DeepSeek: China's AI Breakthrough Challenges U.S. Dominance

'Game on': Tech execs say DeepSeek ramps up China-U.S. competition but won't hurt OpenAI

Tech bosses largely agree the risk DeepSeek poses to OpenAI remains limited for now. The technological advances that Chinese artificial intelligence lab DeepSeek have displayed show the game is on when it comes to U.S.-Sino competition on AI, top tech executives told CNBC. In a series of interviews at France's Artificial Intelligence Action Summit, leaders of several major tech companies told CNBC that the emergence of DeepSeek demonstrates that China can't be counted out as a serious player when it comes to AI innovation. Last month, DeepSeek shocked global markets with a technical paper saying that one of its new AI models was created with a total training cost of less than $6 million -- far less than the billions upon billions of dollars being spent by Big Tech players and Western AI labs such as OpenAI and Anthropic. Chris Lehane, chief global affairs officer at OpenAI, told CNBC that DeepSeek's advanced, low-cost model confirms there is a "very real competition between U.S.-led, small D democratic AI and CCP [Chinese Communist Party] China-led autocratic, authoritarian AI." Many critics of DeepSeek have pointed to apparent censorship by the model when it comes to sensitive topics. For example, when asked about the 1989 Tiananmen Square massacre, DeepSeek's AI assistant app responds with: "Sorry, that's beyond my current scope. Let's talk about something else."

DeepSeek founder Liang Wengfeng spotted in public for first time since company upended AI in the West

Please use Chrome browser for a more accessible video player The elusive founder of disruptive Chinese AI company DeepSeek has been spotted attending a Beijing conference, one week after snubbing a global AI summit in Paris. Liang Wenfeng was seen publicly for the first time since the company's AI model upended the tech industry. Up until DeepSeek's unveiling, the US was thought to hold the global monopoly on artificial intelligence, with every top 10 AI company in the world being based in the States. But when DeepSeek revealed a powerful model that was significantly cheaper to run than OpenAI's ChatGPT, it threw that monopoly into question. Read more: Inside China's Silicon Valley, where founder of DeepSeek Liang keeps a low profile Musk v Altman: The battle to become king of AI The share prices of leading AI chipmaker Nvidia plummeted, as did those of Meta Platforms, Microsoft, and Alphabet. Despite causing chaos in the West, however, Mr Wengfeng didn't attend a global summit on AI in Paris last week. OpenAI chief executive Sam Altman was in the French capital for the conference and told Sky News he would like to "work with China" - although he doesn't know if the US government would let him do that. "Should we try as hard as we absolutely can [to work with them]? Yes," he said. Mr Wengfeng was seen at a rare meeting held by President Xi Jinping on Monday with some of the biggest names in China's technology sector. During the meeting, the president urged Chinese tech leaders to "show their talent" and be confident in the power of China's model and market. One analyst said the meeting was a reflection of the Chinese government's concern that it is lagging behind the US when it comes to technology. "It's a tacit acknowledgement that the Chinese government needs private-sector firms for its tech rivalry with the United States," said Christopher Beddor, deputy China research director at Gavekal Dragonomics in Hong Kong. "The government has no choice but to support them if it wants to compete with the United States."

China's DeepSeek has taken the world by storm. Here are the brains powering the AI sensation

DeepSeek offices in Beijing on Jan. 28, 2025.Peter Catterall | Afp | Getty Images Artificial intelligence startup DeepSeek has rocketed into global prominence, shaking up the AI world, but the team behind it is relatively unknown outside China. DeepSeek's founder, Liang Wenfeng, has been dubbed by some in Western media as China's Sam Altman. But unlike his Silicon Valley counterpart, Liang has maintained a low public profile. Liang's team, comprising young graduates from some of the country's leading universities, is also little known. The team consists of fewer than 140 people, according to Chinese state media, though a research paper on its latest R1 reasoning model lists about 200 contributors. CNBC has been unable to confirm the official size of the team. Outside of its core technology developers, DeepSeek has mostly shared the senior management team, operation staff, human resources and financial accountants of its mothership High-Flyer, according to sources familiar with the company. Here's an overview of the people behind the AI sensation and how the startup came into being.

The story behind DeepSeek's breakthrough

Your browser does not support playing this file but you can still download the MP3 file to play locally. This January, DeepSeek made an announcement that changed the artificial intelligence landscape. The Chinese start-up said that it had trained the latest in a series of competitive AI models, which appeared to be built much more cheaply than those of Silicon Valley competitors. The FT's China Technology Correspondent, Eleanor Olcott, first wrote a story about the company last summer. She's on today to discuss how DeepSeek's founder Liang Wenfeng was able to finance and grow his company at a time when start-up funding is scarce in China, and often comes with too many strings attached. Liang Wenfeng, the DeepSeek founder panicking the tech world The global AI race: is China catching up to the US?

Chinese businesses rush to try DeepSeek AI at 'unprecedented' scale

DeepSeek, founded in 2023 out of a quantitative hedge fund, had released a basic version of R1 in November, and a V3 model in December. It launched a smartphone chatbot app in January. Another attractive factor for businesses is that DeepSeek's models are open-source, allowing individuals and companies to download and customize it. DeepSeek also advertised drastically lower prices for applications to use its tech versus that of OpenAI. ChatGPT is not officially available in mainland China and requires users to provide an overseas phone number and payment method from a supported country such as the U.S. DeepSeek changed the perception that AI models only belong to big companies and have high implementation costs, said James Tong, CEO of Movitech, an enterprise software company which says its clients include Danone and China's State Grid. He said Movitech started integrating an earlier version of DeepSeek in the fourth quarter of last year, helping boost sales by about 25% from the same period in 2023. The company plans to launch a new DeepSeek-integrated application by the end of March to improve clients' ability to make decisions, he said. Many recent videos on Chinese social media have showed off how to run a local version of DeepSeek on Apple's Mac mini. Apple Mac mini online sales in China climbed significantly from November to January, versus the same period the year prior, according to data from consultancy WPIC. The electronics-focused JD.com site recorded unit sales of around 20,200 in January, up from nearly 19,400 in December and around 12,250 in November, the data showed. DeepSeek's affordability is pressuring more expensive AI models to cut prices, enabling more businesses to adopt the tech, said Chim Lee, senior Asia analyst at the Economist Intelligence Unit. He added that open-source models allow finance, banking and healthcare businesses -- which are subject to stringent data protection rules in China -- to develop AI applications locally. "It is still very early to point to concrete business applications, but a key takeaway is that DeepSeek will accelerate the commoditization of AI," Lee said. Beijing is also increasing support. China's national supercomputing network announced Tuesday that eligible companies and individuals can obtain three free months of DeepSeek access, along with subsidized computing power. The network is similar to OpenAI's Trump-backed Stargate project in the U.S. for building AI infrastructure -- with the potential for "even faster scaling," Winston Ma, adjunct professor at NYU School of Law said Wednesday. He is also the author of "The Digital War: How China's Tech Power Shapes the Future of AI, Blockchain and Cyberspace."

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 (and Google Play, as well). 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 data center 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 "over-hyped." 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. Microsoft announced that DeepSeek is available on its Azure AI Foundry service, Microsoft's platform that brings together AI services for enterprises under a single banner. When asked about DeepSeek's impact on Meta's AI spending during its first-quarter earnings call, CEO Mark Zuckerberg said spending on AI infrastructure will continue to be a "strategic advantage" for Meta. At the same time, some companies are banning DeepSeek, and so are entire countries and governments. New York state also banned DeepSeek from being used on government devices. 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. TechCrunch has an AI-focused newsletter! Sign up here to get it in your inbox every Wednesday.

DeepSeek AI tastes big success in China. What you need to know?

The move by the Chinese tech giant is notable as integrating DeepSeek brings in an external AI platform, while tech firms compete fiercely in developing the most advanced AITencent said on Sunday its Weixin messaging app, China's largest, is allowing some users to search via DeepSeek's artificial intelligence model, as firms race to link up with the AI startup that has grabbed global attention, as per a report. In a beta test, Weixin is testing access to DeepSeek for searches, Tencent said in an email statement to Reuters. The move by the Chinese tech giant is notable as integrating DeepSeek brings in an external AI platform, while tech firms compete fiercely in developing the most advanced AI, Reuters reported. Weixin uses Tencent's proprietary Hunyuan-Large language model to enrich its AI search, a spokesperson said. Tencent is exploring the integration of multiple products with DeepSeek, including Tencent Cloud AI Code Assistant and Tencent Yuanbao, another AI assistant app, said a person with knowledge of the matter. DeepSeek is being tested by Weixin, which serves domestic users, not its sister app WeChat, which targets overseas users. The two had a combined 1.38 billion users at the end of September. DeepSeek last month upended the AI world, launching a free AI assistant that it says uses less data at a fraction of the cost of incumbent services. It quickly overtook U.S. rival ChatGPT in downloads from Apple's App Store. Among Chinese companies seeking to capitalise on DeepSeek's potential breakthrough, automaker Great Wall Motor and leading telecoms providers are integrating the AI model released by DeepSeek into their offerings. Q1. Which App is using DeepSeek? A1. Tencent said on Sunday its Weixin messaging app is allowing some users to search via DeepSeek's artificial intelligence model. Q2. How did DeepSeek hog all headlines? A2. DeepSeek last month upended the AI world, launching a free AI assistant that it says uses less data at a fraction of the cost of incumbent services. It quickly overtook U.S. rival ChatGPT in downloads from Apple's App Store.

DeepSeek gives China's chipmakers leg up in race for cheaper AI

Industry executives are now predicting that DeepSeek's open-source nature and its low fees could boost adoption of AI and the development of real-life applications for the technology, helping Chinese firms overcome US export curbs on their most powerful chips.The rise of DeepSeek's artificial intelligence (AI) models is seen providing some Chinese chipmakers such as Huawei a better chance to compete in the domestic market against more powerful U.S. processors. Huawei and its Chinese peers have for years struggled to match Nvidia in building top-end chips that could compete with the U.S. firm's products for training models, a process where data is fed to algorithms to help them learn to make accurate decisions. However, DeepSeek's models, which focus on "inference," or when an AI model produces conclusions, optimise computational efficiency rather than relying solely on raw processing power. That is one reason why the model is expected to partly close the gap between what Chinese-made AI processors and their more powerful US counterparts can do, analysts say. Huawei, and other Chinese AI chipmakers such as Hygon, Tencent-backed EnFlame, Tsingmicro and Moore Threads have in recent weeks issued statements claiming products will support DeepSeek models, although few details have been released. Huawei declined to comment. Moore Threads, Hygon EnFlame and Tsingmicro did not respond to Reuters queries seeking further comment. Industry executives are now predicting that DeepSeek's open-source nature and its low fees could boost adoption of AI and the development of real-life applications for the technology, helping Chinese firms overcome US export curbs on their most powerful chips. Even before DeepSeek made headlines this year, products such as Huawei's Ascend 910B were seen by customers such as ByteDance as better suited for less computationally intensive "inference" tasks, the stage after training that involves trained AI models making predictions or performing tasks, such as through chatbots. In China, dozens of companies from automakers to telecoms providers have announced plans to integrate DeepSeek's models with their products and operations. "This development is very much aligned with the capability of Chinese AI chipset vendors," said Lian Jye Su, a chief analyst at tech research firm Omdia. "Chinese AI chipsets struggle to compete with Nvidia's GPU (graphics processing unit) in AI training, but AI inference workloads are much more forgiving and require a lot more local and industry-specific understanding," he said. Nvidia still dominates However, Bernstein analyst Lin Qingyuan said while Chinese AI chips were cost-competitive for inferencing, this was limited to the Chinese market as Nvidia chips were still better even for inference tasks. While U.S. export restrictions ban Nvidia's most advanced AI training chips from entering China, the company is still allowed to sell less powerful training chips that Chinese customers can use for inference tasks. Nvidia published a blog post on Thursday about how inference time was rising as a new scaling law and argued that its chips will be necessary to make DeepSeek and other "reasoning" models more useful. In addition to computing power, Nvidia's CUDA, a parallel computing platform that allows software developers to use Nvidia GPUs for general-purpose computing, not just AI or graphics, has become a crucial component of its dominance. Previously, many Chinese AI chip companies did not directly challenge Nvidia by asking users to abandon CUDA but instead, claimed their chips were compatible with CUDA. Huawei has been the most aggressive in its efforts to break away from Nvidia by offering a CUDA equivalent called Compute Architecture for Neural Networks (CANN), but experts said it faced obstacles in persuading developers to abandon CUDA. "Software performance of Chinese AI chip firms is also lacking at this stage. CUDA has a rich library and a diverse range of software capability, which requires significant long-term investment," said Omdia's Su.

Is DeepSeek Just a Well-Timed PR Storm?

China's latest A.I. entrant has shaken Silicon Valley and sparked global regulatory backlash -- but does it actually threaten U.S. dominance? China's DeepSeek A.I. has ignited debate across the tech world. Some see it as a watershed moment for China's A.I. ambitions, a challenge to U.S. dominance in artificial intelligence. Others dismiss it as more noise than substance, arguing that it offers nothing revolutionary. To make sense of the buzz, let's unpack the facts behind DeepSeek's sudden rise. 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 The Reality of DeepSeek's A.I.: Innovation or Just Open-Source Hype? For years, U.S. A.I. giants -- OpenAI, Google DeepMind and Meta (META) -- have led the charge in developing "reasoning models," A.I. systems designed to mimic human cognitive functions. In December 2024, OpenAI unveiled GPT-4o1, a closed-source model built for elite commercial applications. This month, OpenAI released Deep Research, an agentic A.I. built on the latest GPT o3 model. By contrast, DeepSeek R1 enters the market as an open-source alternative, triggering speculation about whether it can derail the funding and commercialization roadmaps of U.S. firms like Meta and Anthropic. However, at its core, DeepSeek is a mid-sized model -- not a breakthrough. Its primary distinction is its open-source framework, joining a category that includes LLaMA and its derivatives. But unlike its Western counterparts, DeepSeek does not introduce novel architecture or A.I. advancements. The Nvidia Factor: How Did DeepSeek Build Its Model? Reports suggest that DeepSeek's founders stockpiled Nvidia chips, which have been restricted from export to China since September 2022. Some speculate that by combining advanced GPUs with lower-tier chips, they've found a workaround to U.S. sanctions -- potentially making A.I. training cheaper and more efficient. One widely cited advantage of DeepSeek is its lower memory consumption, which theoretically reduces costs for users. If true, this could also address concerns about A.I.'s carbon footprint, a growing issue in global tech regulation. Environmental Costs: A.I.'s Energy Dilemma Data centers powering A.I. models consume massive amounts of electricity and water, primarily for cooling high-performance servers. While many A.I. firms avoid disclosing their carbon footprint, OpenAI has faced substantial scrutiny. A study by KnownHost estimates that ChatGPT emits around 260 tons of CO2 per month. While DeepSeek claims efficiency, it remains unclear whether it genuinely reduces computational waste or merely redistributes the cost. The Overnight Popularity of DeepSeek: Substance or Sensation? DeepSeek's rapid rise is primarily due to two key factors: Cost: Training an open-source model spreads expenses across multiple participants, reducing the overall financial burden. Hardware Flexibility: If DeepSeek can train models using standard chips, it challenges the idea that A.I.'s success depends on cutting-edge processors. However, despite these advantages, DeepSeek R1 (671B) remains costly to run, just like its counterpart LLaMA 3 (671B). This raises questions about its long-term viability for individual or small-scale developers. Security & Vulnerabilities: DeepSeek's Hidden Risks Early post-market research uncovered a critical flaw: DeepSeek lacks adequate safeguards against malicious requests. Unlike OpenAI, which invests heavily in A.I. safety and content moderation, DeepSeek appears susceptible to jailbreaking, meaning users could manipulate it to generate: Hate speech and misinformation Instructions for illegal activities Malicious code or hacking tools While DeepSeek is lax on Western content restrictions, it enforces censorship on internal Chinese topics, raising concerns about political motivations and selective control. International Backlash: Governments Take Action DeepSeek's rapid adoption has triggered swift responses from regulators worldwide: Italy: The Data Protection Authority was one of the first to block DeepSeek due to concerns over data privacy violations. Taiwan: The Ministry of Digital Affairs banned DeepSeek on January 31, 2025, citing national security risks. Australia: Government agencies were prohibited from installing or using DeepSeek AI following concerns about data security and espionage threats. United States: The U.S. Congress issued an official warning, banning congressional staff from using DeepSeek on government devices. A proposed bill could impose fines or jail time for unauthorized use. While DeepSeek may attempt policy changes to regain access in some markets, its early missteps have already fueled global scrutiny. Final Thoughts: DeepSeek's Reality Check At the height of its media frenzy, DeepSeek was hailed as a game-changer -- but does it hold up under scrutiny? No fundamental breakthroughs: While open-source, DeepSeek lacks technological innovations that set it apart from LLaMA or Qwen. Security concerns: Weak safeguards make it vulnerable to misuse, unlike more regulated Western models. Government resistance: International regulators have already restricted access, limiting DeepSeek's global expansion. According to NewsGuard, DeepSeek's chatbot provided inaccurate information 30 percent of the time and failed to answer 53 percent of queries. By comparison, leading A.I. chatbots averaged 40 percent inaccuracy but only 22 percent failure rates -- placing DeepSeek below industry standards. Ultimately, DeepSeek's overnight success is more about timing than technology. The A.I. sector is hungry for breakthroughs, and DeepSeek's arrival created a narrative of disruption. But for now, its technical and ethical flaws suggest it's more hype than revolution. Roman Eloshvili is the founder of ComplyControl, a UK-based provider of AI-powered risk management and regulatory compliance solutions for financial organizations. A C-level fintech executive, Roman has spent over 20 years developing solutions for banks, with an early career in investment real estate and traditional banking. In 2023, recognizing A.I.'s transformative potential in the financial sector, he launched ComplyControl. The platform leverages AI-driven analysis of transactional data and account behavior to identify anomalies and potential risks, enabling financial institutions to take proactive, data-backed measures.

Behind DeepSeek, Chinese AI pushes ahead

In this frantic race, big companies like Alibaba, Baidu and ByteDance are leading the charge, supported by a dynamic group of innovative startups. Today, Chinese AI relies on a diverse ecosystem, with companies specializing in fundamental models, others focusing on specific applications, and groups investing in the infrastructure needed to run these models. A year ago, the Chinese authorities gave implicit support to some forty players in the sector, from the largest to the smallest, including 14 major language models (LLMs). Alibaba recently announced that its Qwen 2.5 Max model outperformed competitors, including Meta's Llama and DeepSeek's V3, on several benchmarks. With a global presence and recognized expertise, the group confirms its diversification and a certain return to favor on the markets, after a long period in the wilderness. The stock has gained almost 50% in one month, after sinking between 2021 and 2023 and limiting its gains to 9% in 2024. Baidu (Ernie) A forerunner in AI, Baidu plans to launch a next-generation model in 2025. Its Ernie Bot chatbot is already a direct competitor to ChatGPT in China, with integration into various services and applications. China's leading search engine announced this week that Ernie Bot will be free to use from April 1. Despite its early start in the AI arena, however, Baidu has struggled to achieve widespread adoption of its LLM. Management recently stated that its latest version, Ernie 4.0, matched OpenAI's GPT-4 capabilities. ByteDance (Doubao) Known for being the parent company of TikTok, ByteDance is also investing in AI with Doubao, a chatbot popular in China with 75 million monthly active users. After starting 2024 behind the curve, it now boasts more than 15 autonomous AI applications. Among them are the text-to-video generator Jimeng and the image generator Xinghui, as well as Kouzi, a personalized chatbot development platform, and Maoxiang, which offers role-playing games and emotional support.According to information available a few months ago, ByteDance was the biggest buyer of Nvidia's H20 AI chips, a model specially adapted for China in response to US restrictions. Tencent (Hunyan) Digital and entertainment giant Tencent is betting on generative AI by integrating it into its gaming, social media and cloud computing platforms. The group has integrated DeepSeek's R1 model into its cloud service, within its Hyper Application Inventor application service. Tencent's language model is called Hunyuan. Since the end of 2024, it has been enhanced with a video generation tool. Tencent was in talks with Apple to integrate its LLM into the iPhone in China, but Alibaba seems to have won the bid. Huawei In the face of U.S. sanctions on semiconductors, Huawei is developing its own AI solutions, including hardware and optimization of deep learning models. The group has spearheaded the production of chips competing with Nividia's, notably by supplying DeepSeek. Huawei and other Chinese AI chipmakers such as Hygon, EnFlame, Tsingmicro and Moore Threads have issued statements in recent weeks claiming that their products would support DeepSeek models. Rising stars: the "Six Tigers" club A group of companies, dubbed the "Six Tigers", are now seen as the vanguard of Chinese AI, as illustrated by a recent MIT article. Stepfun (Step) Founded in 2023 by a former Microsoft executive, Stepfun has rapidly established itself as a major player thanks to its expertise in fundamental models. Its Step-2 model contains over 1,000 billion parameters, making the company a serious competitor to the market leaders. It also stands out for its ambition to develop General Artificial Intelligence (GAI). Zhipu AI (GLM) A spin-off from Tsinghua University, Zhipu specializes in fundamental models and advanced research. Its GLM-4-Plus model rivals GPT-4, while its Ying video generator is inspired by OpenAI's Sora model. However, the company is under scrutiny by the US administration due to its links with the Chinese state. The US Department of Commerce has added the Chinese unicorn to its commercial blacklist due to allegations that it supports the Chinese military. Minimax (Talkie) Minimax stands out for its pragmatic approach: instead of developing its own models, it focuses on consumer applications. Its flagship product, Talkie, a companion chatbot, surpassed its competitor Character.AI in the number of downloads in 2024, according to MIT. Moonshot (Kimi) The company is behind Kimi, the second most widely used AI chatbot in China after Doubao, at least before DeepSeek's iruption. It is backed by Alibaba and other major investors. Its name had been circulating asApple 's partner for China. 01.AI (Yi) Founded by Kai-Fu Lee, former head of Google China, 01.AI is one of China's leading open-source AI companies, with a strong focus on generative AI innovation. Recent rumors suggest that the group has relied on low-cost infrastructure to train its model, much like DeepSeek. Baichuan AI (Baichuan) Founded in 2023 by Wang Xiaochuan (former CEO of Sogou), Baichuan primarily targets the domestic market with specialized solutions for healthcare and education. It was valued at $2.8 billion in July 2024 at a funding round in which Alibaba participated. Challengers on the rise DeepSeek (DeepSeek) DeepSeek is one of the most promising startups in the sector, so much so that it stirred up trouble in the mighty American AI armada earlier this month. Its DeepSeek-V3 model defies Western standards and has been adopted by the likes of Great Wall Motor and China's three biggest telecom operators (China Mobile, China Unicom and China Telecom). DeepSeek's chatbot application overtook Doubao's a few days ago as the most popular artificial intelligence application in China. ModelBest (Luca, MiniCPM) A specialist in small AI models, ModelBest has focused on energy efficiency and privacy. Its MiniCPM range can perform complex tasks on smartphones and connected objects. Infinigence AI (Megrez) The company focuses on optimizing IT infrastructure to improve the training of AI models. Its major innovation is the heterogeneous computing cluster, which combines various chips to bypass US restrictions. And all the others... China's AI ecosystem is vast and includes many other players, including listed companies whose share prices sometimes explode at the mere mention of a tenuous link to AI. Lately, companies such as Capitalonline Data Service and MeiG Smart Technology have benefited from this. The sector is also under close surveillance by the American authorities, who regularly update their various watch lists of companies too close to the Chinese authorities.

FACTBOX China's AI firms take spotlight with deals, low-cost models

BEIJING, Feb 14 (Reuters) - Alibaba's (9988.HK), opens new tab announcement this week that it will partner with Apple (AAPL.O), opens new tab to support iPhones' artificial intelligence services offering in China has again thrust the spotlight on the country's AI industry on the heels of DeepSeek. Following are some of the prominent AI services and their developers to watch in China. DOUBAO Developed by ByteDance, a latecomer to the large language model (LLM) market, the namesake chatbot is the most popular consumer AI app in China. The company's latest LLM, Doubao-1.5-pro, was released in January and outperforms ChatGPT's GPT-4 in some metrics. ByteDance is among the leaders in developing cost-effective LLMs, and its offering is the cheapest available in the market, even more affordable than DeepSeek, according to a Barclays report. Doubao-1.5 was priced at only 3% to 4% of GPT-4's pricing. ByteDance's low-cost advantage is attributed to its use of the "mixture of experts" (MoE) framework, which is common among various AI models in China, including those of DeepSeek. QWEN Developed by e-commerce giant Alibaba (9988.HK), opens new tab, Qwen is the company's flagship LLM and chatbot. It was released in March 2023 and has been rated among the top-tier in global benchmarks. In January, Alibaba released Qwen 2.5-Max and claimed its functionality surpassed that of highly-acclaimed DeepSeek-V3. A strong advocate of open AI development, Alibaba announced in September that many of its AI models would be made open-source. ERNIE Chinese search giant Baidu (9888.HK), opens new tab was among the first in China to launch a ChatGPT equivalent, Ernie Bot. Baidu claims its latest model, Ernie 4.0, matches the capabilities of GPT-4. On February 14, Baidu announced a plan to make its next-generation LLM, Ernie 4.5, open-source from June 30, a major shift in strategy as competition heats up. HUNYUAN Tencent (0700.HK), opens new tab released its flagship model Hunyuan in September and launched its chatbot, Yuanbao, in May 2024. The company's massive user base, acquired through its popular consumer apps like WeChat, gives Tencent a significant potential advantage in pushing out consumer applications in the future. Tencent also offers open-source models and advocates for open-source development. Another advantage of Hunyuan is its focus on multi-modality, which includes text, image, and video generation capabilities. GLM Zhipu AI developed GLM and claims its latest LLM, GLM4, outperforms GPT-4 in some metrics. Initially born out of a Tsinghua University laboratory, Zhipu counts Alibaba, Tencent, and the state-owned fund Zhongguancun Science City Innovation Development among its backers. In January, the U.S. Commerce Department placed Zhipu on an export control entity list. KIMI It was developed by Moonshot, which is backed by the likes of Alibaba, Tencent, and Hongshan. Its chatbot Kimi is known for its long-context processing technology and the company claims the model supports two million input tokens, or units of words, for each query. On January 20, just two days before DeepSeek released the R1 model, Moonshot unveiled its own reasoning model, Kimi 1.5. MINIMAX Minimax is a pioneer among Chinese firms in researching the MoE framework and it open-sourced its most advanced model, MiniMax-01, in January. MiniMax was founded in December 2021, well before the AI craze sparked by the launch of OpenAI's ChatGPT. It is backed by Alibaba and other prominent investors, including Hongshan, Gaorong Capital, and IDG. In addition to LLMs, MiniMax has gained recognition for its social consumer apps, such as Takie and Xingye. 01.AI Similar to DeepSeek, 01.AI has adopted an open-source approach and pioneered the MoE framework. The company claims that its models are trained with fewer resources and have a cost-effective advantage, achieving the lowest cost level of LLM inference in the industry, 1/40 less than GPT-4's list price. Its founder Kai-Fu Lee, a former head of Google China, said recently in a written interview with Reuters that 01.AI has formed a joint partnership with Alibaba to create a lab focused on continuing the development of LLM technologies. The company is backed by companies like Alibaba and Sinovation Ventures. BAICHUAN Baichuan was founded in April 2023 by a team of former executives from the search engine Sogou, including founder and CEO Wang Xiaochuan, who previously served as the CEO of Sogou. The company gained recognition in June 2023 by becoming the first in China to launch an open-source model. (This story has been refiled to correct spelling of 'Doubao,' in paragraphs 5 and 6) Reporting by Liam Mo and Brenda Goh; Editing by Miyoung Kim and Sharon Singleton Our Standards: The Thomson Reuters Trust Principles., opens new tab Suggested Topics:Artificial Intelligence

DeepSeek AI Empowers Chinese Chipmakers, Reducing Reliance on U.S. Tech Amid Export Limits

DeepSeek AI Enhances Memory Efficiency, Strengthening China's Chip Industry Against U.S. Restrictions The Chinese artificial intelligence (AI) company DeepSeek succeeded in its domestic market by providing AI models that strengthen the competitive position of local chip manufacturers. Recently, DeepSeek received model support from four firms, including Huawei, Hygon, Moore Threads and EnFlame, backed by Tencent for their "inference" task-optimized AI algorithms. Implementing this approach allows Chinese AI processors to perform as well as U.S.-made chips despite American export restrictions on advanced chips. DeepSeek optimizes its models to deliver better computational results instead of relying on megahertz-based processing. AI models optimized for inference work efficiently with less powerful chips, thus presenting an economic choice instead of using high-end American hardware. The open-source platform, together with DeepSeek's low fees, has the potential to speed up AI adoption and drive the creation of practical applications, which, according to industry experts, will give Chinese businesses an advantage against US-based companies like in the dominant market.

DeepSeek Is Reportedly Exploring To Develop In-House AI Chips; Stepping Into The Computational Race As Well

DeepSeek is reportedly exploring a "semiconductor" venture, as the firm is now said to be eager to develop in-house AI chips, adding to its computational capabilities. DeepSeek has evolved massively over the past few months, going from a "side project" to a firm that managed to disrupt the global AI industry with the release of its cutting-edge LLM models. Not only was the firm able to compete with OpenAI's GPT o1, but it was also said to have access to limited computing power, although this was refuted in a report that we discussed here. Now, according to DigiTimes, DeepSeek is exploring the possibility of creating its own AI chips, joining the bandwagon of other mainstream AI firms looking to opt for the same route. While the report doesn't mention much about DeepSeek's chip projects, it claims that the company has started a "major recruitment drive," hiring semiconductor experts to lead the project. It isn't as easy as it might sound since developing an AI chip requires an extensive supply chain process, and for Chinese companies, the primary issue lies in acquiring the necessary semiconductor processes due to global sanctions. The only chip access they have is through sources like SMIC, but they, too, are way behind the worldwide semiconductor cadence. For now, it is claimed that DeepSeek has access to around 10,000 of NVIDIA's "China-specific" H800 AI GPUs and 10,000 of the higher-end H100 AI chips, totaling around $1 billion of computing resources. Despite seeing trade restrictions from the US, it hasn't held DeepSeek back at all since the AI firm does have equipment on par with what its competitors own, and likely there's much more as well, which is undisclosed for now. Apart from this, the firm also operates inferencing workloads on Huawei's Ascend AI chips. Hence, they do have a diverse arsenal. The concept of in-house chips for DeepSeek is undoubtedly questionable, given that the firm hasn't evolved as large as competitors like OpenAI, but it is great to see the company exploring the option since it will increase diversity in the AI market. The key question lies in whether DeepSeek manages to reach the implementation stages.

IBM Integrates DeepSeek's AI Models Into WatsonX Platform

DeepSeek's reported smaller budget has come under scrutiny from some in the industry. International Business Machines (IBM) has announced the arrival of China-based DeepSeek models to its enterprise AI development platform. The move comes as more U.S. firms have begun working with the Chinese firm, potentially challenging the long-held dominance of more expensive domestic AI models. IBM and DeepSeek On Thursday, Feb. 13, the U.S. technology giant said it had started using distilled versions of DeepSeek-R1 to help enterprises build AI models for secure reasoning. In a news release, IBM said including DeepSeek-R1 in its watsonx.ai platform was part of the company's commitment to open-source innovation in AI. By including the "best open source models," which IBM has labeled DeepSeek, the company said it was promoting a culture of collaboration. DeepSeek released its R1 AI model last month, claiming it cost significantly less to train than its U.S. rivals. Cheaper AI Will Boost Adoption, Says IBM CEO The potential breakthrough of much cheaper AI with the same power as larger, more established players has sent shockwaves through the U.S. technology industry. IBM CEO Arvind Krishna said the potential cutback on AI costs following DeepSeek's success will significantly boost widespread adoption of the emerging tech. "We will find that the usage will explode as costs come down," Krishna told Bloomberg TV on Tuesday, Feb. 11. "I think it is a validation -- we have been on the point that you do not have to spend so much money to get these models," he added. DeepSeek's Hurdles The level of success DeepSeek has achieved with such a smaller budget has come under fire from some in the industry. In January, Anthropic CEO Dario Amodei claimed that the Chinese company's spending was not significantly different from that of its U.S. rivals. DeepSeek's overall threat to U.S. dominance was 'greatly overstated,' and he claimed Nvidia's stock drop after its release was "baffling." At the same time, a study from BofA Global Research claimed that the $5.58 million cost of training DeepSeek's model was "misleading" as it did not include research, algorithms, and data expenses. However, the researchers claimed that reducing AI training costs was possible. BofA Global Research analysts Brad Sills and Carly Liu wrote: "Cost is the biggest hurdle to adopting AI applications. "We believe advancements in cost could drive the price even lower and therefore adoption higher."

DeepSeek is rumored to be working on in-house AI chips, would compete with NVIDIA

TL;DR: Chinese AI company DeepSeek is developing an in-house AI chip to compete with NVIDIA and AMD, amid global sanctions affecting semiconductor access. The company is recruiting semiconductor experts for this venture. Despite challenges, DeepSeek aims to innovate in the AI chip market, similar to efforts by other companies like OpenAI. Chinese AI company DeepSeek is reportedly working on a new "semiconductor" venture, a new in-house AI chip that would compete directly with the likes of NVIDIA, AMD, and other companies. In a new report from DigiTimes, we're learning that DeepSeek is looking into making its own AI chips, adding that the company has started a "major recruitment drive" hiring semiconductor experts to lead the AI chip project. Chinese companies have multiple hurdles in their way, the biggest being securing the semiconductor equipment and processes because of global sanctions. The only chip access the likes of DeepSeek has access to would be SMIC, but they're many years behind the rest of the semiconductor industry, namely TSMC. DeepSeek isn't the only AI startup looking into developing its own in-house AI chips, with the likes of OpenAI also looking into diving into the AI chip market.