, energy and capital while guarding their training methods as trade secrets. Centralised, closed models create a dangerous feedback loop. The more data they amass, the more powerful they become, further marginalising anyone outside their gates. For consumers this means large fees, surrendered data and watching AI's future unfold without meaningful participation. The promise of DeepSeek's reasoning R1 model lies in its adaptability. Being open sourced it can be tailored to local needs. It avoids redundant calculations using something called sparse neural network training, meaning that its efficiency reduces compute and energy needs by orders of magnitude. This means that advanced AI can benefit the masses, not just the few. It proves that the technology is a commodity. Billions of dollars need not be wasted on competition between tech giants with closed models. AI's value should not lie in proprietary models but in what we are all able to do with it. As an investor, I am concerned that DeepSeek's prominence might lead the US to opt for even tougher sanctions. In China, export restrictions of graphics processing units (GPUs) such as Nvidia's powerful H100s have hindered start-up growth. Funding from foreign investors is limited due to compliance risk concerns. But the real danger lies in limiting access to global education and research collaboration, which stifles the global knowledge flow that is critical to sustaining progress. Talent can circumvent chip shortages, but erecting barriers to learning risks long-term stagnation. Yet even additional US restrictions, conspiracy theories and smear campaigns targeting DeepSeek cannot change the reality that the Chinese start-up has put AI into the hands of humanity. Against all the noise, let's consider this as a moment in history. In 1440, Johannes Gutenberg brought Europe the printing press, an invention that broke the monopoly on knowledge previously held by elites. DeepSeek's achievement joins this tradition of making information more accessible. Its low-cost reasoning model proves that AI can belong to everyone, not just those who are hoarding codes, chips and capital.

DeepSeek's rise: How a Chinese startup went from stock trader to AI star

In the rivalry between China and the United States over domination of artificial intelligence, DeepSeek seemed to come out of nowhere. In fact, it has skyrocketed through China's tech world in recent years with a path that was anything but conventional.Two years ago, when big-name Chinese technology companies like Baidu and Alibaba were chasing Silicon Valley's advances in artificial intelligence with splashy announcements and new chatbots, DeepSeek took a different approach. It zeroed in on research. The strategy paid off. The Chinese startup has jolted the tech world with its claim that it created a powerful AI model that was significantly cheaper to build than the offerings of its better-funded American rivals. In the rivalry between China and the United States over domination of artificial intelligence, DeepSeek seemed to come out of nowhere. In fact, it has skyrocketed through China's tech world in recent years with a path that was anything but conventional. Its mission to pursue research mirrors that of companies like OpenAI, the Silicon Valley firm that marked an American signature over AI in fall 2022. But the similarities mostly end there. DeepSeek's origins are in finance, not technology for technology's sake. Its parent company, a Chinese hedge fund called High-Flyer, began not as a laboratory devoted to safeguarding humanity from AI like Open AI, but as a business using AI to make bets in the Chinese stock market. High-Flyer had thrived by capitalizing on a market dominated by China's retail investors, who are known for jumping in and out of stocks impulsively. In 2021, High-Flyer found itself pressured by regulatory crackdowns in China on speculative trading, which authorities in Beijing felt was at odds with their attempts to keep markets calm. So High-Flyer pursued a new opportunity that it said aligned better with Chinese government priorities: advanced AI. "We want to do things with greater value and things that go beyond the investment industry, but it has been misinterpreted as AI stock speculation," High-Flyer CEO Lu Zhengzhe told Chinese state media in 2023. "We have set up a new team independent of investment, which is equivalent to a second startup." DeepSeek was born. As with many other Chinese startups, DeepSeek came at an established market with a different business approach. DeepSeek's latest model for artificial intelligence is believed to be nearly as powerful as American rivals but far more efficient. Its success suggests that Silicon Valley's AI lead has shrunk. DeepSeek's breakthrough, despite efforts by Washington to limit Chinese access to the advanced chips needed for AI, raises questions about how effective those controls can be over the long term -- although DeepSeek's founder has acknowledged that the chip restrictions are a limitation. DeepSeek did not rely on making consumer-facing AI products for revenue, and only this month released its first chatbot, which allows anyone to generate text and photos with simple commands. Instead, the company used the money that High-Flyer made from stock trading to bankroll ambitious research. The approach set it apart from U.S. rivals, all of which are ultimately consumer technology companies. This unconventional approach also allowed DeepSeek to sidestep stringent regulations the Chinese government has placed on AI use by the public. Because its focus was research and selling to businesses that use its model -- and, until the release of its chatbot this month, not consumer applications -- its early work did not face the same government restrictions. DeepSeek is run by its CEO, Liang Wenfeng, a thin, bespectacled engineer who studied at Zhejiang University in the eastern city of Hangzhou. He has said repeatedly in the few interviews he has given to Chinese media that to catch up with American innovation, Chinese companies must put research before profits. DeepSeek and High-Flyer did not respond to requests for comment. What Chinese technology companies "lack in innovation is certainly not capital, but a lack of confidence and knowledge about how to organize a high density of talent to achieve effective innovation," he said in a widely circulated interview with Chinese tech outlet 36Kr. Those who have worked with Liang describe him as a capable manager with a deep technical background, according to interviews and public accounts. "He's definitely an INTP," said Zihan Wang, a computer engineer who worked on an earlier DeepSeek model, referring to an introspective personality type from the Myers-Briggs test, a popular personality test among young people in China. "INTPs are really good researchers and they have a willingness to explore," Wang said. "He is not one of those people who wants to control everything." Liang was not too bothered with details like project timelines, and occasionally sent thought-provoking research questions to the entire team of researchers, Wang said. But mostly, Liang seemed driven to advance the technology and was not focused on profits. Unlike many Chinese companies, which tend to focus on hiring programmers, Liang has gained a reputation for employing people from outside computing. Poets and humanities majors from China's top universities on DeepSeek's staff train the model to write classical Chinese poetry and ace questions taken from the country's difficult college entrance examination. "Most of the team graduated from the top universities in China," said Yineng Zhang, a lead software engineer at Baseten in San Francisco who works on the SGLang, a project not part of DeepSeek that helps people build on top of DeepSeek's system. "They are very smart and very young." For years, Chinese tech companies pioneered AI applications used in computer vision, like facial recognition. But OpenAI's release of ChatGPT prompted a reckoning. When no Chinese company immediately released anything comparable, many concluded that American companies had a lead in advanced AI. In China, computer scientists were determined to prove they could compete. In 2023, many companies in China released their own large language models, the technology that underpins chatbots like ChatGPT. But making advanced models would require using a large number of chips that would cost hundreds of millions of dollars. High-Flyer was spending, too. By 2021, it was one of just a handful of Chinese companies that had been able to stockpile more than 10,000 advanced Nvidia A100 chips. Yet DeepSeek's research gave it a surprising advantage. Last year, it dramatically cut the prices it charged developers who build applications using its model, prompting a price war with larger rivals. Wang said there was little discussion of commercial applications for the technology they were building. Instead, he said, the company was focused on making an AI system that could be used by a range of people for many purposes. "During my time there, we did not talk much about how we make money," Wang said. "They just focused on making a great foundation model." A crucial part of DeepSeek's popularity is that it has made its developers' work public. This kind of information sharing, called open source, has been a cornerstone of the development of computer software, the internet and now artificial intelligence. In the United States, AI researchers and entrepreneurs have long followed the progress of DeepSeek's technology. Last year, the company turned heads when it released systems designed to generate their own computer programs. A new challenge for the company may come with its new high profile. The same day it released R1, the model behind its new chatbot, last week, Liang appeared at a round table discussion with Li Qiang, China's premier. DeepSeek's sudden popularity has thrust it to the center of the Chinese Communist Party's efforts to spur innovation, and that could prove difficult to manage, said Jimmy Goodrich, a senior adviser for technology analysis to the RAND Corp., a federally funded think tank. "It's a big predicament for DeepSeek. I'm sure they weren't on the government's five-year plan," he said. "Can they maintain this chaotic carefree vision when both the party and the world is watching?" Goodrich asked.

DeepSeek Isn't the Only Low-Cost AI Startup. Here's What It Means for OpenAI and Nvidia. | The Motley Fool

DeepSeek is an artificial intelligence (AI) research lab based in China. It was spun out of the country's most successful hedge fund, High-Flyer, in 2023. The fund had been using AI for years to develop trading algorithms. The DeepSeek team found a way to develop powerful large language models (LLMs) for a tiny fraction of the money being spent by America's leading AI companies. It triggered a panic in the U.S. stock market on Monday as investors considered the impact on chip suppliers like Nvidia (NVDA 0.77%) and prominent developers like OpenAI (which is backed by Microsoft). This could be a transformational moment in the AI race. Not only are DeepSeek's methods potentially valid, but there is at least one other Chinese AI start-up that seems to have produced similar results. Here's what it could mean for Nvidia and OpenAI. Ilya Sutskever is one of the co-founders of America's leading AI developer, OpenAI. He once believed data and computing power were the key ingredients to training the best AI models and producing the smartest AI software. This is known as pre-training scaling, and it meant the developers with the most financial resources could build the best data centers, buy the best chips, and win the AI race. But in November 2024, he told Reuters the results from using that method have plateaued. OpenAI has since developed models with better "reasoning" skills, meaning they spend more time "thinking" to produce the best responses from the ChatGPT chatbot. This is known as test-time scaling, and models that use it (GPT-4o1 to GPT-4o3) are better at problem solving, and bring AI closer to human intelligence on an academic level. It cost OpenAI around $20 billion to reach this point (money it mostly raised from investors since 2015). But DeepSeek recently released its V3 model, which was created for just $5.6 million, and yet it's competitive with OpenAI's GPT-4o models across several performance benchmarks. The U.S. government banned Nvidia from selling its latest graphics processors (GPUs) to Chinese AI companies, so DeepSeek developed V3 using less powerful versions like the H100 and the H800. To compensate for the lesser performance, DeepSeek had to innovate on the software side by creating more efficient algorithms and data input methods. The company also used a technique called distillation to create V3. It involves taking a small model and training it using a successful model like GPT-4o1 to produce a similar final product. This strategy supercharges the speed with which an AI company can train a competitive LLM, and it will potentially lead to commoditization. In other words, there could be hundreds of LLMs on the market in the future with similar capabilities, and they will mostly be interchangeable. That could be a real threat to OpenAI and even Nvidia. OpenAI could lose the advantage it established thanks to its considerable financial resources, and since less LLM training will be required, Nvidia could suffer from reduced demand for GPUs. Training is only one side of the equation. There is also the inference process, which involves the AI model turning prompts into accurate responses. But like with any business, lower overall costs can translate into lower prices for customers. As of this writing, DeepSeek charges just $0.14 per 1 million input tokens, which is 94% cheaper than OpenAI's rate of $2.50 per 1 million input tokens (input tokens are calculated based on the number of words in a user's prompt). But DeepSeek isn't the only AI lab which seems to have cracked this code. Kai-Fu Lee, who used to run Alphabet's Google operations in China, launched an AI start-up called 01.ai. According to its website, its Yi models perform well against competing models from DeepSeek. The company charges just $0.10 per 1 million input tokens, which is even cheaper than its Chinese rival -- and substantially cheaper than OpenAI. I think OpenAI is in trouble if LLMs continue trending toward commoditization. Plus, its models are closed-source, so developers are locked into the company's ecosystem, which won't be a desirable feature once competitive open-source LLMs are widely available. By contrast, DeepSeek uses an open-source approach that gives developers more freedom to tweak its models as necessary to build AI software. Developers can also download open-source models locally so they never have to share their sensitive data with the creator. But while OpenAI faces uncertainty, Nvidia might actually benefit from plunging inference costs, which could offset some of the lost GPU demand on the training side. Think about the progression of the cellphone. When we had to pay a fee every time we sent a text message or browsed the internet, we didn't use our phones as frequently as we do now. Unlimited plans with uncapped calls, texts, and data enable us to spend hours on our phone each day for a nominal monthly fee -- simply put, when costs came down, usage skyrocketed. AI could follow a similar path, and as usage increases, companies will need more of Nvidia's GPUs to cover demand for inference. That will be especially true as reasoning capabilities evolve, because more thinking requires substantially more computational power. The short-term picture is a little less certain. Will some of Nvidia's customers reduce their data center spending as they optimize their training methods like DeepSeek did? It's hard to say, but a new quarterly earnings season just started, so we should receive an update from almost every one of them over the next few weeks.

Experts Say DeepSeek Proves High-Performance AI Possible Without Silicon Valley's Spending

Security concerns emerge as DeepSeek's open-source approach could challenge regulatory frameworks and raise geopolitical AI tensions. A new player in the AI arena, DeepSeek, has unveiled an open-source language model, R1, that experts say is a game-changer in efficiency, raising questions about the future of AI development. The DeepSeek Effect Developed for under $6 million, DeepSeek's R1 has not only outperformed established competitors in several tests but has also called into question the established big-spending approach prevalent in Silicon Valley. This development has triggered a mix of optimism about democratized AI and fears of an AI stock bubble bursting, sending ripples throughout financial markets. "Another DeepSeek is hiding in plain sight," said Roi Ginat, CEO of Endless AI, a multimodal AI company, highlighting DeepSeek's influence in making impactful AI with limited resources. Ginat's own company, EndlessAI, has been working on efficient video processing, echoing DeepSeek's stand to using LLMs in practical, cost-effective ways. Ginat told Benzinga that EndlessAI provides the tools to make existing models work smarter by using proprietary video transfer optimization which enables massive cost reduction for video processing, adding that this is complementary, not competitive, with what DeepSeek is doing. Speaking with Benzinga, Satya Nitta, CEO of Emergence AI, an AI orchestration company, echoed the sentiment, calling DeepSeek R1 a "meaningful advance in broadening access to AI reasoning." Nitta emphasized the importance of open-source solutions that are not only trusted but also scalable and adaptable to evolving needs. Matthew Putman, CEO of Nanotronics, an industrial AI company, told Benzinga that DeepSeek's achievement demonstrates that "it is possible to do more with less," suggesting that resource constraints may fuel innovation. Putman sees these developments as essential for broader adoption of AI for industries beyond just text generation, stressing it's critical for advancing semiconductor production, which makes AI technology possible. Mitesh Agrawal, CEO of Positron, an AI hardware company, pointed to DeepSeek's success as evidence that "inference-time compute is the new scaling factor" in AI development, delivering real results with optimized systems. Agrawal told Benzinga that performance per dollar and performance per watt are critical to make the best AI models accessible. Also Read: Helix Debuts AI Index Perpetual Market, Bridging Crypto And Traditional Equities On-Chain Is DeepSeek A Cause For Concern? However, concerns also accompany this disruptive innovation. Speaking with Benzinga, Pukar Hamal, CEO of SecurityPal, an AI-enhanced security review company, warned about the security and regulatory implications of an "open-source" model with unknown alignment. Hamal highlighted the "gas law" of computing, whereby cheaper compute expands markets, but also stressed the need for guardrails to protect the public and maintain compliance. These concerns translated into market jitters on the day DeepSeek news first broke. The Nasdaq Composite dropped 2.7% on Monday, the S&P 500 slid 1.6% and the Dow Jones Industrial Average dipped 0.1% following the DeepSeek announcement. The announcement of a new low-cost, high-performance open source AI model that competes with existing proprietary AI systems also drove volatility in the cryptocurrency market. Crypto markets also took a hit, with Bitcoin BTC/USD falling 5.5% and Ethereum ETH/USD dropping 7% on Monday before recovering. Other major altcoins also experienced significant losses, as traders grappled with the potential implications of DeepSeek's emergence. On Monday, altcoins like Solana SOL/USD, BNB BNB/USD, Dogecoin DOGE/USD and Cardano ADA/USD were down 11%, 3.5%, 9% and 8.5%, respectively. What's Next For DeepSeek DeepSeek's disruptive impact is already rippling through financial markets. The sharp declines in AI-related stocks and cryptocurrencies suggest that investors are reassessing the sustainability of current AI valuations. Analysts warn that DeepSeek's ability to deliver high-performance AI at a fraction of traditional costs may force industry leaders to rethink their strategies. Moreover, experts believe DeepSeek's approach could reshape the $25 billion computer vision market and redefine the scalability of AI applications. As companies continue to push the boundaries of multimodal AI, the industry appears poised for a shift toward leaner, more efficient AI models that can thrive outside the confines of billion-dollar research labs. In the face of this transformation, the question remains: will legacy AI firms adapt to the new paradigm, or will DeepSeek's breakthrough mark the beginning of a broader industry upheaval? Read Next: Grayscale Seeks SEC Approval To Convert XRP Trust Into Exchange-Traded Fund Image created using artificial intelligence with Midjourney. $BTCBitcoin$104698.71-0.04%WatchlistOverview$ADACardano$0.97501.60%$BNBBNB--%$DOGEDogecoin$0.33300.44%$ETHEthereum$3353.653.27%$SOLSolana$240.170.74%Market News and Data brought to you by Benzinga APIs

Ex-OpenAI VP's Shocking DeepSeek Warning - Wes Roth

It's no secret that artificial intelligence (AI) is reshaping the world as we know it, but what if the race to dominate this technology could redefine global power itself? Dario Amodei, a leading voice in AI and former VP at OpenAI, has raised a red flag about China's rapid advancements in AI, spearheaded by its trailblazing company, DeepSeek. With breakthroughs that rival -- and in some cases, outpace -- Western innovations, DeepSeek's progress is more than just a technological feat; it's a wake-up call for the global community. Amodei's insights reveal a high-stakes competition where the balance of power, ethical oversight, and the future of AI leadership hang in the balance. But this isn't just about chips, algorithms, or geopolitical rivalries -- it's about the kind of world we want to live in. As DeepSeek pushes the boundaries of AI efficiency and capability, questions arise about who gets to wield this power and how it will be used. Amodei's warning isn't just a call to action for governments and tech leaders; it's a reminder to all of us that the choices made today will shape the role AI plays in our lives tomorrow. So, how do we ensure that this fantastic technology remains a force for good? DeepSeek has positioned itself as a central force in China's AI ecosystem, achieving remarkable progress in developing advanced AI systems. Its latest models, V3 and R1, showcase significant improvements in reasoning, efficiency, and adaptability. The R1 model, in particular, employs reinforcement learning to enhance problem-solving capabilities, placing it on par with some of the most sophisticated AI systems developed in the United States. What distinguishes DeepSeek is its ability to achieve these advancements at a fraction of the cost. This efficiency is driven by several key factors: These developments are more than just technical achievements; they represent a paradigm shift in the global AI landscape. By reducing costs and improving efficiency, DeepSeek is closing the gap with Western AI leaders, raising critical questions about the implications for global governance, economic influence, and the balance of power in AI development. AI scaling laws provide a framework for understanding how increased resources -- such as computational power, data, and training -- translate into improved model performance. These laws reveal that while greater investment often leads to better outcomes, the costs associated with innovative AI research are escalating rapidly. DeepSeek, alongside other leading AI organizations like OpenAI, is actively pushing the boundaries of these scaling laws. By reinvesting efficiency gains into the development of more advanced systems, they are driving the next wave of innovation. A cornerstone of this progress is reinforcement learning, which allows AI models to learn through trial and error. This approach not only enhances reasoning capabilities but also enables AI systems to tackle increasingly complex tasks. These advancements are accelerating the development of large language models and other fantastic AI technologies, reshaping industries and scientific research alike. Master Artificial Intelligence (AI) with the help of our in-depth articles and helpful guides. The competition for AI supremacy between the United States and China is intensifying, with DeepSeek's advancements underscoring China's growing capabilities. This rivalry extends beyond technological innovation, influencing critical areas such as national security, economic power, and global influence. To maintain its leadership, the United States has implemented export controls on advanced chips, such as Nvidia's H100, which are essential for training and deploying state-of-the-art AI systems. However, loopholes in these regulations and the risk of smuggling pose significant challenges to enforcement. Dario Amodei highlights the importance of strengthening these controls to prevent China from further scaling its AI capabilities. The prospect of a "bipolar" AI world, where both the U.S. and China possess equally advanced systems, raises concerns about heightened military and strategic competition. Conversely, a "unipolar" AI world dominated by democratic nations could foster stronger oversight and ensure the ethical use of AI technologies. This underscores the need for democratic nations to collaborate in shaping the future of AI development. Amodei predicts that AI systems capable of surpassing human intelligence in most tasks could emerge as early as 2026-2027. This accelerated timeline highlights the critical need for strategic foresight to guide AI development responsibly. The convergence of reinforcement learning and large language models is driving a pivotal transformation, with the potential to transform scientific research, technological innovation, and industrial processes. However, the risks of unchecked AI development are profound. Authoritarian regimes could exploit AI for surveillance, control, and other unethical purposes, threatening individual freedoms and global stability. To counter these risks, democratic nations must prioritize responsible AI development. This involves making sure that AI technologies are used to benefit humanity while upholding ethical standards and safeguarding individual rights. Export controls are a vital tool for shaping the trajectory of AI development. By restricting access to advanced chips and other critical technologies, democratic nations can slow the progress of authoritarian regimes in building powerful AI systems. Amodei advocates for a coordinated international effort to close regulatory loopholes and enforce these controls effectively. Beyond technology, the geopolitical stakes of AI leadership extend to broader issues of global governance. The ability to set norms and standards for AI use will shape international relations for decades to come. Democratic oversight is essential to ensure that AI serves as a force for progress and innovation, rather than a tool for oppression or conflict. By fostering collaboration and ethical practices, nations can work together to navigate the challenges and opportunities presented by AI.

How DeepSeek Took a Chunk Out of Big AI

US companies have been pouring billions and billions of dollars into advancing AI, warning the expense will only get higher now that there's no more information to copy from the internet to teach the machines. But a little-known Chinese startup just revealed how all that outlay (a quarter of a trillion dollars projected for 2025 alone) may not be necessary -- and that maybe Silicon Valley isn't as far ahead as it thinks it is. In the Bloomberg Originals mini-documentary How China's DeepSeek Came for Big AI, we introduce you to the young man behind the company and how he may have gone about building a chatbot on the cheap. It's an app that's suddenly put the likes of OpenAI, Google, Microsoft and Meta on the back foot. But it was technology stock darling Nvidia that was especially bloodied by this sudden revelation: its market capitalization took a dive of galactic proportions -- the biggest ever, in fact.

The AI business model is built on hype. That's the real reason the tech bros fear DeepSeek | Kenan Malik

While privacy fears are justified, the main beef Silicon Valley has is that China's chatbot is democratising the technology No, it was not a "Sputnik moment". The launch last month of DeepSeek R1, the Chinese generative AI or chatbot, created mayhem in the tech world, with stocks plummeting and much chatter about the US losing its supremacy in AI technology. Yet, for all the disruption, the Sputnik analogy reveals less about DeepSeek than about American neuroses. The original Sputnik moment came on 4 October 1957 when the Soviet Union shocked the world by launching Sputnik 1, the first time humanity had sent a satellite into orbit. It was, to anachronistically borrow a phrase from a later and even more momentous landmark, "one giant leap for mankind", in Neil Armstrong's historic words as he took a "small step" on to the surface of the moon. It was a significant moment in the cold war, too. A confidential White House report worried that "American prestige" had "sustained a severe blow", giving the USSR "clear advantage in the cold war". That fear spurred Washington into reshaping its space programme, and catalysed the Apollo missions, culminating with Armstrong and Buzz Aldrin becoming, on 20 July 1969, the first humans to walk upon another celestial body. DeepSeek, sponsored by a Chinese hedge fund, is a notable achievement. Technically, though, it is no advance on large language models (LLMs) that already exist. It is neither faster nor "cleverer" than OpenAI's ChatGPT or Anthropic's Claude and just as prone to "hallucinations" - the tendency, exhibited by all LLMs, to give false answers or to make up "facts" to fill gaps in its data. According to NewsGuard, a rating system for news and information websites, DeepSeek's chatbot made false claims 30% of the time and gave no answers to 53% of questions, compared with 40% and 22% respectively for the 10 leading chatbots in NewsGuard's most recent audit. The figures expose the profound unreliability of all LLMs. DeepSeek's particularly high non-response rate is likely to be the product of its censoriousness; it refuses to provide answers on any issue that China finds sensitive or about which it wants facts restricted, whether Tiananmen Square or Taiwan. The true impact of DeepSeek is not on the technology but on the economics of AI. It is a chatbot as capable, and as flawed, as other current leading models, but built at a fraction of the cost and from inferior technology. The US ban on the sale to China of the most advanced chips and chip-making equipment, imposed by the Biden administration in 2022, and tightened several times since, was designed to curtail Beijing's access to cutting-edge technology. Paradoxically, it may have spurred Chinese researchers into becoming more innovative. DeepSeek is also free to use, and open source. The combination of low cost and openness may help democratise AI technology, enabling others, especially from outside America, to enter the market. There is a certain irony that it should be China that is opening up the technology while US firms continue to create as many barriers as possible to competitors attempting to enter the field. And here lies perhaps the biggest impact of DeepSeek. It has ripped off the veil of mystique that previously surrounded AI. Silicon Valley has nurtured the image of AI technology as a precious and miraculous accomplishment, and portrayed its leading figures, from Elon Musk to Sam Altman, as prophets guiding us into a new world. The technology itself has been endowed with almost magical powers, including the promise of "artificial general intelligence", or AGI - superintelligent machines capable of surpassing human abilities on any cognitive task - as being almost within our grasp. Last April, Musk predicted that AI would be "smarter than any human" by the end of 2025. Last month, Altman, the CEO of OpenAI, the driving force behind the current generative AI boom, similarly claimed to be "confident we know how to build AGI" and that "in 2025, we may see the first AI agents 'join the workforce'". Almost a decade ago, the Nobel prize-winning computer scientist Geoff Hinton urged nations to "stop training radiologists", and similar medical technicians, because "it's completely obvious within five years, deep learning [AI] is going to do better". Dario Amodei, the CEO of Anthropic, a corporation founded by former OpenAI employees, has claimed that AI could double the human lifespan within five to 10 years. These fantasy claims have been shredded by critics such as the American cognitive scientist Gary Marcus, who has even challenged Musk to a $1m bet over his "smarter than any human" claim for AI. Nevertheless, for all the pushback, each time one fantasy prediction fails to materialise, another takes its place. Such claims derive less from technological possibilities than from political and economic needs. While AI technology has provided hugely important tools, capable of surpassing humans in specific fields, from the solving of mathematical problems to the recognition of disease patterns, the business model depends on hype. It is the hype that drives the billion-dollar investment and buys political influence, including a seat at the presidential inauguration. It is also an approach that seeks to advance AI less through major scientific breakthroughs than through a brute force strategy of "scaling up" - building bigger models, using larger datasets, and deploying vastly greater computational power. The disruptive quality of DeepSeek lies in questioning this approach, demonstrating that the best generative AI models can be matched with much less computational power and a lower financial burden. The hype around DeepSeek is in part a reflection of the hype around AI. It is a reflection, too, of geopolitical tensions. Had DeepSeek been created by geeks at a US university, it would most likely have been feted but without the global tumult of the past two weeks. Beneath the panic lies fear of DeepSeek's Chinese origins and ownership. Yet, too great an obsession with the geopolitics of DeepSeek can distort the lessons we take from it. The promise of more open access to such vital technology becomes subsumed into a fear of its Chinese provenance. Concerns about privacy, censorship and surveillance, rightly raised by a model such as DeepSeek, can help obscure the reality that such issues bedevil all AI technology, not just that from China. Particularly at a time of threatened trade wars and threats to democracy, our capacity to navigate between the hype and the fear assumes new importance.

The DeepSeek Wake-Up Call

This is Atlantic Intelligence, a newsletter in which our writers help you wrap your mind around artificial intelligence and a new machine age. Sign up here. Earlier this week, almost overnight, the American tech industry entered a full-on panic. The latest version of DeepSeek, an AI model from a Chinese start-up of the same name, appeared to equal OpenAI's most advanced program, o1. On Monday, DeepSeek overtook ChatGPT as the No. 1 free app on Apple's mobile-app store in the United States. So far, China has lagged the U.S. in the AI race. DeepSeek suggests that the country has gained significant ground: The chatbot was built more quickly and with less money than analogous models in the U.S., and also appears to use less computing power. Software developers using DeepSeek pay roughly 95 percent less per word than they do with OpenAI's top model. One prominent AI executive wrote that DeepSeek was a "wake up call for America." Because DeepSeek appears to be cheaper and more efficient than similarly capable American AI models, the tech industry's enormous investments in computer chips and data centers have been thrown into doubt -- so much that the top AI chipmaker, Nvidia, lost $600 billion in market value on Monday, the largest single-day drop in U.S. history. Sam Altman, the CEO of OpenAI, said that it was "invigorating to have a new competitor" and that, in response, the company would move up some new software announcements. (Yesterday morning, OpenAI said that it is investigating whether DeepSeek used ChatGPT outputs to train its own model.) But many prominent American researchers and tech executives celebrated DeepSeek, as well. That's because "the most notable feature of DeepSeek may be not that it is Chinese, but that it is relatively open," I wrote on Monday. Whereas the top American AI labs at OpenAI, Google, and Anthropic have kept their technology top-secret, DeepSeek published an in-depth technical report and is allowing anybody to download and modify the program's code. "Being democratic -- in the sense of vesting power in software developers and users -- is precisely what has made DeepSeek a success," I wrote. Start-ups and researchers love this relative transparency. In theory, competitors can use DeepSeek's code and research to rapidly catch up to OpenAI with far fewer resources -- you might not need colossal data centers to get to the front of the AI race. (The Atlantic recently entered into a corporate partnership with OpenAI.) However, there's substantial uncertainty about just how much cheaper DeepSeek was to build, based on reports about the start-up's hardware acquisitions and uncertainty about how the model was trained. Meanwhile, for national-security hawks, the fear is that an open-source program that won't answer questions about the Tiananmen Square protests could become a global technological touchpoint. DeepSeek could face similar privacy concerns as TikTok: Already, the U.S. Navy has banned its use, citing security concerns. Any predictions, for now, are highly speculative. The global AI race is far from over, and forthcoming products from Silicon Valley could leap ahead once again. At the very least, U.S. tech companies may have to reconsider whether the best way to build AI is by keeping their models a secret. After several major tech executives announced their support for Donald Trump, many liberal internet users are now alleging that they are being censored on certain social-media platforms. "To some, this pattern was as unmistakable as it was malicious," my colleague Kaitlyn Tiffany writes. "Social media was turning against Democrats." And they are panicking.

DeepSeek's success challenges assumptions about Chinese tech companies - and the US-China competition

The release of the new DeepSeek-R1 artificial intelligence (AI) model has shocked the tech world. Launched on January 20 with little fanfare, the Chinese AI model was reportedly developed at only a fraction of the cost of OpenAI's GPT-4o, and over a much shorter period of time. One Chinese commentator has called its release a "Pearl Harbor attack" on the AI world. Though the reference to an "attack" may be a strong word, it alludes to the growing competition between the United States and China over dominance in the AI sphere, which the US had been leading thus far. Indeed, people across China were celebrating a homegrown success story on Wednesday, as DeepSeek's AI app soared to the top of the Apple and Google stores in the US. So, what does the emergence of DeepSeek's model say about US-China competition in this space? Chinese government control First, DeepSeek's success is undoubtedly sending a message to the Chinese government that excessive control kills innovation. Until mid-2023, enthusiasm for innovation in China's tech companies had been stifled by increasingly restrictive regulations. The Chinese government had embarked on a sweeping crackdown of tech companies like Alibaba and others in order to prevent the spread of rampant entrepreneurial capitalism in China. The launch of ChatGPT in 2023 promised to open up exciting new frontiers for the development of AI in the West. But it must have come as a rude shock to China's tech companies. The Chinese government changed tact and reassured them that it recognised the crucial role of the digital economy as a key driver of economic growth. It soon began to relax its tight grip over the sector. But the elephant in the room is how DeepSeek - and China's AI companies in general - will deal with censorship. As it stands, politically sensitive words and questions seem to be no-go areas for DeepSeek. When asked what happened on June 4 1989 in Tiananmen Square (the site of the government's crackdown on democracy protesters), the chatbot's answer was along the lines of, "Sorry, that's beyond my current scope. Let's talk about something else." This raises the question: can a Chinese AI tool be truly competitive in the global tech race without a solution to the challenge of censorship? US efforts to contain Chinese tech development Meanwhile, the US has adopted a wide array of measures aiming at curbing China's AI development over the past few years. These included the Biden administration's attempts to restrict China's access to the advanced chips needed for AI, as well as the export of chip-making equipment and other technology to China. The US has also blacklisted a large number of Chinese entities that it has identified as having both military and commercial technology. The launch of DeepSeek raises questions over the effectiveness of these US attempts to "de-risk" from China in relation to scientific and academic collaboration. For one, DeepSeek was able to evade US restrictions on advanced chips by stockpiling downgraded chips made by Nvidia before the Biden administration moved to ban them. Western observers have often portrayed China's AI initiatives as limited due to these US controls. However, these observers have somehow failed to take seriously the emergence of a new generation of Chinese entrepreneurs who prioritise foundational research and long-term technological advancement over quick profits. DeepSeek is a good example of this approach. It has embraced open-source methods, pooling collective expertise and fostering collaborative innovation. This approach not only mitigates resource constraints, but also accelerates the development of cutting-edge technologies. Another common assumption in the West is that Chinese companies are mere followers or imitators. DeepSeek's achievements likewise challenge this perception. As the company's chief executive, Liang Wenfeng, said to one Chinese media outlet: Innovation such as ours happens all the time in the US. The Americans are surprised by us, mainly because we are a Chinese company, and we are entering their game as an innovator with original contribution, not as followers. DeepSeek's success also calls into question the legislation supported by both the Biden and Trump administrations that aims to prevent Chinese graduate students from attending universities in the US. The assumption behind what researchers call "STEM talent de-coupling" is that the Chinese government may use some of these students to engage in knowledge and technology transfer when they return to China. Liang, however, never studied outside China. And he recruited graduates and students from top Chinese universities to staff his research team. None studied overseas. These developers belong to a generation of young, patriotic Chinese who harbour personal ambition, as well as a broader commitment to advancing China's position as a global innovation leader. What does this mean for Australia? In Australia, the initial reaction to DeepSeek's AI chatbot has been one of caution, even concern. Clare O'Neil, the former cyber security minister, said the government would examine more closely how the app works before providing guidance to Australians on potential data security concerns. But DeepSeek may also be a reminder that Australia's scientific collaborations should be guided primarily by research excellence rather than geopolitical considerations. To stay competitive and reduce its reliance on external technology providers, Australia needs to invest in its own AI research infrastructure and build its own talent pool. A narrow focus on political alignments and a growing paranoia about partnering with Chinese researchers means that Australia risks missing out on the next wave of breakthrough technologies.

DeepSeek Sends Artificial Intelligence (AI) Giants Plunging. Here's What You May Want to Do Now | The Motley Fool

Tech giants around the globe were rattled on Jan. 27 after Chinese AI start-up DeepSeek unveiled an impressive, low-cost artificial intelligence (AI) model, sparking widespread concerns about the scale of investment being poured into expensive hardware and data centers. DeepSeek released a free AI assistant last week that quickly saw its downloads overtake OpenAI's ChatGPT on the Apple App Store in the U.S. The Chinese start-up claims the free, open-source large language model known as V3 took just two months to train and cost less than $6 million. What's more, DeepSeek claims its latest cost-effective model, the R1, can match the performance of OpenAI's o1 reasoning model on certain benchmarks. The company also claims R1 is 20 to 50 times cheaper to use than o1. The advent of this cheaper model hammered the shares of tech giants that have been investing billions of dollars in AI infrastructure. Nvidia (NVDA 0.77%), which is the leading provider of graphics processing units (GPUs) used for AI model training and inferencing, saw its stock price plunge 17% on Jan. 27. The company has been one of the biggest beneficiaries of the massive spending on AI infrastructure as it controls an estimated 85% of the AI chip market. Nvidia's top-of-the-line data center GPUs are reportedly priced between $30,000 to $40,000 per unit. Tech giants such as Microsoft and Meta Platforms have been purchasing hundreds of thousands of these chips from Nvidia to train and deploy AI models. These two companies reportedly spent $9 billion in 2023 to purchase a total of 300,000 AI chips from Nvidia. That investment most likely shot up last year as Microsoft was reportedly planning to get its hands on 1.8 million Nvidia GPUs, while Meta hoped to procure 350,000 chips last year. DeepSeek, meanwhile, claims it used 2,048 of Nvidia's performance-capped H800 GPUs (a downgraded version of the chipmaker's popular H100 AI GPU) to train its model. For comparison, OpenAI reportedly used more than 10,000 flagship H100 chips to train its GPT-4 and GPT-4o models. If DeepSeek's claims are true, then it means competitive AI models can be trained at a fraction of the cost of what the likes of Microsoft, Meta, and others have been spending, putting in doubt the entire tech industry's huge investment plans. It was no surprise that the entire AI ecosystem was deep in the red following this development. Broadcom (AVGO 4.51%), which supplies custom AI processors to Microsoft, Alphabet, Meta, and OpenAI to develop in-house chips, saw its stock plunge 17% as well. Foundry giant Taiwan Semiconductor Manufacturing (TSM 2.88%), popularly known as TSMC, fell more than 13%. Both Broadcom and TSMC have also been riding the surge in AI chip demand with their share prices soaring in the past couple of years. Broadcom dominates the custom AI processor market with an estimated market share of 55% to 60%, according to JPMorgan. The company sees a massive addressable market worth $60 billion to $90 billion for its custom processors and networking chips over the next three years. That points toward a major increase in its AI revenue from last year's level of just over $12 billion. TSMC, meanwhile, is the go-to manufacturer of chips for fabless chipmakers such as Nvidia and Broadcom. It is the dominant player in the global foundry market by quite some distance, and it has built a solid clientele that allows it to benefit from the adoption of AI in multiple industries. The foundry giant reported impressive revenue growth of 37% in 2024 and is expecting solid performance in 2025 as well. Wall Street analysts believe the DeepSeek-inspired fears about a slowdown in AI infrastructure spending are overblown. The compute capacity that will be freed up by the launch of more efficient AI models is likely to be absorbed elsewhere, according to Bernstein Research. Others share a similar view, believing that the use of AI should only expand, sustaining demand for chips from Nvidia, Broadcom, and TSMC. At the same time, there are suspicions that DeepSeek may be downplaying how much it spent training its models. According to one source, it may actually have access to 50,000 of Nvidia's powerful H100 GPUs. Given the uncertainty of the situation, investors would do well to avoid a knee-jerk reaction as seen with the recent sell-off. After all, spending by major tech giants on AI infrastructure is still likely to head higher once again in 2025, as evidenced by the capital spending plans laid out by the likes of Microsoft and Meta, as well as the announcement of the Stargate project. Investors who remain bullish on AI can now buy many top AI stocks at a modest discount.

How DeepSeek Is Disrupting AI and the Global Economy

A new Chinese AI model, has become an instant sensation around the world. The assistant, which was developed by a startup based in Hangzhou, quickly climbed to the top of Apple's App Store, stunning investors and technology analysts. Its rapid success has also spurred discussions about ways for artificial intelligence to get better without depending on huge infrastructure and costly computing power. The main advantage of DeepSeek is its efficiency. While OpenAI's top performers cost more than $100 million to develop, DeepSeek reportedly took less than two months and less than $6 million. This development has questioned the assumption that state-of-the-art A.I. has to rely on high-end microchips and lots of computing power. If so, we could see broader adoption of AI tools as this AI may be affordable to each and every business around the world.

OpenEuroLLM vs. DeepSeek and ChatGPT: Key Differences Explained

Given the millions of users and powerful capabilities of ChatGPT and DeepSeek, can OpenEuroLLM compete? As the U.S. and China continue to speed ahead with new AI developments, spearheaded by OpenAI and DeepSeek, Europe has somewhat fallen into the background of the global AI race. In an attempt to change this, a European alliance has announced the development of OpenEuroLLM -- a European group of large language foundation models. The aim is to become a competitive alternative to OpenAI and DeepSeek while remaining grounded in the EU's stringent safety and privacy guidelines. However, with OpenAI's ChatGPT and DeepSeek's R1 leading the AI race with millions of users, how does OpenEuroLLM currently compare? Developed by European AI allies, OpenEuroLLM is designed as an open‐source alternative that strengthens Europe's independent AI capabilities. The mission is to reduce reliance on U.S. and Chinese models by fostering collaborative, home‐grown AI innovation in the continent. OpenEuroLLM has said it would incorporate AI into European values, focusing on transparency, openness, and democracy. ChatGPT Developed by OpenAI, ChatGPT is part of the AI firm's mission to "ensure that artificial general intelligence -- AI systems that are generally smarter than humans -- benefits all of humanity." It is offered in both free and subscription-based versions, with continuous refinements driven by extensive usage feedback. ChatGPT is designed to be a widely accessible, safe, and useful tool that helps people solve problems and streamline everyday tasks. Thousands of companies use OpenAI's chatbot, which is strong in generating content and handling general-purpose inquiries. DeepSeek Created by a Chinese research team, DeepSeek emphasizes cost efficiency and rapid deployment. It is offered free and is built to perform well on technical, logic‐oriented tasks. However, it has also drawn attention to issues such as strict censorship on politically sensitive topics and data privacy concerns, given that user data is stored on servers in China. With European data protection as a core design principle, OpenEuroLLM will be expected to adhere to the sweeping rules being enforced in the continent. The EU's AI Act, which came into force on August 1, requires transparency for all AI systems that pose potential risks. All AI systems deemed high-risk will need to comply with specific legal requirements, while those deemed unacceptable will be banned. OpenEuroLLM, which focuses on transparency and ethical development, will likely play a pivotal role in the EU's heavily regulated AI landscape. ChatGPT ChatGPT adheres to Western data protection standards and is subject to ongoing debates over bias and data handling. OpenAI's closed-source nature means that many of its internal processes remain proprietary, sometimes limiting transparency compared to open-source models. DeepSeek Despite its technical prowess, DeepSeek has faced regulatory scrutiny in Italy and Ireland over its data storage practices. The fact that user data may be stored on servers in China, combined with the model's built-in censorship mechanisms, has raised concerns about transparency and user rights. While detailed technical documentation on the project is still emerging, OpenEuroLLM is built with a focus on transparency and open collaboration. The EU AI Act states that any AI developers acquiring data for AI training from the web must ensure they have first received consent. This means that OpenEuroLLM's architecture is unlikely to scrape as much data from the web as ChatGPT and DeepSeek. ChatGPT ChatGPT uses a traditional transformer-based architecture that processes queries at the full capacity of the model. This type of architecture demands vast computational resources, meaning it comes with higher training and operational costs. DeepSeek employs a Mixture-of-Experts (MoE) architecture, which means that for each query, only a subset of its 671 billion parameters is activated. This selective processing significantly reduces training and operational costs and allows it to excel in technical tasks and logical reasoning. "Think of it like assembling the Avengers for every input  --  only the best-suited heroes (experts) get called in," Harshada Jivane, machine learning engineer, explained in a blog post. The new European AI project currently has a budget of €52 million, along with a compute commitment of a possibly higher value, Peter Sarlin, the co-founder of Silo AI who is backing the project, told TNW . In addition to being backed by the European Commission, OpenEuroLLM has partnered with over 20 research institutions and leading AI startups. The project also receives funding from STEP, an EU scheme to boost investment in innovative technologies. ChatGPT Over the years, OpenAI has attracted substantial investments from major backers, including Microsoft, SoftBank, and Oracle. According to the Wall Street Journal , the U.S. AI firm is in talks to raise a new round of funding to double its valuation to $340 billion. These billions of dollars of investments underscore the heavy capital requirements behind building and running advanced AI models like ChatGPT. DeepSeek In contrast to the enormous sums raised by OpenAI, DeepSeek is reportedly developed on a much leaner budget. DeepSeek's development is fully funded by High-Flyer , a quantitative hedge fund in China. According to public disclosures, its flagship model, R1, was trained with an investment of roughly $9 million.

What Is DeepSeek and How Should It Change How You Invest in AI?

Peter Gratton, M.A.P.P., Ph.D., is a New Orleans-based editor and professor with over 20 years of experience in investing, risk management, and public policy. Peter began covering markets at Multex (Reuters) and has expanded his coverage to include investments, ethics, public policy, and the health and travel industries. The generative AI industry in the U.S. is getting a wake-up call, as are jittery investors who have seen their stocks rise with the AI boom in the last few years. That's because a Chinese startup, DeepSeek, upended conventional wisdom about how advanced AI models are built and at what cost. The company reported in early 2025 that its models rival those of OpenAI's Chat GPT, all for a reported $6 million in training costs. That triggered a record $600 billion single-day drop in Nvidia's (NVDA) stock and forced investors to rethink their AI-based bets going forward. For anyone investing in AI, understanding DeepSeek's rise is critical for navigating a new era in this sector. We get you up to speed below. What Is DeepSeek? DeepSeek is a Hangzhou, China-based AI research company founded in July 2023 by former hedge fund executive Liang Wenfeng and backed by quantitative investment giant High-Flyer Quant. It has prioritized algorithmic efficiency and open-source collaboration to challenge the AI dominance of U.S. tech giants. Since its launch, DeepSeek has released a series of impressive models, including DeepSeek-V3 and DeepSeek-R1, which it says match OpenAI's o1 reasoning capabilities at a fraction of the cost. In addition, DeepSeek's models are open source, meaning they are freely available for anyone to use, modify, and distribute. Most crucially -- and potentially most devastating for competitors -- DeepSeek achieved these advances despite U.S. export restrictions on advanced AI chips, such as Nvidia's H100 and A100 models. In fact, those limits might have been a godsend for the company's breakthroughs. DeepSeek says it found workarounds for greater chip capacity and applied other crucial strategies that competitors are now likely poring over in detail: This combined approach enabled the company to train its models using about 2,000 Nvidia GPUs over 55 days at a cost of around $5.6 million, a fraction of what U.S. tech giants are spending. How Does This Affect AI Investing? DeepSeek's success challenges the prevailing idea fueling massive investments in AI in the U.S. -- that AI development requires endless piles of cash for massive spending on Nvidia-type chips and other expensive technology. Major tech stocks in the U.S. had significant declines on news of these developments. Nvidia's stock plummeted nearly 17%, the largest single-day loss in U.S. stock market history. Other stocks in the AI space, including Microsoft Corporation (MSFT), Alphabet Inc. (GOOGL), and ASML Holding NV (ASML), also plummeted. The DeepSeek moment also creates opportunities for investors in the AI space. Here's what savvy investors are likely to do: DeepSeek's Generative Error Problem Investors should also stay updated as experts get a look under the hood at DeepSeek. An early study from NewsGuard, which rates the trustworthiness of news and information sites, included reasons for significant concerns about DeepSeek's reliability. Despite topping App Store downloads, the Chinese AI chatbot failed accuracy tests 83% of the time, placing it near the bottom of evaluated AI chatbots -- ranking 10th out of 11 competitors. NewsGuard's assessment uncovered several critical issues: DeepSeek's approach to accuracy thus seems to shift responsibility to users, with its terms of use notifying them to "proactively verify the authenticity and accuracy of the output content." The Bottom Line DeepSeek may be a harbinger of a less costly future for AI. This could mean pivoting to a focus on software changes over the brute force of more and more expensive technology, open-source collaboration, and scalable infrastructure. But it also means looking past the hyped-up headlines and assessing whether DeepSeek offers something new and different or, given some early tests of its abilities, if it's just another AI-produced hallucination.

DeepSeek shocked the AI world this week. Here's how tech CEOs responded

"We have to run harder, run faster, have an all-country effort," Palantir CEO Alex Karp told CNBC, in an interview that aired on Friday. Microsoft CEO Satya Nadella commented on its "real innovations." OpenAI's Sam Altman described it as "clearly a great model." Apple CEO Tim Cook said "innovation that drives efficiency is a good thing." And Palantir's Alex Karp said it shows the importance of "an all-country effort." The tech CEOs were all talking about China's DeepSeek, which burst out of obscurity and into the center of the tech universe this week. In the past few days, those execs and many of their peers have addressed questions about the startup lab's new artificial intelligence model, which has stunned experts and was reportedly much more cost effective to create than competitive models in the U.S. DeepSeek's mobile app shot up to the top of the charts on Apple's App Store early in the week and remained in the lead spot as of Friday, ahead of OpenAI's ChatGPT. Reports that its new R1 model, which rivals OpenAI's o1, cost just $6 million to create sent shares of chipmakers Nvidia and Broadcom down 17% on Monday, wiping out a combined $800 billion in market cap. The timing was stark. DeepSeek's rollout landed just as tech earnings season was about to begin, with Meta, Microsoft, Tesla and Apple all reporting between Wednesday and Thursday, and a week into President Donald Trump's second term in office. Trump has emphasized the importance of the U.S. winning in AI, particularly against China, and in his first week back in the White House announced a project called Stargate that calls on OpenAI, Oracle and SoftBank to invest billions dollars to boost domestic AI infrastructure.

Chinese AI app DeepSeek was downloaded by millions of Americans. Deleting it might come next

This week's news that the DeepSeek Chatbot app, developed by China, was downloaded from the Apple app store significantly more times than the US-developed ChatGPT from Open AI, wiped billions off the global tech market. Plenty of Americans are discovering the AI search powers of DeepSeek, the breakthrough Chinese generative AI app that surged to No. 1 downloaded status on Apple's App Store last week. But in an era of U.S.-China technology rivalry and mistrust, and entities from NASA to the U.S. Navy and Taiwanese government prohibiting use of DeepSeek within days, is it wise of millions of Americans to let the app start playing around with their personal search inquiries? The sudden rise of DeepSeek -- created on a rapid timeline and on a budget reportedly much lower than previously thought possible -- caught AI experts off guard, though skepticism over the claims remain and some estimates suggest the Chinese company understated costs by hundreds of millions of dollars. Privacy advocates were caught off guard, too, and their concerns aren't predicated on AI development costs, and they already warning that Americans are putting themselves and their privacy at risk. The amount of data and information that bad actors in China could harvest from DeepSeek is 20 times worse than what could be collected from a Google search, says Dewardric McNeal, managing director and senior policy analyst at risk management firm Longview Global, which advises companies on China strategy. "It is a rich trove of intelligence," said McNeal, who has studied the details of Chinese government data sharing requirements for domestic firms. There are obvious risks, he said, such as personal banking or health information that can be stolen, and prominent cybersecurity firms are already reporting vulnerabilities in DeepSeek. DeepSeek itself reported being hit with a major cyberattack last week. But McNeal is just as worried about the "bigger picture" competition between nations. "I want us to speak broader than just the narrow data; we often don't speak about the degree to which this information paints a mental map through understanding queries," McNeal said. For example, Chinese intelligence could use the broader patterns of queries in DeepSeek to learn about various American industries and to sow division among the public. "The world won't end tomorrow because I logged into DeepSeek," McNeal said, but he added that does not mean there isn't considerable risk involved. The AI's open-source approach, for one, could give China access to US-based supply chains at an industry level, allowing them to learn what companies are doing and better compete against them. "National security professionals are thinking about it in those terms," McNeal said. Matt Pearl, a special advisor to the deputy national security advisor at the National Security Council in the Biden administration and now the Strategic Technologies Program director at the Center for Strategic and International Studies, said DeepSeek's privacy policy implies that people have control over what is collected, but it should induce alarm. "DeepSeek's privacy policy is not worth the paper it is written on," Pearl said. DeepSeek is subjected to PRC laws and anything entered into the app is fair game. Through keystroke patterns, a DeepSeek user can be tracked across all devices, information gathered from advertisers, and DeepSeek could also seek to leverage cameras and microphones, according to Pearl. "If they can do it technically in the app and the PRC determines it is something they want to do, then it poses a danger," Pearl said. But the threat that Pearl said most keeps him up at night is related to cybersecurity and the potential for a mass malware injection. "It is hard to emphasize all the different potential ways in which it could be used. And, in theory, it could be done in a single update to the app," he said. Officials at High Flyer, the Chinese-backed hedge fund which created DeepSeek, did not respond to a request for comment. Despite the outsized impact on the markets and leading AI firms including Nvidia, DeepSeek still has a long way to go to catch up to rival ChatGPT, which is continuing to raise a formidable war chest -- a few days after the DeepSeek headlines dominated the tech and markets news cycle, OpenAI was reportedly in talks for a $40 billion funding round. DeepSeek remains far behind ChatGPT in consumer activity, according to online analytics platform Semrush, with the OpenAI app maintaining an average daily visit count in the tens of millions. But ChatGPT has experienced a recent dip in traffic -- it had 22.1 million visitors on October 1, 2024, but that had declined to 14.9 million by January 19, according to Semrush. At the same time, even before it became a major national news story, DeepSeek's online footprint was growing -- from 2.3K average U.S. daily visits on October 1, 2024, to 71.2K by January 19 (a week before it caused the stock market to tank). Joe Jones, director of research and insights for The International Association of Privacy Professionals, a policy-neutral nonprofit that promotes privacy and AI governance, says that disruptors like DeepSeek can make the organization's job more difficult. "It is challenging for people to do that work when you have proliferating laws that are complex, diverse, and often in tension, and technologies like DeepSeek that come at you from left field, upend status quos and make you rethink good governance," Jones said. The fact that the debate is playing out across borders makes it more contentious. "This has gotten a whole lot more complex in this turbocharged geopolitical environment," Jones added.

What Silicon Valley Leaders Are Saying About China's A.I. Miracle DeepSeek

Influential tech figures like Satya Nadella, Yann LeCun and Dario Amodei shared their takes on the Chinese startup's newfound success. The sudden arrival of breakthrough A.I. models from DeepSeek, a Chinese A.I. startup, has drawn ire, envy and admiration from Silicon Valley's elite. Demonstrating capabilities that rival A.I. leaders like OpenAI, Meta (META) and Anthropic, DeepSeek's models were developed with a fraction of the costs of their U.S. counterparts -- a feat that has caught much of the tech industry by surprise. Sign Up For Our Daily Newsletter Sign Up Thank you for signing up! By clicking submit, you agree to our <a href="http://observermedia.com/terms">terms of service</a> and acknowledge we may use your information to send you emails, product samples, and promotions on this website and other properties. You can opt out anytime. See all of our newsletters DeepSeek's V3 model, released in December and followed by a new reasoning model earlier this month, was created with only 2,000 of Nvidia (NVDA)'s pricey graphics processing units (GPUs), according to the Hangzhou-based startup, which grew out of the hedge fund High-Flyer just last year. The company's self-claimed ability to create advanced A.I. models on the cheap spooked investors betting on American A.I. giants. Nvidia shares tumbled 17 percent today (Jan. 27). Stocks of AMD, Alphabet (GOOGL) and Microsoft (MSFT) also fell. As some tech leaders question whether DeepSeek has more A.I. chips than it's letting on, others advocate for tougher GPU export controls to prevent China from gaining an edge in the new technology. The venture capitalist Marc Andreessen called DeepSeek's breakthrough a modern-day "Sputnik moment." Microsoft's Satya Nadella praised the company's efficiency, and Meta's chief A.I. scientist Yann LeCun lauded its decision to open source. One thing's for sure: DeepSeek's developments have got Silicon Valley talking. Here's a look at what prominent tech figures are saying about the startup's surprising success: Marc Andreessen, co-founder of Andreessen Horowitz Marc Andreessen, general partner of the venture capital firm Andreessen Horowitz, described DeepSeek's R1 reasoning model, released on Jan. 20, as A.I.'s "Sputnik moment" in a post on X yesterday (Jan. 26), referencing how the unexpected success of the Soviet Union's 1957 satellite launch spurred the U.S. to step up its space efforts. DeepSeek-R1 "is one of the most amazing and impressive breakthroughs I've ever seen," added Andreessen, who praised the company's decision to open source the model as "a profound gift to the world." Yann LeCun, chief A.I. scientist at Meta DeepSeek's performance isn't necessarily telling of China's A.I. capabilities surpassing those of the U.S., according to Yann LeCun, but instead highlights the power of open source. "The correct reading is: 'Open source models are surpassing proprietary ones,'" the computer scientist said in a recent post on Meta's Threads. LeCun additionally called the A.I. stock selloff in reaction to DeepSeek "woefully unjustified," noting that much of the money tech companies spend on A.I. isn't necessarily used to train models but to run them. Dario Amodei, CEO of Anthropic For some tech leaders, DeepSeek's breakthrough is a sign that the U.S. should continue cracking down on chip export controls. Dario Amodei recently told CNBC that he believes Chinese A.I. companies have more GPUs than expected, as many were able to procure vast amounts of the hardware before such restrictions were implemented. In the case of DeepSeek, "it's been reported, at least, that they have a cluster of 50,000," said Amodei. According to the Anthropic CEO, export controls must focus on preventing chip stockpiles from growing into the millions. "I think if the United States can't lead in this technology, we're going to be in a very bad place geopolitically," Amodei told CNBC. Marc Benioff, CEO of Salesforce To Marc Benioff, DeepSeek's newfound popularity proves that data, not compute or models, will make the difference for tech companies going forward. "The real treasure of A.I. isn't the [user interface] or the model -- they've become commodities," said Benioff in a post on X yesterday, where he noted that DeepSeek's success was achieved without vast funds or Nvidia chips. "The true value lies in data and metadata, the oxygen fueling A.I.'s potential." Satya Nadella, CEO of Microsoft While speaking with CNBC in Davos earlier this month, Satya Nadella lauded the Chinese startup's progress. "I think we should take the development out of China very, very seriously," said Nadella, who described the DeepSeek models as "super impressive." The CEO followed up on his comments today with an X post seemingly linking DeepSeek to "Jevons paradox," an economic theory that posits heightened efficiency leads to resource consumption instead of reduction. "As A.I. gets more efficient and accessible, we will see its use skyrocket, turning it into a commodity we just can't get enough of," he said.

DeepSeek's 'Sputnik moment' prompts investors to sell big AI players

The DeepSeek app is seen on a mobile phone in Beijing, Jan. 27. AFP-Yonhap Investors hammered technology stocks Monday, sending the likes of Nvidia and Oracle plummeting, as the emergence of a low-cost Chinese artificial intelligence model cast doubts on Western companies' dominance in this sector. Startup DeepSeek last week launched a free assistant it says uses less data at a fraction of the cost of incumbent players' models, possibly marking a turning point in the level of investment needed for AI. Futures on the Nasdaq 100 slid almost 4 percent, suggesting the index could see its biggest daily slide since September 2022 later on, if those losses were sustained. Those on the S&P 500 dropped 2 percent. Shares in AI chipmaker Nvidia fell 10 percent, rival Oracle dropped 8 percent and AI data analytics company Palantir lost 7 percent in pre-market trading. DeepSeek, which by Monday had overtaken U.S. rival ChatGPT in terms of downloads on the Apple Store, offers the prospect of a viable, cheaper AI alternative which has raised questions about the sustainability of the level of spending and investment on AI by Western companies, including Apple and Microsoft. From Tokyo to Amsterdam, shares in AI players tumbled. "We still don't know the details and nothing has been 100 percent confirmed in regard to the claims, but if there truly has been a breakthrough in the cost to train models from $100 million+ to this alleged $6 million number this is actually very positive for productivity and AI end users as cost is obviously much lower meaning lower cost of access," Jon Withaar, a senior portfolio manager at Pictet Asset Management, said. The hype around AI has powered a huge inflow of capital into the equity markets in the last 18 months in particular, as investors have bought into the technology, inflating company valuations and sending stock markets to record highs. Little is known about the small Hangzhou startup behind DeepSeek. Its researchers wrote in a paper last month that the DeepSeek-V3 model, launched Jan. 10, used Nvidia's H800 chips for training, spending less than $6 million - the figure referenced by Pictet's Withaar. H800 chips are not top-of-the-line. Initially developed as a reduced-capability product to get around restrictions on sales to China, they were subsequently banned by U.S. sanctions. The DeepSeek app is seen on a mobile phone in Beijing, Jan. 27. AFP-Yonhap 'Sputnik moment' Marc Andreessen, the Silicon Valley venture capitalist, said in a post on X Sunday that DeepSeek's R1 model was AI's "Sputnik moment", referencing the former Soviet Union's launch of a satellite that marked the start of the space race in the late 1950s. "Deepseek R1 is one of the most amazing and impressive breakthroughs I've ever seen -- and as open source, a profound gift to the world," he said in a separate post. In Europe, ASML which counts Taiwan's TSMC , Intel and Samsung as its customers, dropped almost 11 percent, while in Japan, startup investor SoftBank Group slid more than 8 percent. Last week it announced a $19 billion commitment to fund Stargate, a datacenter joint venture with OpenAI. Big Tech has ramped up spending on developing AI capabilities and optimism over the possible returns has driven stock valuations sky-high. Nvidia alone has risen by over 200 percent in about 18 months and trades at 56 times the value of its earnings, compared with a 53 percent rise in the Nasdaq, which trades at a multiple of 16 to the value of its constituents' earnings, according to LSEG data. Nick Ferres, chief investment officer at Vantage Point Asset Management in Singapore said the market was questioning the capex spend of the major tech companies. Masahiro Ichikawa, chief market strategist at Sumitomo Mitsui DS Asset Management said: "The idea that the most cutting-edge technologies in America, like Nvidia and ChatGPT, are the most superior globally, there's concern that this perspective might start to change." "I think it might be a bit premature," Ichikawa said. (Reuters)

The DeepSeek effect: Why China's ability to sneak past US barriers should be a wake up call for Indian IT giants

Author of Running with the Dragon: How India Should Do Business with China DeepSeek is making US experts and businessmen look like defeated soldiers. The model, which many regard as a superior version of ChatGPT, has also dealt a body blow to the reputation of Indian IT prowess. The AI platform represents a lot more than a tech advancement. It will have strong political and security implications at a time when Donald Trump has set out to reshape global trade in a manner that suits him. Washington's efforts to slow down the growth of AI tech in China by blocking the export of Nvidia chips did not work, as DeepSeek managed to get around it. China Daily said that the company, established in 2023, used as few as 2,048 H800 GPUs from Nvidia. But reports in tech portals noted that the company had access to many more chips that got past gov controls. 'DeepSeek didn't come out of nowhere - they've been at model-building for years,' Jimmy Goodrich, a senior adviser to RAND Corp, said. 'It's been long known that DeepSeek has a really good team, and if they had access to even more compute, God knows how capable they would be.' DeepSeek, which is available to developers at low prices compared to rivals like ChatGPT, is set to become a game changer. Thousands of Western app developers and software firms might move to the new platform made in China, which is accused of indulging in cybertheft. This is a paradigm shift that no one imagined until a few days ago. What happens after the Chinese company makes Western developers dependent on its platform is something Americans need to worry about. The platform represents a huge source of useful data for DeepSeek, and possibly Chinese authorities. There are signs that the Chinese gov controls DeepSeek, which refuses to answer critical questions about Chinese leaders and events like the Tiananmen massacre. The popularity of a Chinese AI model in foreign countries might give Beijing some military advantage in applications like simulations, geopolitical analysis and real-time decision-making. After the bloodbath on Wall Street, many expected Trump to react sharply and possibly threaten to ban DeepSeek. But Trump, who recently saved TikTok, described the development as a 'wake-up call' for American business and said that he saw the emergence of 'a much less expensive method' as 'a positive, as an asset'. The Chinese platform, which is overwhelmed by requests from new users trying to download it, will require more Nvidia chips, preferably the latest version, to expand its base. The question is whether US will allow the supply of more chips to the firm. A Wired report said that DeepSeek is collecting and sending data to the Chinese authorities, pointing out, 'The English-language DeepSeek privacy policy, which lays out how the company handles user data, is unequivocal: 'We store the information we collect in secure servers located in the People's Republic of China'.' China's ability to sneak past US barriers and give the likes of OpenAI, Meta and Google a run for their money should be a lesson for Indian IT companies, which have been reluctant to invest much in R&D, blaming high costs and insufficient data required for training a new model. Now that the Chinese have reduced the cost of using an AI model to a fraction, will this enthuse Indian firms to participate in the development of AI applications? The first signs are not encouraging. TCS CEO-MD K Krithivasan indicated that most Indian IT companies are 'system integrators' and may not be keen on building their own LLMs for AI. 'I don't think there's going to be a huge incremental advantage in building your own LLMs since there are so many already available,' he said. Indian companies use the products as software to serve their customers' needs. 'If you take a similar approach here, we are better off - at least in the short run - using the LLMs available, but ensuring that enterprises are able to use them to really get the value,' he said. The DeepSeek episode shows that GoI needs to provide sufficient incentives to IT companies so that they can think big. GoI can also provide greater access to its data to help them train AI models. Is this the beginning of a bigger Chinese role or the introduction of censorship in the AI universe? (The writer is author of Running with the Dragon: How India Should Do Business with China)

DeepSeek hysteria could quickly fade as skepticism on Wall Street...

The markets want you to think the DeepSeek freak-out is an immediate existential threat to US tech and AI dominance. But don't be so sure DeepSeek isn't more of a Deep Fake from an investor's perspective. Recall why exactly the markets imploded Monday when DeepSeek mania began to make its way around social media on some accounts of savvy influencers like Marc Andreessen: The story goes that some hedge fund trader in China created a hyper competitive artificial intelligence platform on a shoe-string budget. Big AI outfits are spending boatloads of money on R&D, and the chips needed to support AI infrastructure. That's why chip-maker Nvidia has been such a market darling, and why if DeepSeek can do it better and for less, its shares tanked Monday, losing $500 billion in market value before I finished my lunch. In the coming weeks, Wall Street tech analysts will be unspooling how the Chinese company was able to "leapfrog" US tech giants in AI. Here's why this story is so treacherous from a standpoint if you're betting against US AI and companies like Nvidia. The Chinese are better known for their ability to borrow stuff from the US than innovate on their own. It's why Trump, during his first term, engaged in the trade war after hearing complaints from big companies in the US that the cost of admission in doing business on the Mainland and with China's massive consumer market, is sharing intellectual property with the ruling Chinese communist party. Every China-based company is controlled by the CCP; China's ByteDance is on the verge of selling its popular short-video app TikTok or it faces a ban from US app stores, because of the CCP. I have no idea how DeepSeek got to where it is other than what I'm reading, and it may indeed be a miracle of the first order. It was started by a math genius hedge fund trader named Liang Wenfeng who "outsmarted" US tech giants (as a headline in the Wall Street Journal stated). He couldn't get those great US chips because of many reasons, including they're very expensive, so he innovated on his home turf and came up with a better, faster AI model on the cheap. You can see why this would signal bad things for US tech stocks if this is in fact true, and why the freakout is about to reverse itself if this story doesn't hold. Some Wall Street analysts and tech companies like Microsoft and OpenAI are investigating if DeepSeek is actually using technology or chips from Jensen Huang's Nvidia in a substantial way, or if it was substantially financed by the Communist Chinese government to do so. Noted tech analyst Dan Ives is one who doesn't quite buy the DeepSeek rags-to-riches story. He doesn't dispute the quality of the product, just that it's in his opinion impossible to create something that good with a mere $6 million investment story that is being sold to the markets. Markets are now factoring in such skepticism as witnessed by Nvidia's Tuesday rebound along with other tech names. And if Ives is right, and there's more to DeepSeek than meets the eyes, what just went down, will now go up even more. A DeepSeek rep didn't return a request for comment.

China's DeepSeek AI rattles Wall Street, but questions remain

Chinese researchers backed by a Hangzhou-based hedge fund recently released a new version of a large language model (LLM) called DeepSeek-R1 that rivals the capabilities of the most advanced U.S.-built products but reportedly does so with fewer computing resources and at much lower cost. High Flyer, the hedge fund that backs DeepSeek, said that the model nearly matches the performance of LLMs built by U.S. firms like OpenAI, Google and Meta, but does so using only about 2,000 older generation computer chips manufactured by U.S.-based industry leader Nvidia while costing only about $6 million worth of computing power to train. By comparison, Meta's AI system, Llama, uses about 16,000 chips, and reportedly costs Meta vastly more money to train. Open-source model The apparent advance in Chinese AI capabilities comes after years of efforts by the U.S. government to restrict China's access to advanced semiconductors and the equipment used to manufacture them. Over the past two years, under President Joe Biden, the U.S. put multiple export control measures in place with the specific aim of throttling China's progress on AI development. DeepSeek appears to have innovated its way to some of its success, developing new and more efficient algorithms that allow the chips in the system to communicate with each other more effectively, thereby improving performance. At least some of what DeepSeek R1's developers did to improve its performance is visible to observers outside the company, because the model is open source, meaning that the algorithms it uses to answer queries are public. Market reaction The news about DeepSeek's capabilities sparked a broad sell-off of technology stocks on U.S. markets on Monday, as investors began to question whether U.S. companies' well-publicized plans to invest hundreds of billions of dollars in AI data centers and other infrastructure would preserve their dominance in the field. When the markets closed on Monday, the tech-heavy Nasdaq index was down by 3.1%, and Nvidia's share price had plummeted by nearly 17%. However, not all AI experts believe the markets' reaction to the release of DeepSeek R1 is justified, or that the claims about the model's development should be taken at face value. Mel Morris, CEO of U.K.-based Corpora.ai, an AI research engine, told VOA that while DeepSeek is an impressive piece of technology, he believes the market reaction has been excessive and that more information is needed to accurately judge the impact DeepSeek will have on the AI market. "There's always an overreaction to things, and there is today, so let's just step back and analyze what we're seeing here," Morris said. "Firstly, we have no real understanding of exactly what the cost was or the time scale involved in building this product. We just don't know. ... They claim that it's significantly cheaper and more efficient, but we have no proof of that." Morris said that while DeepSeek's performance may be comparable to that of OpenAI products, "I've not seen anything yet that convinces me that they've actually cracked the quantum step in the cost of operating these sorts of models." Doubts about origins Lennart Heim, a data scientist with the RAND Corporation, told VOA that while it is plain that DeepSeek R1 benefits from innovative algorithms that boost its performance, he agreed that the general public actually knows relatively little about how the underlying technology was developed. Heim said that it is unclear whether the $6 million training cost cited by High Flyer actually covers the whole of the company's expenditures -- including personnel, training data costs and other factors -- or is just an estimate of what a final training "run" would have cost in terms of raw computing power. If the latter, Heim said, the figure is comparable to the costs incurred by better U.S. models. He also questioned the assertion that DeepSeek was developed with only 2,000 chips. In a blog post written over the weekend, he noted that the company is believed to have existing operations with tens of thousands of Nvidia chips that could have been used to do the work necessary to develop a model that is capable of running on just 2,000. "This extensive compute access was likely crucial for developing their efficiency techniques through trial and error and for serving their models to customers," he wrote. He also pointed out that the company's decision to release version R1 of its LLM last week -- on the heels of the inauguration of a new U.S. president -- appeared political in nature. He said that it was "clearly intended to rattle the public's confidence in the United States' AI leadership during a pivotal moment in U.S. policy." Dean W. Ball, a research fellow at George Mason University's Mercatus Center, was also cautious about declaring that DeepSeek R1 has somehow upended the AI landscape. "I think Silicon Valley and Wall Street are overreacting to some extent," he told VOA. "But at the end of the day, R1 means that the competition between the U.S. and China is likely to remain fierce, and that we need to take it seriously." Export control debate The apparent success of DeepSeek has been used as evidence by some experts to suggest that the export controls put in place under the Biden administration may not have had the intended effects. "At a minimum, this suggests that U.S. approaches to AI and export controls may not be as effective as proponents claim," Paul Triolo, a partner with DGA-Albright Stonebridge Group, told VOA. "The availability of very good but not cutting-edge GPUs -- for example, that a company like DeepSeek can optimize for specific training and inference workloads -- suggests that the focus of export controls on the most advanced hardware and models may be misplaced," Triolo said. "That said, it remains unclear how DeepSeek will be able to keep pace with global leaders such as OpenAI, Google, Anthropic, Mistral, Meta and others that will continue to have access to the best hardware systems." Other experts, however, argued that export controls have simply not been in place long enough to show results. Sam Bresnick, a research fellow at Georgetown's University's Center for Security and Emerging Technology told VOA that it would be "very premature" to call the measures a failure. "The CEO of DeepSeek has gone on record saying the biggest constraint they face is access to high-level compute resources," Bresnick said. "If [DeepSeek] had as much compute at their fingertips as Google, Microsoft, OpenAI, etc, there would be a significant boost in their performance. So ... I don't think that DeepSeek is the smoking gun that some people are claiming it is [to show that export controls] do not work." Bresnick noted that the toughest export controls were imposed in only 2023, meaning that their effects may just be starting to be felt. He said that the real test of their effectiveness will be whether U.S. firms are able to continue to outpace China in coming years.

Tech selloff deepens as DeepSeek triggers AI rethink

The logo of DeepSeek is displayed alongside its AI assistant app on a mobile phone, in this illustration picture taken Jan. 28. Reuters-Yonhap Japanese technology shares fell on Tuesday as a global market rout sparked by the emergence of a low-cost Chinese artificial intelligence model entered day two, with investors questioning the sky-high valuation and dominance of AI bellwethers. Shares of Nvidia, the poster child of the AI boom in recent years, dragged U.S. stocks lower, sinking 17 percent on Monday and wiping $593 billion from the chipmaker's market value, a record one-day loss for any company. It all stemmed from a free AI assistant launched by Chinese startup DeepSeek last week that the firm said uses less data at a fraction of the cost of services available currently. That garnered attention worldwide, although skepticism lingers. OpenAI CEO Sam Altman called it an "impressive model." "We will obviously deliver much better models and also it's legit invigorating to have a new competitor!," Altman, the head of the AI firm behind ChatGPT, said in a social media post. The launch and increasing popularity of DeepSeek spurred investors to dump tech stocks globally, with ripples felt from Tokyo to Amsterdam to Silicon Valley. Markets in tech-heavy South Korea and Taiwan are closed for the next few days for Lunar New Year. Mainland China is closed until Feb. 4, leaving the spotlight firmly on Japanese firms. On Tuesday, chip-testing equipment maker Advantest, a supplier to Nvidia lost 10 percent after diving nearly 9 percent on Monday. Chip-making equipment maker Tokyo Electron and technology start-up investor SoftBank Group slid 5 percent. "It's clearly a sell first, ask questions later approach, and we've actually seen that kind of move in the past in Japan," said Kei Okamura, a portfolio manager at Neuberger Berman, referring to a global market meltdown in August headlined by Japan's Nikkei. Over in the U.S., Broadcom finished down 17.4 percent, while ChatGPT backer Microsoft fell 2.1 percent and Google parent Alphabet closed down 4.2 percent. The Philadelphia semiconductor index tumbled 9.2 percent, for its deepest percentage drop since March 2020. No margin of error The selloff has brought into the spotlight the crowded positioning among investors and the billions of dollars U.S. tech giants are pouring in to develop AI capabilities, as well as the extremely high valuation of some of these firms. "What makes Monday's tech selloff so jarring is that the valuations of many of these AI and tech companies offer no margin of error," said David Bahnsen, chief investment officer at The Bahnsen Group. "The excessive weighting these tech stocks have in many investor portfolios and the high concentration these tech stocks have in the market indices was a significant and under-appreciated risk issue." The hype around AI has powered a huge flow of capital into equities, inflating valuations and lifting stock markets to record highs, leading to an increase of around $10 trillion in the market value of "Magnificent Seven" companies since ChatGPT kicked off the AI boom in November 2022. It is not just the chipmakers and tech companies but companies focused on datacenters also taking a hit, with Malaysia's utility conglomerate YTL Power down 9 percent on Tuesday, its third session of steep loss. "We're still, like many investors, gathering information," said Neuberger Berman's Okamura, noting that a lot of investors are scrambling to gather more information and decide their next move. "I think we're going to see many more of these (developments) going forward. And we've seen technological advancements like this that have had implications for cost spend." Investor focus will be on the flurry of tech earnings this week, with executives likely keen to calm frayed nerves and ease concerns about capital spending. This photo illustration shows the DeepSeek and Nvidia logos on screens in Hangzhou, in China's eastern Zhejiang province, Jan. 27. Fears of upheaval in the AI gold rush rocked Wall Street, following the emergence of a popular ChatGPT-like model from China, with U.S. President Donald Trump saying it was a "wake-up call" for Silicon Valley. AFP-Yonhap AI race Little is known about the Hangzhou startup behind DeepSeek, whose controlling shareholder is Liang Wenfeng, co-founder of quantitative hedge fund High-Flyer, records showed. Its researchers wrote in a paper last month that its DeepSeek-V3 model, launched on Jan. 10, used Nvidia's lower-capability H800 chips for training, at a cost of less than $6 million. The launch and the popularity of the app contrasts with the lackluster reception that met the Chinese ChatGPT equivalent made by search engine giant Baidu, which exposed the gap in AI capabilities between U.S. and Chinese firms. The quality and cost efficiency of DeepSeek's models have flipped this narrative on its head and spurred a warning from U.S. President Donald Trump, who called it "a wakeup call for our industries." Japan's digital minister Masaaki Taira said DeepSeek's emergence had upended conventional wisdom that Chinese AI was years behind. "It's been said that Chinese generative AI might be about five years behind, but that turned out to be wrong and it seems to be on a fairly good track," Taira said, adding that Japan was taking a closer look into suggestions that Chinese AI may be more cost effective. (Reuters)

DeepSeek Isn't a Reason for Big Tech to Become More Powerful

Tom Wheeler was chairman of the Federal Communications Commission from 2013 to 2017. He is currently a senior fellow at Harvard Kennedy School Shorenstein Center and a visiting fellow at the Brookings Institution. He is also the author of, TECHLASH: Who Makes the Rules in the New Gilded Age. The Chinese artificial intelligence (AI) lab DeepSeek grabbed headlines and tanked the stock market with its announcement of a new AI model nearly equivalent to the United States' most recent reasoning models but at a fraction of the cost. If the past is prologue, the DeepSeek development will be seized upon by some as rationale for eliminating domestic oversight and allowing Big Tech to become more powerful. Such a response is exactly the opposite of what America needs. If anything, DeepSeek proves the importance of protecting American innovation by promoting American competition. A typical Silicon Valley argument has been that allowing big companies to gobble up smaller rivals allows the incredible resources of those big companies to drive the AI race forward and protect American interests. Such a thesis conveniently overlooks that the breakthroughs of DeepSeek, OpenAI, and Anthropic were breakthroughs from disruptive startups, not national champions. "We often hear that pursuing antitrust cases against or regulating [dominant tech] firms will weaken American innovation and cede the global stage to China," former Federal Trade Commission (FTC) chair Lina Khan has said. But "history and experience show that lumbering monopolies...cannot deliver the breakthrough technological advancements that hungry startups tend to create...To stay ahead globally, we don't need to protect our monopolies from innovation - we need to protect innovation from our monopolies. We need to choose competition over national champions." Not only are big companies lumbering but cutting-edge innovations often conflict with corporate interest. According to The Wall Street Journal, Google engineers had built a generative AI chatbot over two years before OpenAI unveiled ChatGPT. Despite entreaties from its engineers, Google management sat on the breakthrough. One theory is that the ability to ask an AI chatbot a question and receive an answer threatened Google search, the company's cash cow. "Gmail creator warns Google is 'only a year or two away from total disruption' because of AI like ChatGPT," one headline proclaimed. An incumbent like Google -- especially a dominant incumbent -- must continually measure the impact of new technology it may be developing on its existing business. The company's first fiduciary responsibility is to shareholders, not to the expansion of knowledge. An innovative startup such as OpenAI, however, has no such qualms. "If it's going to happen anyway, it seems like it would be good for someone other than Google to do it first," OpenAI's CEO Sam Altman wrote in an email to co-founder Elon Musk. A strategy of "national champions" is at the heart of China's AI policies. The Chinese government anointed big companies such as Baidu, Tencent, and Alibaba.Then, the Chinese government subsidized them with cash and helpful policies. Yet DeepSeek was built, not by these favored companies but by a hedge fund that originally started using AI for trading decisions. DeepSeek's breakthrough took advantage of the commoditization of AI. There are over one million open-source models freely available on the Hugging Face open-source repository. These open-source models, built on breakthroughs in the original foundation models, are free to be modified and developed as the user sees fit. DeepSeek studied those open-source models, trained their own model, and optimized it to use less computing power. Then, they open-sourced their breakthrough to make it available to everyone. The evolution of AI from amazing proprietary capabilities to an openly available commodity is a watershed that will enable the proliferation of innovation, not just in the foundation models, but in the widespread application of the technology. The lesson of history is that it is not the primary technology that is transformative, but its secondary applications. The open availability of a low-cost, low-compute model opens the door to the Jevons paradox, an economic principle that increased efficiency leads to greater overall consumption rather than a reduction. As Microsoft CEO Satya Nadella posted on X after the DeepSeek announcement, "Jevons paradox strikes again!" The AI race has now begun its second lap. Model development will continue to be important, but the future lies in what easily available AI will enable. As Jack Clark, co-founder of Anthropic, explained, "DeepSeek means AI proliferation is guaranteed." The dispersal of AI applications in the United States is driven by for-profit enterprises seeking to gain a competitive advantage. Its opportunities are virtually boundless in areas as diverse as healthcare, consumer tech, finance, or farming. This second leg of the AI race, however, requires the maintenance of an open marketplace environment that avoids innovations being gobbled up by the kind of market dominating power that characterized the last quarter century. The Big Tech companies, often looked to as America's national champions, have become big through anticompetitive activities. The Department of Justice and multiple state attorneys general sued Google for violating antitrust laws to dominate the search market (and won.) They also sued Google's online advertising market and expects a decision soon. The FTC has sued Meta alleging an unlawful effort to maintain their social media monopoly by acquiring Instagram and WhatsApp, a case that is expected to go to trial in April. If we are concerned about the AI race with China, we need to focus less on lobbying to let the big guys get bigger, and more on making sure there are competitive opportunities to spur innovation. As today's AI developers mature and as AI disperses into applications, the historical lesson remains critical: unchecked consolidation of power stifles the innovation necessary for economic growth, national security, and consumer protection. Ensuring a competitive market drives innovation. This includes not only antitrust enforcement, but also sectoral regulation built on promoting competition while providing consumer protection guardrails. Innovative competition also requires support for the innovators. The venture capitalist model predicated on the sale of the startup to a dominant company is broken. Investing with the goal of ultimately consolidating the new competition into existing powerhouses may maximize VC returns, but does not maximize returns to the public interest. Building the competition necessary for a vibrant AI market requires alternative support vehicles for innovators. Such support could include initiatives by the Small Business Administration and tax policies. At the same time, easing the path for initial public offerings could provide an alternative exit strategy for those who do invest. Innovation proliferation also proliferates the risks of existential harm from unsupervised AI. That a hedge fund can revolutionize the AI world seemingly overnight highlights the importance of assuring the headlong rush forward includes behavioral expectations and guardrails. President Trump's ill-advised repeal of President Biden's executive order removed some of the safeguards against runaway AI. Positive AI developments require balancing open source technology with safety standards and the enforceable expectation they will be followed. This includes red teams to actively seek problems in new models and report their findings. Such standards and behaviors require audits at least to the level we require for the financial system. Ultimately, the scare headlines that a new Chinese AI model threatens America's AI dominance is just that -- a scare headline. If appropriately applied in a competitive marketplace, the DeepSeek development is a roadmap for continued American leadership. Built on U.S. technology, it commoditizes AI and accelerates the race to to disperse AI throughout the economy of the world. Such threats and promises require oversight. A competitive market that will incentivize innovation, must be accompanied by common sense guardrails to protect against the technology's runaway potential. The arrival of DeepSeek shows that competition works; it represents an opportunity for the United States to continue its AI leadership.

The DeepSeek mania proves it's finally -- finally! -- time to talk about AI privacy

It's been precisely one week since a ChatGPT rival surprised the AI world. DeepSeek, a China-based company, unleashed its R1 model and is suddenly at the center of the AI world's attention -- for good reason. On top of being seemingly as sophisticated as OpenAI's o1 model of ChatGPT, DeepSeek's R1 model is free. Yes, that's right... zero dollars. That proposition, coupled with the fact that it was seemingly developed for a small fraction of the cost of other splashy LLMs, has sent ripples throughout the tech world, especially from the financial side. In just one day, Nividia saw $600 billion knocked off its market cap, marking the most significant one-day loss in market history -- that's less of a ripple and more of a tidal wave. And while all of that is impressive, DeepSeek might be having another unintentional impact that's just as big, and it's got almost nothing to do with the stock market. While DeepSeek's ability and price have dominated the conversation, there's another throughline that's decidedly less flattering for the company -- privacy. One of the main differences between DeepSeek and the rest of the LLM field right now is that it's not American or, more specifically, based in China. For obvious TikTok-related reasons, that's a bad thing for an American app to be right now. The idea of an app that Hoovers up your data and sends it to a server in China is problematic for many reasons. One is that privacy -- whether for an American app or a foreign one -- is essential. The more your personal data is collected, the higher the risk is to your digital safety. Secondly, there's little difference between a private company and the government in China, which raises the question of what strictly user data could be used for. That argument is for another day. Still, it doesn't take much extrapolation to see how a potentially adversarial government could use massive data troves for purposes that aren't in Americans' interest. And in many ways, none of that is surprising. Of course, a non-American coming in and wiping out the value of a titan like Nvidia will make some waves, and, of course, that will dredge up skepticism. But what I find more interesting isn't what DeepSeek says about a Chinese AI company but about AI broadly. It didn't take long for DeepSeek's R1 to find controversy. Just a day after its ascension into the public conversation, Bloomberg and the Financial Times reported that Microsoft and OpenAI are investigating whether DeepSeek used less-than-scrupulous methods to train R1. Bloomberg reports: "Microsoft Corp. and OpenAI are investigating whether data output from OpenAI's technology was obtained in an unauthorized manner by a group linked to Chinese artificial intelligence startup DeepSeek, according to people familiar with the matter." Naturally, OpenAI seems less than enthusiastic about the idea that DeepSeek copied the company's homework to build its R1 model. As many have pointed out, that aversion is blatant hypocrisy. OpenAI notoriously trains its algorithm using copyrighted material and other people's intellectual property. It is currently in a legal battle with the New York Times over using the paper's content to train ChatGPT. That's not even addressing privacy concerns with whatever data OpenAI stores about inputs from its users or anyone who signs up to use its platform. And there's a distinction between data scraped by a private company in the U.S. and one in China, where the lines between private and public entities are incredibly blurry. Still, all of that is almost irrelevant because DeepSeek, whether meaning to or not, is making us talk about AI privacy. Every new digital platform has some privacy pitfalls. Consider the rise of social media, voice assistants, and e-commerce. However, the difference between those platforms and the AI platforms created before us is that we now have years of history from which to draw. That means, at least theoretically, things could be different. I say theoretically because to protect your privacy, history says you must fight for it. For example, voice assistants like Alexa, Siri, or Google Assistant didn't become opt-in until scandals emerged over recording programs that inadvertently captured sensitive conversations. After whistleblowers leaked details of those programs, tech giants quickly rolled them back or at least gave users the option to opt-out. We aren't there yet, but moments like DeepSeek bring us closer to putting privacy on the map. However, we'll likely have to learn the hard way. On Thursday, security researchers found that DeepSeek was storing millions of log lines in an unsecured database accessible without authentication. That information could theoretically give bad actors access to DeepSeek's internal systems, which is not great. But if it could happen to DeepSeek, it could happen to anyone. Therefore, it's more apparent than ever that the time to have an AI privacy conversation is now. Whether that happens is anyone's guess, but when all of our data ends up in the hands of someone who doesn't have users' well-being in mind, we'll at least know that we should have seen it coming.

DeepSeek shows China playbook to deal an even bigger shock to US

DeepSeek's new AI model showcases China's potential to achieve technological breakthroughs despite US export curbs. Leveraging a vast pool of IT talent and government support, the firm signifies China's innovative edge. However, challenges remain in the advanced chipmaking domain, indicating the need for persistent efforts to stay competitive on the global stage.The success of DeepSeek's new AI model points to how China might eventually achieve an even bigger technological breakthrough in the face of US export curbs: Producing its own cutting-edge chips. As tech leaders and politicians marvel at DeepSeek's apparent ability to build an innovative AI model without spending nearly as much as rivals in the US, a development that roiled markets on Monday, the question now is how exactly the Hangzhou-founded company pulled it off -- and what it means for American efforts to stay ahead of China in the tech race. While much remains unclear, such as the exact chips DeepSeek used and whether it has enough on hand to further develop its AI models, its success highlights some of China's key advantages. The country has a deep pool of highly skilled software engineers, a vast domestic market and government support in the form of subsidies as well as funding for research institutes. It also has a pressing necessity to find a way to do more with fewer resources. "China has an obvious advantage in IT talent, both in terms of the sheer number and labor cost," said Liu Xu, a research fellow at the National Strategy Institute of Tsinghua University, which is helping to spearhead China's AI push. "The biggest resource for China is the country's vast demand -- as one of the world's most populous nations and as a gigantic manufacturing hub." In many ways, DeepSeek is an example of China at its best. From Tencent Holdings Ltd.'s WeChat to ByteDance Ltd.'s TikTok, Chinese developers have crafted world-leading apps for consumers, pioneering new modes of communication and ecommerce along the way. DeepSeek's R1 model is now set to challenge similar products from OpenAI, Alphabet Inc.'s Google and Meta Platforms Inc. China, however, is still playing catchup in the hardware space -- the main focus of US export controls in recent years. The US has banned China from buying the most advanced AI chips from Nvidia Corp. and Advanced Micro Devices Inc. It has also blocked President Xi Jinping's government from obtaining ASML Holding NV's extreme ultraviolet lithography (EUV) machines, which are essential to producing high-end chips. President Xi has poured billions of dollars into making a breakthrough in advanced semiconductors, part of his wider Made in China 2025 push to make the nation a leader in emerging technologies. Research by Bloomberg Economics and Bloomberg Intelligence shows that the broader effort has largely been a success, pushing China's manufacturing prowess to historic heights as the nation becomes dominant in producing goods like electric vehicles, batteries and solar panels. Although Chinese companies such as Huawei Technologies Co. have made progress in producing AI chips, as of now those parts are less powerful than the ones produced by Nvidia. And the lack of access to high-end chips, particularly after the Biden administration tightened trade controls, still poses a major hurdle to China's development. In 2023, Huawei introduced a smartphone with a 7-nanometer chip -- something the US thought was unfeasible for Chinese firms to manufacture with less-advanced chipmaking technology. Yet while that phone was hailed as a breakthrough -- and spurred patriotic memes on social media as its release coincided with then-Commerce Secretary Gina Raimondo's visit to the country -- since then the firm has struggled to make advancements toward 5nm chips, while rivals like Apple Inc. and Samsung Electronics Co. have moved on to 3nm. It remains uncertain if DeepSeek, which released its new model just as Donald Trump was inaugurated, will be able to continue accessing the high-end chips that led to its current product. Liang Wenfeng, the company's 40-year-old founder, has cited US export controls as a hindrance to the company's future growth. On Monday, President Trump said DeepSeek's R1 release "should be a wake-up call for our industries that we need to be laser-focused on competing to win." He also lauded the model as a "positive development" that could allow for less expensive AI advancements across the board. While DeepSeek's advance is positive for China's economic transition away from property toward high-tech growth drivers, it may also spur American lawmakers to redouble efforts to stop the nation from getting the most advanced technology. The same data security concerns that led legislators to take action against TikTok may also lead to similar action against DeepSeek. Michelle Giuda, chief executive officer at the Krach Institute for Tech Diplomacy at Purdue, told Bloomberg TV that it was essential for the US to maintain "a really strong defense" with tighter export controls. At the same time, she added, the US needs to catch up with China in producing talented engineers if it's going to stay ahead in the tech race. "It's a game changer in the sense that all we should do is double down American efforts to move faster, smarter, and maybe more cheaply on how we innovate in artificial intelligence," she said, adding that China graduated more than twice as many engineers as the US in 2020. "Unless we have the engineers to build and design the data centers, to design more advanced AI, we are not gonna be able to be the world capital of AI." For now, it looks like China remains several generations behind in chipmaking equipment. Beijing recently advised state-linked organizations to use a new homemade lithography machine with a resolution of 65nm or better -- far from the 8nm resolution of ASML's best machines. Still, China recently publicized a patent application from Shanghai Micro Electronics Equipment Group Co., known as SMEE, for an EUV lithography machine -- which, if it came to market, would be the only one in the world to compete with those produced by ASML. After DeepSeek's revelation, China has shown it has the potential to continue surprising the world with even bigger breakthroughs. "Bunch of talented 22-year-olds without access to the world's best chips, without access to Nvidia chips, seem to have created something that's even better than even the best companies in the Western world have done," Tony Haymet, the Australian government's incoming chief scientist, told reporters on Tuesday in Canberra. "It shows you how disruptive technology can be," he added. "And how quickly things can happen."

China rejoice DeepSeek's global rise, send veiled message to US, Donald Trump

DeepSeek's success has become a national toast in China. DeepSeek last week launched a free AI assistant.Chinese bloggers, state media and local citizens have welcomed DeepSeek's global success with pride and glee, with some saying the homegrown AI startup's meteoric rise is a sign China is beating back Washington's attempts to contain the country's tech industry, as per a report. DeepSeek last week launched a free AI assistant that it says uses less data at a fraction of the cost of incumbent services. By Monday, it had overtaken U.S. rival ChatGPT in downloads from Apple's App Store, triggering a global selloff in tech shares, Reuters report. The Chinese company's apparent ability to match OpenAI's capabilities at a much lower cost has posed questions over the sustainability of the business models and profit margins of U.S. AI giants such as Nvidia and Microsoft. In China, it has raised hopes that the country can successfully resist Washington's export controls targeting access to cutting-edge semiconductors. "This also symbolises U.S. containment, persecution, and sanctions against China in the field of advanced technology has completely failed," military affairs commentator Chen Xi wrote on his WeChat account on Wednesday. U.S. President Donald Trump said on Monday that DeepSeek's technology should act as a spur for American companies and it was good that Chinese firms had come up with a cheaper, faster method of artificial intelligence. The provincial government's media office in Zhejiang, where DeepSeek is based, published a lengthy essay on Wednesday that quickly went viral and was read more than 100,000 times. "The moon overseas is not actually more round, whatever others can do, we can also do it and even do it better," the essay said, while criticising online voices that were both overly triumphant and overly pessimistic about China's technological development. "We need to leave the narrow prism of triumphalism," the department argued. Still, the sentiment around DeepSeek echoes public reaction to Huawei's 2023 surprise release of its high-end Mate 60 Pro smartphone during a visit by then U.S. Commerce Secretary Gina Raimondo, who led the Biden Administration's efforts to restrict Chinese access to high-end AI chips. At the time, the state-backed Global Times said that Huawei's ability to produce a high-end smartphone despite years of targeted U.S. sanctions showed Washington had failed in its "extreme crackdown" on China. Chen Jianuo, a 38-year-old employee at a sustainable development magazine in Beijing, said she felt proud of DeepSeek's popularity overseas after noticing it was a trending topic on Chinese social media platform Weibo. "China has made great progress in the development of artificial intelligence, and I hope that the technological development of our country will get better and better," she said. Leo Li, a 24-year-old student, said that he was happy a Chinese company could be on a par with the likes of Meta and OpenAI and that he would consider using DeepSeek's AI tools. Q1. When did DeepSeek launch a free AI assistant? A1. DeepSeek last week launched a free AI assistant that it says uses less data at a fraction of the cost of incumbent services. By Monday, it had overtaken U.S. rival ChatGPT in downloads from Apple's App Store, triggering a global selloff in tech shares, Reuters report. Q2. Why has DeepSeek become a talking point? A2. The Chinese company's apparent ability to match OpenAI's capabilities at a much lower cost has posed questions over the sustainability of the business models and profit margins of U.S. AI giants such as Nvidia and Microsoft.

How DeepSeek stunned the AI industry - podcast

Why is the US technology industry worried about Chinese company DeepSeek? Robert Booth reports DeepSeek, the Chinese company behind the new AI chatbot R1, uses less computing power and fewer chips than its rivals, and claims the model is far cheaper. "It's sort of the biggest news in this space of AI chatbots since November 2022 when ChatGPT came out," Robert Booth, the Guardian's UK technology editor, tells Helen Pidd. "What they seem to have done is that they've worked out a way of doing more with less," Robert explains. "What they're doing then is activating only the relevant chips when you put your search in, rather than all of them, so it's answering more efficiently." Robert and Helen also discuss concerns over the model's security and censorship, and why the US AI industry has been shaken by its launch.

DeepSeek A.I. Is a Win for China, but a Danger to Party Control

In 2017, China watched in awe -- and shock -- as AlphaGo, an artificial intelligence program backed by Google, defeated a Chinese prodigy at a complex board game, Go. The decisive loss to a foreign computer program, which had similarly trounced a South Korean player, was a sort of Sputnik moment for China. That year, Chinese officials laid out a bold plan to lead the world in A.I. by 2030, pledging billions to companies and researchers focused on the technology. From this fervor emerged DeepSeek, the largely unknown Chinese start-up that upended the technology landscape by creating a powerful A.I. model with far less money than experts had thought possible. DeepSeek is private, with no apparent state backing, but its success embodies the ambitions of China's top leader, Xi Jinping, who has exhorted his country to "occupy the commanding heights" of technology. Mr. Xi wants the Chinese economy to be powered not by old growth engines like debt-fueled real estate and cheap exports, but by the most advanced technologies like A.I., supercomputing and green energy. For Mr. Xi, this moment helps dent the aura of superiority the United States has held in A.I., a critical field in a fierce superpower rivalry. China has cast itself as a benevolent global partner to developing countries, willing to share its know-how, with Mr. Xi saying that A.I. should not be a "game of rich countries and the wealthy." Now, DeepSeek has shown that it might be possible for China to make A.I. cheaper and more accessible for everyone. The question, though, is how the ruling Communist Party manages the rise of a technology that could one day be so disruptive that it could threaten its interests -- and its grip on power. Chinese regulation of A.I. has varied in intensity over the years, depending on where the country assesses its strengths and weaknesses. When the Chinese government was worried it had fallen behind the United States in 2022 after the launch of OpenAI's ChatGPT, it took a more hands-off approach that ultimately allowed ventures like DeepSeek and others to thrive. Now that the pendulum has swung the other way, that confidence in the industry could prove to be a "double-edged sword," said Matt Sheehan, who studies Chinese A.I. as a fellow at the Carnegie Endowment for International Peace. The party's "core instincts are toward control," Mr. Sheehan said. "As they regain confidence in China's A.I. capabilities, they may have a hard time resisting the urge to take a more hands-on approach to these companies." As if to underscore that possibility, DeepSeek's founder, Liang Wenfeng, was invited to a discussion with Premier Li Qiang on Jan. 20, the same day that the company released its latest and most powerful A.I. model, known as R1. Mr. Liang's attendance was all the more remarkable considering DeepSeek had not been considered one of China's so-called A.I. Tigers. That distinction is reserved for high-profile firms like Zhipu AI, a Beijing-based start-up that has received substantial state investment. DeepSeek is no stranger to the party's urge to interfere; that may have inadvertently played a role in its eventual success. DeepSeek had originally trained its A.I. models to make bets on the Chinese stock market. But when regulators targeted such behavior, it pivoted in 2023 to advanced A.I. to conform with China's industrial policy. Then it stunned the world by rivaling the performance of its American competitors despite using far fewer of the advanced computer chips that are hard for China to obtain -- a technological feat that until recently had not been available. At home, Chinese commentators have held up DeepSeek's achievement as evidence that U.S. restrictions on exports of A.I. chips to China are ultimately futile (even though the company's founder has said such limits are a major concern). Even the recent allegations by OpenAI that DeepSeek improperly harvested its data to build its models have not deterred its fans in China, who accuse the San Francisco company of spreading rumors. "The U.S. technological sanctions on China have left China with no choice but to develop," said Sun Chenghao, a foreign relations expert at Tsinghua University in Beijing, echoing a popular sentiment in China. "We can only rely on ourselves." A.I. holds a special place in Mr. Xi's vision of China's rise, with its potential to help the country overcome many of its biggest challenges like its shrinking work force. China has used facial recognition and algorithms to supercharge its ability to surveil its people and snuff out dissent. The technology is also factoring into China's military modernization with autonomous weapons systems and even battlefield strategy. DeepSeek's development could also advance China's geopolitical goals. DeepSeek uses an open source model, meaning anyone can peer under its hood and use its technology, unlike leading American companies that use more expensive proprietary software. "The low cost and open source nature of DeepSeek's model bolsters the Chinese government's narrative that China is the place developing countries can look to for A.I. solutions," Mr. Sheehan said. How big a player China becomes on the global stage in A.I. could ultimately depend on how the government decides to balance regulations with the freedom that companies and researchers need to do cutting-edge work that allows them to compete with the United States. Some analysts like Gregory C. Allen, a researcher at the Center for Strategic and International Studies and a former U.S. defense official, said there were most likely no restraints on A.I. development when it comes to China's military. "The only thing holding them back is performance," said Mr. Allen, who in his former job held talks with members of the People's Liberation Army responsible for assessing the risks of A.I. The same does not hold true for regulating A.I. in the private sector. The landscape there is dictated by the competing priorities of China's regulatory agencies, each feeling their way around a technology that many in the world still do not fully understand. It is clear that the more widely used a technology is, the more the party will want to rein it in. In 2023, just months after ChatGPT set off an investment frenzy over artificial intelligence, China issued rules aimed at controlling what Chinese chatbots say to users, requiring them to reflect "socialist core values" and avoid information that undermines "state power." In the case of DeepSeek's chatbot, this has led to awkward responses to seemingly benign questions like, "Who is Xi Jinping?" Researchers testing its capabilities have found that the bot gives answers that spread Chinese propaganda and even parrot disinformation campaigns. Some concerns are more existential in nature. A growing chorus of scholars have been sounding the alarm about the potentially catastrophic consequences of losing human control over A.I. Chief among those voices has been Andrew Yao, a giant in A.I. at Tsinghua University and a recipient of the Turing Award, the equivalent of the Nobel Prize for computing. His influence helped establish what China calls the Global AI Governance Initiative, which was introduced by Mr. Xi in 2023 and included a call to always keep A.I. under human control. Last year, the government also called for the enhancement of A.I. governance "on the basis of human decision-making and supervision." Ultimately, A.I. in China may only advance as far as the government decides it can mitigate those risks, said Barath Harithas, an expert on A.I. policy at the Center for Strategic and International Studies, a Washington think tank. "Overregulation and the need to adhere to 'core socialist values' could risk neutering A.I.'s potential," Mr. Harithas said.

European AI allies unveils LLM alternative to Big Tech, DeepSeek

OpenEuroLLM is building open-source foundation models for Europe As China's DeepSeek threatens to dismantle Silicon Valley's AI monopoly, a European alliance has emerged with an alternative to tech's global order. They call their project OpenEuroLLM. Like DeepSeek, they aim to develop next-generation open-source language models -- but their agenda is very different. Their mission: forging European AI that will foster digital leaders and impactful public services across the continent. To support these objectives, OpenEuroLLM is building a family of high-performing, multilingual large language foundation models. The models will be available for commercial, industrial, and public services. Over 20 leading European research institutions, companies, and high-performance computing (HPC) centres have enlisted in the the project. Leading their alliance is Jan Hajič, a renowned computational linguist at Charles University, Czechia, and Peter Sarlin, the co-founder of Silo AI, Europe's largest private AI lab, which was acquired last year by US chipmaker AMD for $665mn. They're joined by an array of European tech luminaries. Among them are Aleph Alpha, the leading light of Germany's AI sector, Finland's CSC, which hosts one of the world's most powerful supercomputers., and France's Lights On, which recently became Europe's first publicly-traded GenAI company. Their alliance has been backed by the European Commission. According to Sarlin, the initiative could be the Commission's largest-ever AI project. "What's unique about this initiative is that we're bringing together many Europe's leading AI organisations in one focused effort, rather than having many small, fragmented projects," he told TNW via email. "This concentrated approach is what Europe needs to build open European AI models that eventually enable innovation at scale." The project has a budget of €52mn, as well as compute commitment that may have a larger monetary value, Sarlin said. Alongside funding from the Commission, OpenEuroLLM has received support from STEP, an EU scheme to boost investment in strategic technologies. The project also aligns with the EU's plans to fortify Europe's digital sovereignty, which is becoming vulnerable. With China and the US developing new AI capabilities at breakneck speeds, Europe faces an uncertain future in the digital landscape. OpenEuroLLM hopes to strengthen the continent's position with new digital infrastructure. The project has also pledged to embed AI with European values of democracy, transparency, openness, and community involvement. According to OpenEuroLLM, the models, software, data, and evaluation will be fully open. They will also be capable of fine-tuning and instruction-tuning for specific industry and public sector needs. Additionally, the alliance promises to preserve both linguistic and cultural diversity. The plans arrive in testing times for European tech. With US and Chinese firms racing to deliver new AI breakthroughs, fears are growing that European companies, economies, and even culture are under threat. Sarlin wants OpenEuroLLM to bring new hope to the continent. "This isn't about creating a general purpose chatbot -- it's about building the digital and AI infrastructure that enables European companies to innovate with AI," he said. "Whether it's a healthcare company developing specialised assistants to medical doctors or a bank creating personalised financial services, they need AI models adapted to the context in which they operate, and that they can control and own. "This project is about giving European businesses tools to build models and solutions in their languages that they own and control."

Meet the founder of DeepSeek, the China AI startup rattling Nvidia and Big Tech

This story incorporates reporting from devdiscourse, FIRSTonline, Forbes and BBC. Liang Wenfeng's DeepSeek is bringing Chinese innovation to the fore in the artificial intelligence landscape. Launched in December 2023, the company has already challenged industry giants like OpenAI. DeepSeek is spearheading a shift in AI discourse, where Chinese technology is not merely an echo of Western advancements but a distinct and independent force. The vision for DeepSeek is clear: not only to create an AI product that competes globally but also to catalyze a new paradigm of more affordable AI. Liang's journey to founding DeepSeek began in 2015 with the establishment of High-Flyer Quantitative Investment Management. There he garnered valuable experience developing computer models for stock trading. The financial markets' inherent volatility offered a fertile ground for refining AI models -- a process that laid the groundwork for DeepSeek. By 2023, Liang was ready to translate his finance-based AI acumen into a new venture, marking the inception of DeepSeek. Within months, the startup emerged as a formidable player, owing to its leveraging of computational resources that rival those of established market leaders. The company's rapid ascent has further spotlighted China's technological self-sufficiency. Traditionally, China's tech sector has been perceived as trailing behind the U.S. and other Western powers. Now, however, DeepSeek exemplifies a shift towards autonomous innovation. The firm's focus on cost-effective AI solutions also means broader access to advanced technology -- a democratization of AI capabilities that could significantly alter industry dynamics globally. Unlike older models that require expansive infrastructure, DeepSeek's approach fosters greater technological inclusivity. This strategic push aligns with China's broader economic transformation from traditional manufacturing into advanced tech sectors like AI, chips, and electric vehicles. DeepSeek's impact has been palpable; the app's rise to the top of the Apple Store's download ranks stands as a testament to its global relevance and appeal. Investors have taken notice, and some major tech stocks have reacted accordingly. The international AI race, once squarely in America's domain, now feels the competition heating up from China's burgeoning tech sphere. Liang's academic credentials -- having graduated in artificial intelligence from Zhejiang University with expertise in computer vision and big data -- have significantly contributed to his visionary leadership at DeepSeek. His work is celebrated as a source of national pride, with experts like Marina Zhang from the University of Technology Sydney recognizing DeepSeek's advances as emblematic of China's growing technological prowess. These achievements, however, are shaded by considerations of regulatory compliance, especially concerning politically sensitive content -- a common requirement for Chinese tech companies. Like other AI systems from prominent Chinese firms, DeepSeek is trained to navigate these challenges carefully. This nuanced balancing act between innovation and regulation reflects the complexities Liang must manage as he steers DeepSeek towards further breakthroughs. As DeepSeek continues to evolve, the global AI industry must reckon with its potential -- particularly if Liang's ambitions for a low-cost, widespread AI model come to fruition. With each advancement, DeepSeek not only redefines China's role in the technological arena but also challenges global perceptions of what Chinese innovation can achieve.

The U.S. restricted China's access to AI chips. That didn't stop DeepSeek

This story incorporates reporting from China Briefing, The Washington Post and The Associated Press on MSN.com. China's AI sector is advancing rapidly, led by innovation and resource-efficient practices. Pioneering companies like DeepSeek have significantly contributed to these advancements with open-source models that challenge conventional AI development methods. These models have reduced reliance on traditional chip-intensive processes, allowing China to maintain its competitive edge despite U.S. chip restrictions. This progress marks a shift in China's AI industry toward sustainability and collaboration, driving it to the forefront of international discussions on AI governance and policy. DeepSeek, a notable player in China's AI landscape, exemplifies this shift with its cutting-edge R1 AI model. The company's approach has captivated the tech community by cutting costs and minimizing dependency on high-performance chips. By leveraging open-source strategies, DeepSeek and other companies offer free access to their models worldwide, fostering a collaborative ecosystem that supports innovation beyond national borders. One of DeepSeek's significant contributions is its R1-Zero model, which challenges the conventional belief in the necessity of supervised fine-tuning during AI model development. By eliminating this stage, DeepSeek has streamlined the AI training process -- showing that large models can be trained efficiently without extensive computational resources. The growing trend of open-source AI models in China, from startups like Minimax to tech giants such as Alibaba, signals a strategic movement toward shared development and cost efficiency. This trend allows Chinese companies to make impressive strides without the high-performance chips that are restricted by U.S. sanctions. By emphasizing collaboration and open access, China not only bypasses some effects of these restrictions but also actively participates in reshaping global AI research dynamics. China's advancements in AI, driven by resource efficiency and open-source contributions, influence the competitive dynamics between nations. These developments are spurring discussions around the need for a revised global AI policy, as China's approach disrupts traditional models of AI creation and governance. As these conversations continue, China's innovations stand as a testament to the potential of alternative strategies in overcoming geopolitical and technological barriers. Looking ahead, the focus will likely remain on how well China can sustain this momentum and how their approach might influence international policy frameworks. As of January 2025, the AI landscape continues to evolve rapidly, with China prominently positioned to shape its trajectory.

DeepSeek data breach: A grim warning for AI security

The Chinese AI startup DeepSeek, known for its advanced AI chatbot DeepSeek R1, has found itself at the centre of a major data breach controversy. Security researchers uncovered a critical vulnerability in its database infrastructure, exposing sensitive user data and operational secrets. This incident has sparked widespread concerns about the security practices of AI companies, especially as DeepSeek eyes global expansion. The breach was discovered by Wiz Research, a cybersecurity firm based in New York. Within minutes of scanning DeepSeek's systems, researchers identified a publicly accessible ClickHouse database hosted on the company's domains. Also read: DeepSeek AI: How this free LLM is shaking up AI industry The database lacked authentication, leaving it open to anyone online. This misconfiguration allowed unrestricted access to over one million log entries containing sensitive information, including chat histories, API keys, backend operational details, and metadata. What made the situation even more alarming was that the database had full administrative privileges. This meant that not only could an attacker view the data, but they could also modify or delete it. Researchers noted that depending on the database's configuration, attackers could potentially retrieve plaintext passwords and proprietary files using simple SQL commands. Apart from the data breach, DeepSeek R1 has been criticised for its susceptibility to cyberattacks. Security researchers have demonstrated how the model can be exploited using techniques like "Evil Jailbreak," which bypasses safety mechanisms to generate harmful content. These vulnerabilities further compound concerns about the company's ability to safeguard its systems and users. Also read: DeepSeek vs Meta: 5 Things Mark Zuckerberg Teased About Llama 4 and the Future of Open-Source AI Once alerted by Wiz Research, DeepSeek acted swiftly to secure the exposed database. However, this quick response does little to mitigate the broader implications of such a lapse. Also read: Qwen 2.5 Max better than DeepSeek, beats ChatGPT in coding, costs 10x less than Claude 3.5 Security experts have criticised the company for failing to implement basic security measures, such as authentication protocols and encryption. Gal Nagli, a cloud security researcher at Wiz, highlighted that while much of the focus in AI security is on futuristic threats like adversarial attacks, basic oversights such as exposed databases pose far greater risks. DeepSeek's data breach is not an isolated case but rather a symptom of a larger issue within the rapidly growing AI industry. As companies rush to deploy generative AI models and expand their user base, many overlook essential security protocols. This negligence not only jeopardises user trust but also exposes businesses to regulatory scrutiny and potential legal consequences. The breach has also raised questions about DeepSeek's readiness for global expansion. Also read: DeepSeek vs OpenAI: Why ChatGPT maker says DeepSeek stole its tech to build rival AI The company recently announced plans to host its services on local servers in India, aligning with the country's data localisation policies. However, this incident casts doubt on whether DeepSeek can meet India's stringent data protection standards. The breach has attracted attention from regulators worldwide. Authorities in Italy and Ireland have launched investigations into DeepSeek's data handling practices, while the U.S. Navy has warned personnel against using its services due to security concerns. These developments highlight the growing scrutiny faced by Chinese tech companies operating in international markets. On forums like Reddit, users have expressed outrage over DeepSeek's negligence. Many have compared this incident to hypothetical scenarios involving U.S.-based companies like Google or OpenAI, emphasising that such lapses would provoke even greater backlash if they occurred in Western firms.

How China's DeepSeek Came for Big AI

As US companies pour billions of dollars into advancing artificial intelligence, a little-known Chinese startup has seemingly done the impossible. DeepSeek unveiled a chatbot app that performs as well if not better than those of Silicon Valley giants, and at a fraction of the cost. (Source: Bloomberg)

Opinion | DeepSeek has created a 21st-century Sputnik moment

Export controls may have helped turbocharge Chinese innovation. Is this a Sputnik moment? The world has reacted with astonishment to the release of a disruptive AI model from Chinese company DeepSeek, which appears to be able to perform as well or, in some cases, better than ChatGPT and other cutting-edge models put out by U.S. companies. Americans had assumed their massive lead in funding, access to high-quality chips, and innovation would keep them well ahead. That assumption now looks like hubris. The episode is in some ways a much bigger deal than Sputnik. Sputnik was about the Soviet Union's space program competing with that of the United States. Few thought the Soviet economy in general was more technologically advanced than America's. But DeepSeek is a private Chinese company that demonstrated its stunning prowess on the cheap in the most important technology for the future. It's not exactly clear just how much DeepSeek's model actually cost, to what extent it needed to use U.S. models for training, and whether there was any closet Chinese government help. But given the enormous efforts that the U.S. government has made over the last few years to preserve its advantage -- chip bans, export controls, etc. -- DeepSeek has made a remarkable achievement. It suggests to me two lessons and two questions. The first lesson is that, over time, open artificial-intelligence systems are likely to outperform closed systems. (An open system is like Lego blocks with instructions; a closed one is the built Lego structure with the instructions kept secret.) Many have pointed out that DeepSeek used Meta's open-source Llama model to train. It also used Qwen, a family of AI models, also open-source, put out by the Chinese technology giant Alibaba. While DeepSeek is currently the best, China's big technology companies have been releasing a number of AI models, mostly open-source, that are getting better and better. If the history of technology is any guide, the ability to see the innards of these models and understand their reasoning should lead to greater and faster technological innovation than using closed models that others cannot use for collaboration. Second, constraints can be useful, as former Intel CEO Pat Gelsinger has noted. Just as art sometimes flourishes in repressive environments, in which restrictions force artists to be creative, so also engineers often operate best under constraints. Forced to use second-tier chips, Chinese engineers produced creative workarounds. This is not just true with DeepSeek. In 2023, Chinese telecommunications giant Huawei released a smartphone with a 7-nanometer chip, a kind that had been explicitly banned by U.S. export controls. There is some evidence that, after years of sclerosis, China's chipmakers have responded to U.S. bans by becoming much more innovative. Follow Fareed Zakaria Follow In a fascinating interview last year, Liang Wenfeng, the CEO of DeepSeek, argues that his engineers are more motivated by doing research than making money, and appears to contrast that attitude with the one prevalent in Silicon Valley, which is all about maximizing revenue, providing cloud services and generating cash flow. Demis Hassabis, who leads Google's DeepMind and also shared the 2024 Nobel Prize in chemistry for AI-related scientific breakthroughs, is said to have fought to keep his team in London, far from Silicon Valley, so that it can focus on basic research. The first question that DeepSeek raises is: Can the United States stop China from advancing along the technological frontier? Some argue that DeepSeek shows that export controls work: Its model needed many Nvidia chips, which it managed to procure before export bans were fully in place. Soon, China will not have access to the best chips and will suffer even more from the ban. But as we have learned with the rounds and rounds of global sanctions against Russia, the world economy is large and porous. Stuff gets through. And China is not Russia. It is a vast, technologically sophisticated economy with millions of software developers and hundreds of high-quality firms in the technology space. Human talent on that scale will find ways to innovate, even if those measures keep China slightly behind. The second question: What is the cost of this approach? If technology bans and export controls at best keep China behind a year -- maybe just several months -- is that gain worth the cost? That cost is Chinese retaliation, limiting the United States' access to key materials that it needs for high technology. More important, a decoupled global economy also creates a closed ecosystem in which U.S. technology companies will not face competition from the best. Is Tesla going to innovate at the highest level if it is not facing its strongest Chinese rival? A technology decoupling means that AI will become the central part of a new global arms race, totally unregulated and unconstrained, with the world's two largest economies hurtling toward superintelligence no-holds-barred, and incorporating it into all military applications -- including nuclear weapons. If artificial intelligence is as revolutionary a technology as predicted, having it unleashed in every realm of human life with absolutely no guardrails points to a scary future -- one far more dangerous than anything people imagined because of the Sputnik satellite.

High flyer to AI flyer: Quant whiz turns tech disruptor

The 40-year-old founder of China's DeepSeek, an AI startup that has startled markets with its capacity to compete with industry leaders like OpenAI, kept a low profile as he built up a hedge fund and then refined its quantitative models to branch into artificial intelligence. Liang Wenfeng, who founded DeepSeek in 2023, was born in southern China's Guangdong and studied in eastern China's Zhejiang province, home to e-commerce giant Alibaba and other tech firms, according to Chinese media reports. The hedge fund he set up in 2015, High-Flyer Quantitative Investment Management, developed models for computerized stock trading and began using machine-learning techniques to refine those strategies. Like many Chinese quantitative traders, High-Flyer was hit by losses when regulators cracked down on such trading in the past year. However, it reportedly manages $8 billion in assets, ample resources for funding DeepSeek's AI research. It also has abundant computing power for AI, since High-Flyer had by 2022 amassed a cluster of 10,000 of California-based Nvidia's high-performance A100 graphics processor chips that are used to build and run AI systems, according to a post that summer on Chinese social media platform WeChat. The U.S. soon after restricted sales of those chips to China. "Thing is, we are sure now that we want to do this, can do this, and are capable of doing this, so we're among the best-suited candidates to tackle it at this moment," Liang told Waves, a tech media outlet, in 2023. "Currently, neither tech giants nor startups have an unassailable lead. With OpenAI paving the way, everyone is working with published papers and open-source code," it quoted him as saying. Liang said he spends his days reading papers, writing code, and participating in group discussions, like other researchers. DeepSeek is exploring what intelligence means, he said. "People may think there's some hidden business logic behind this, but it's mainly driven by curiosity," Liang said. When DeepSeek was asked, "Who is Liang Wenfeng?" its first answer was to name a different Chinese entrepreneur with the same name, at least as spelled in English letters. When asked: "Where is Liang Wenfeng from and where did he go to university?" it said that as of October 2023, the most recent knowledge cutoff for DeepSeek's R1 AI model, "there is no publicly available information about Liang Wenfeng's background, including his place of origin or educational history." "If you are referring to the founder of DeepSeek, details about his personal life or academic background have not been disclosed publicly. For more information about DeepSeek, you can visit its official website," it said. Liang's focused approach fits in with his determination to push AI learning forward. After decades of relying on innovation from the West, he says China should be making its own contributions. "What we see is that Chinese AI can't be in the position of following forever. We often say that there is a gap of one or two years between Chinese AI and the United States, but the real gap is the difference between originality and imitation," he said in another Waves interview in November. "If this doesn't change, China will always be only a follower - so some exploration is inescapable."

From copycat to innovator: How China trumped US with DeepSeek

China, previously known for producing low-quality knock-offs, has emerged as an innovator in various high-tech fields. Recent breakthroughs in iron-making, stealth fighter jets, satellite communication, and artificial intelligence highlight China's progress, posing a challenge to Western technology and innovation dominance.Last month, China unveiled a new iron-making technology that can lead to a 3,600-fold or more increase in the speed of iron-making. It can improve the energy-use efficiency of China's steel industry by more than one-third. A few weeks later, it stunned the world with a new stealth fighter jet, believed to be a 6th-generation model, a big achievement for a country considered a laggard in aviation. Nearly a week later, China achieved a significant milestone in satellite-to-ground laser communication while reaching a data transmission rate of 100 gigabits per second (Gbps), getting ahead of Elon Musk's Starlink in the race for advanced satellite communication technologies. Last week, a Chinese artificial intelligence startup's chatbot, DeepSeek, surged to become the most downloaded free app on Apple's US App Store, displacing OpenAI's ChatGPT. The rise of DeepSeek hammered AI-linked stocks and wiped out close to a trillion dollars in market capitalisation of companies such as Nvidia. What truly rattled the industry was DeepSeek's claim that it developed its latest model, the R1, at a fraction of the cost that major companies are investing in AI development, primarily on expensive Nvidia chips and software. The news sent shockwaves through the US tech sector, exposing a critical concern: should tech giants continue to pour hundreds of billions of dollars into AI investment when a Chinese company can apparently produce a comparable model so economically? Elon Musk, who has invested heavily in Nvidia chips for his company xAI, suspects DeepSeek of secretly accessing banned H100 chips -- an accusation also made by the CEO of ScaleAI, a prominent Silicon Valley startup backed by Amazon and Meta. DeepSeek, which has unsettled Silicon Valley as well as the Donald Trump government, is a cheap copy of an American invention, something China has come to excel in. But DeepSeek is not another notorious Chinese knock-off. It poses a challenge to Silicon Valley with its efficiency. From fake Mercedes-Benz built by Geely Motors to DeepSeek, China has evolved from a copycat to an innovator, worrying the West with its advances in technology in different fields --- from iron-making to AI. How China evolved from a copycat to an innovator Not long back, China was synonymous with low-quality copycats. You could find in China knock-offs of most Western products, from Nokia mobile phone to Mercedes-Benz. As it got more deeply integrated into Western value chains, becoming the factory of the world, it also started pilfering Western technology at an industrial scale. With state support which included deep spying in the West as well as large-scale funding at home, China was able put behind its past image of a producer of low-quality, cheap and counterfeit goods. But there was more at work besides stealing and copying western technology and technology transfers from Western companies producing in China to their domestic partners. The very imitation in which China excelled laid the ground for innovation more than a decade ago. "Companies like Alibaba, Tencent and Xiaomi began by emulating Western models," wrote Vivek Wadhwa the CEO, Vionix Biosciences, ET last year. "But they soon adapted these ideas to suit their local markets, innovating in ways that their Western counterparts hadn't. Take Tencent's WeChat. Originally inspired by WhatsApp and other messaging platforms, WeChat quickly evolved into something far more comprehensive and transformative. It became a 'super app', offering services that span from messaging and social networking to payments, ecommerce and even government services. Today, WeChat is not just a Chinese success story but a model of innovation that other companies worldwide like Meta, X, Tata Group and Grab are looking to replicate." "This strategy is particularly effective in regions where markets and consumer behaviours differ significantly from those in the West. By copying a proven concept and then localising and expanding upon it, companies can create products that are more finely tuned to their specific markets. This process often leads to innovations that go beyond the original idea, offering features and services that the 'originators' never considered," Wadhwa said. A huge number of students opting to study STEM subjects, building a formidable local talent base, also helped the Chinese state that was trying to foster innovation with financial, strategic, corporate and regulatory efforts. China's innovation project was initially marked by poor research and junk patents, with quantity overtaking quality. However, China followed the dictum 'you fake it till you make it' with great zeal which can be gauged from the fact that last year for the first time, the number of international patents filed from inventors in China surpassed applications from the US, as per a report by 'Science and Engineering Indicators'. Another report by the World Intellectual Property Organization showed last year that China-based inventors are filing the highest number of generative artificial intelligence (GenAI) patents, far outpacing inventors in the US, Republic of Korea, Japan and India that comprise the rest of the top five locations. New data from Georgetown University's Center for Security and Emerging Technology (CSET) shared first with Axios showed China leading the US as a top producer of research in more than half of AI's hottest fields. The GenAI patents lead of China over the US has also changed the narrative that Chinese patents are more quantity than quality. When CSET researchers narrowed their analysis to highly cited papers, the Chinese Academy of Sciences was still the leader. Google is second, followed by China's Tsinghua University, Stanford and then MIT. Another research showed last year that China had by some metrics eclipsed the US as the biggest producer of AI talent, with the country generating almost half the world's top AI researchers. By contrast, about 18 percent come from US undergraduate institutions, according to the study, from MacroPolo, a think tank run by the Paulson Institute, which promotes constructive ties between the US and China. The research, as per an NYT report, also shows another new trend. Earlier, the US benefited as large numbers of China's top minds moved to American universities for research and most of them stayed back. But now a growing numbers of Chinese researchers are choosing to staying in China. China faces 'middle-technology trap' While there could be questions over China's latest technological advances, China is also worried about falling in a 'middle-technology trap'. Nearly a year ago, a top state-run Chinese think-tank warned China of a potential "middle-technology trap" unless it throws doors "wide open" for new investments and scientific innovation to deal with increasing containment in the ongoing tech war with the US. "China's manufacturing sector is still in the downstream of the global value chain, and it faces a risk of being hamstrung at the low and mid-end by developed countries such as the United States, Germany and Japan," the Chinese Academy of Sciences said. "The countries that develop later usually have difficulties in industrial upgrading and transitioning to high-income countries because they lack original technological advances after technology importation, imitation, absorption, and tracking," the report said. The report came amid stepped-up technology curbs by the US, while Chinese manufacturers were finding it increasingly difficult to move up value chains. The "middle-technology trap", as per an SCMP report, describes a scenario in which developing countries benefit from industrial transfers due to their low-cost advantages, but face long-term economic stagnation when the advantages diminish, and local firms struggle to catch up with the core technologies retained by developed nations. China is seen to be innovating despite tight Western curbs on tech exports, and many argue because of them as it is forced to make its own technology. But it remains to be seen if it can become global high-tech power on its own. (With inputs from agencies)

DeepSeek's AI model tests limits of U.S. restrictions on Nvidia chips

Powerful artificial intelligence software from Chinese startup DeepSeek indicates that its engineers built a competitive model despite U.S. attempts to curtail China's tech development, raising questions about the effectiveness of Washington's trade curbs. The company's recently released R1 model, which it claims to have developed at a fraction of the cost borne by rival AI companies, sent tech stocks into a tailspin Monday as investors questioned the need to spend billions on advanced hardware. It's also sparked a debate in Washington about the best strategy to prevent China from developing cutting-edge AI, which U.S. policymakers see as a national security risk. The U.S. imposed sweeping controls on the sale of the most advanced Nvidia chips to China in October 2022, and has ratcheted up the measures each year since. But Nvidia has responded by designing new semiconductors for the Chinese market -- including those DeepSeek likely used to build R1.