Moonshot AI pauses Kimi K3 subscriptions as overwhelming demand strains compute capacity

China's Kimi K3 and the rise of open-weight AI models

Moonshot AI's Kimi K3 shows how opening a model to outsiders can turn other companies' computing power into a competitive advantage Just three days after the Chinese developer Moonshot AI unveiled Kimi K3 on July 17, it stopped accepting new subscriptions. Demand for the enormous artificial intelligence model had overwhelmed the company's available computing capacity. Yet Moonshot says it plans to release K3's full weights by July 27, which would allow other organizations to host and modify the model themselves. In a post on X, Dean W. Ball, OpenAI's head of strategic futures, argued that a world dominated by open-weight models could lead to "full AI communism" -- a future he described as "a dystopian hellscape." But giving away the weights of a top-tier AI model may actually make practical sense. For Moonshot, doing so could spread K3 far beyond the company's own computing infrastructure and help it compete with leading U.S. systems whose developers keep their weights private. Before reaching for dystopian prophecies, Ball acknowledged in the same post that Kimi K3, a 2.8-trillion-parameter system, appears to be a very good model. In benchmarks published by Moonshot, Kimi K3 generally lands ahead of OpenAI's GPT-5.5 and Anthropic's Claude Opus 4.8 but behind Claude Fable 5 and, on some tests, GPT-5.6 Sol. The company reports that it performs especially well on web searches and business workflows. These results place Kimi K3 among the strongest modern AI systems without showing that it has surpassed the leading American models. What most clearly sets K3 apart from GPT-5.6 Sol or Claude Fable 5 is Moonshot's plan to release its weights openly. On supporting science journalism If you're enjoying this article, consider supporting our award-winning journalism by subscribing. By purchasing a subscription you are helping to ensure the future of impactful stories about the discoveries and ideas shaping our world today. A large language model such as Kimi, GPT or Claude is, at bottom, an enormous collection of numbers. During training, the model ingests vast amounts of text and other data while an algorithm adjusts billions or trillions of numerical dials -- the weights -- until the system can predict and eventually generate humanlike content. Much of what the model has learned is encoded in those numbers. American AI labs generally keep the weights of their most capable models on private servers. A user can talk to GPT or Claude through an app but never possess the model itself. An open-weight release inverts this arrangement: The developer posts the trained weights publicly, allowing anyone with sufficient hardware to run the model privately or customize it. Chinese leaders have embraced that approach as part of a broader political message. At the 2026 World Artificial Intelligence Conference in Shanghai, Chinese president Xi Jinping called for "open source, openness, collaboration and sharing" to prevent "new historical injustice in AI." But that rhetoric blurs an important distinction. "Open weight is not the same as open source," says James Landay, a professor of computer science at Stanford University. There's been a lot of mixing up between the two." An open-source AI model should provide more than its weights, but also enough information and code for outsiders to study and modify the system -- although researchers and standards groups continue to debate how much of the training process must be disclosed. An open-weight release can leave the model's data and development history opaque. Landay says that uncertainty should make organizations cautious about adopting models whose provenance they cannot fully examine. "We might not know what's in there, we might not know if they phone home in some ways with our data," he warns. But such opacity does not erase the commercial logic of releasing the weights. "They still make money in a number of ways," says Kyle Chan, a fellow at the Brookings Institution who studies Chinese technology policy. Moonshot can continue selling access through its application-programming interface and subscription products even after other companies begin hosting K3. Moonshot is younger and less richly resourced than the largest U.S. frontier-model developers. Chan argues that releasing a strong model's weights gives such a company another way to compete: widespread adoption can expand its influence even when it lacks enough hardware to serve every user itself. U.S. export controls introduced in 2022 have restricted Chinese laboratories' access to advanced AI chips. "This constrained compute capacity for the Chinese AI labs," Chan says, "they talk about it all the time." The restrictions do not fully explain Chinese developers' embrace of open weights, but Chan says limited compute makes the strategy more attractive. Chan expects major hosting platforms such as Databricks to begin offering K3 after its weights are released. "By open-weighting it, you basically unlock all that extra compute capacity that other people have invested in and built up," he says, effectively turning outside providers' infrastructure into part of the model's distribution system. "It's like an amplifying effect." Meta helped popularize open-weight large language models when it released Llama in 2023. DeepSeek brought new attention to China's open-weight strategy with its R1 model in early 2025. OpenAI and Google now offer open-weight families of their own while reserving their most capable systems for controlled services. The U.S. startup Thinking Machines Labs joined the field on July 15 with its first model, Inkling. Chan believes the leading U.S. labs risk ceding ground if Chinese models become the systems that companies and developers around the world can most readily adopt. "I think it's a mistake to give up on open weight," he says. "The success of the Chinese models is showing its value." That doesn't imply China will necessarily win the AI race, Landay says. "New open models may come from those big players and not from Alibaba or the Kimi people," he says. "But if I could predict it, I'd be one of those rich guys driving an expensive car." Still, Landay expects competition from Chinese developers and smaller U.S. laboratories to put greater pressure on leading companies to release more capable open models. "I think the bigger lesson is that open ecosystems in the long run win," he says.

The US may find it hard to shrug off Moonshot's AI shockwaves

The latest Chinese models such as Kimi K3 look set to bring greater price competition More than a year after the DeepSeek shock, another "open" AI model from China has sent a severe tremor through Silicon Valley. This time, it may be harder for the US AI companies to shrug off the challenge. That has made a response from America all the more likely -- with unpredictable consequences for both the AI companies and their customers. The latest shockwaves were caused by last week's launch of Kimi K3, a model from Chinese start-up Moonshot that has come close to matching the most advanced competitors from the US frontier AI labs. The angst this has induced is reminiscent of the worries over DeepSeek's R1. Its low training costs appeared to threaten the far more expensive models being developed in the US. Wall Street came to terms with that threat, believing that US frontier models still had a meaningful lead, even if it had been cut to less than a year. But Kimi K3 -- along with other Chinese models like last month's GLM 5.2 and the Qwen 3.8 model that Alibaba announced at the weekend -- have all but erased the time advantage. Some US officials have been quick to accuse Moonshot of copying US rivals (using a technique called distillation) and getting around export controls on advanced US chips. There are signs, though, that innovation is playing an increasingly important part. Even a top executive at OpenAI conceded that Kimi K3's advances did not look like the result of copying, while its architectural improvements over earlier models have drawn admiration in the US. This looks like real competition, not just emulation. In economic terms, the competitive effects are not clear-cut. Open-weight models (a limited form of open-source software) are cheaper because the companies that develop them do not look to recover their training costs, and because they are run either on a customer's own systems or hosted by cloud-computing companies that compete on cost. Very large models like Kimi K3, though, are expensive to run. Also, the true cost to a user is not the price of a token (the basic unit of output, on which pricing is based), but the cost of completing a task. Some models use fewer tokens to do that, or generate fewer hallucinations (meaning less human labour is needed to validate their responses). That said, the latest spate of Chinese models looks set to bring greater price competition to the most advanced forms of AI. The question now is how US AI companies, as well as policymakers in Washington and Beijing, respond. The US frontier labs need to accelerate their shift from selling raw intelligence to packaging it into agents that can complete more valuable tasks. This worked with coding, but models like Kimi K3 are catching up fast on that front, forcing the labs to keep finding new uses. The stiffer competition also increases the urgency for frontier labs to get closer to their users, tapping customer data and business context to improve their relevance, while embedding their models into customers' workflows. The question for OpenAI and Anthropic is whether they can show real headway on these measures in time for their mammoth IPOs. The geopolitical ramifications are harder to predict. The Kimi K3 launch came just as President Xi Jinping was promoting open source AI as central to China's technology pitch to the world. Just days later, though, the FT reported that the country was considering export controls on some technologies -- including restricting access to the most advanced models' weights, the parameters which shape how they respond. This points to an emerging tension in the tech strategy of China Inc. Alongside an export-led drive geared to wide availability and low cost, there is a growing awareness of the need to defend homegrown technologies. It is not clear how that will play out. Washington, for its part, shows every sign of a knee-jerk response. Treasury secretary Scott Bessent signalled a possible crackdown on Chinese companies that use distillation -- though IP leakage from models that are publicly available would be hard to prevent. The US has other reasons to try to stem advanced Chinese AI, for instance the cyber risk the models might represent. Even if it succeeded in making its domestic market a protected zone for US frontier models, though, the result might be counter-productive in global terms. Companies in the rest of the world would still be able to access cheap Chinese models. And simply barring their use by US entities wouldn't solve the cyber threat.

China's Moonshot pauses Kimi subscriptions amid hot demand, IPO push

SHANGHAI/HONG KONG, July 20 (Reuters) - Chinese startup Moonshot AI has temporarily paused new subscriptions after demand for its newly launched Kimi K3 model strained capacity, a bottleneck that comes as the company seeks fresh funding and prepares for a potential Hong Kong listing. Moonshot is in the process of unwinding its current offshore structure ahead of a Hong Kong initial public offering, two sources with knowledge of the matter said. The company has engaged financial advisers including Goldman Sachs and China International Capital Corp to discuss the IPO plan, although the timetable remains fluid, said one of the sources and a third person with knowledge of the IPO plan, declining to be named as the information was confidential. CICC did not immediately respond to a request for comment. Goldman Sachs and Moonshot declined to comment. Founded in 2023 by Yang Zhilin, an AI researcher who pursued doctoral studies at Pittsburgh-based Carnegie ⁠Mellon University, Moonshot is one of China's most closely watched AI startups. It raised more than $2 billion in May from investors including Meituan, China Mobile and CPE, according to a fundraising teaser seen by Reuters, bringing the company's total historical fundraising to over $5.5 billion. It has since begun seeking up to $2 billion in fresh capital, with its valuation reaching $30 billion in June, the teaser showed. Strong demand for increasingly powerful models is boosting investor interest in China's leading AI startups, but it is also raising the need for costly computing infrastructure. Moonshot's competitors, including DeepSeek, have recently sought external capital to expand compute capacity as Chinese AI firms race to narrow the gap with U.S. rivals. Moonshot said on Sunday that since Kimi K3's release, it has drawn massive user interest, leading to "unprecedented compute challenges." Over the past 48 hours, user requests had sharply exceeded forecasts and were approaching the limits of existing clusters, the company said. Moonshot said it would pause new consumer subscriptions immediately and allocate available computing power to current paid users, who would be ⁠unaffected by the shortage. The company also said it would split future memberships into two plans, including one just for coding, a move aimed at matching compute resources more precisely with user demand. CAPACITY CRUNCH FOLLOWS KIMI K3 RELEASE The capacity crunch follows a strong reception for Kimi K3, which Moonshot unveiled on Friday as a 2.8 trillion-parameter model, making it the world's largest open-weight AI system, according to the company. "Kimi K3 has received far more love than we expected, and our GPUs are feeling it," Moonshot said on X, adding that new subscription spots ⁠would reopen in batches as capacity was added. Kimi K3's size and focus on coding and agent-style tasks make it more expensive to serve at scale, as such workflows typically require repeated model calls and heavy inference capacity. While open-weight models allow users to download and customise the underlying system, analysts say few users are likely to host a model of Kimi K3's scale ⁠themselves because of the hardware costs involved. Moonshot said Kimi K3 performed competitively with leading U.S. models in some technical tasks while independent evaluations have also pointed to strong performance. The launch comes as other Chinese AI firms such as Z.ai and MiniMax release more capable models at lower cost, challenging assumptions that China's model developers lag U.S. peers by ⁠months. Alibaba, an investor in Moonshot, said on Sunday Qwen3.8-Max-Preview, its 2.4-trillion-parameter model, had debuted on its AI platforms ahead of a planned open-weight release. U.S. export controls on advanced Nvidia chips, however, have made access to computing power a key constraint for these ambitious companies. Writing by Eduardo Baptista in Beijing; Reporting by Kane Wu in Hong Kong and Samuel Shen in Shanghai; additional reporting by Yantoultra Ngui in Singapore and Laurie Chen in Beijing; Editing by Sam Holmes Our Standards: The Thomson Reuters Trust Principles., opens new tab * Suggested Topics: * Disrupted * Capital Markets Kane Wu Thomson Reuters Kane Wu covers M&A, private equity, venture capital and investment banks in Asia. She tracks the region's most high-profile deals, fundraisings as well as investment trends amidst geopolitical, macroeconomic and regulatory changes. She was nominated for a SOPA Excellence in Business Reporting award for coverage of China regulatory crackdown in 2021. Prior to Reuters, she worked at the Wall Street Journal and also wrote about Asia's loan market for Thomson Reuters Basis Point. She is based in Hong Kong.

ETtech Explainer: Why Moonshot AI paused Kimi K3 subscriptions within days

Chinese AI startup Moonshot AI has stopped new subscriptions for its Kimi K3 model. This decision follows a sharp surge in user demand for the advanced AI. Additionally, Kimi K3 is attracting attention for its high-end coding performance as well as for offering this capability at a much lower cost than competitors. Following a sharp surge in user demand, Chinese AI startup Moonshot AI has temporarily stopped accepting new subscriptions for its latest AI model, Kimi K3, which it launched last week.Kimi K3 has been attracting attention for offering high-end coding performance at a much lower cost than leading frontier AI models such as Anthropic's Fable 5 and OpenAI's GPT models.Subscriptions on holdIn a post on microblogging platform X, the official Kimi AI

This Chinese AI Model Became So Popular It Stopped Taking New Users -- and Analysts Say Nvidia, Micron Coul

Two days after its launch on July 17, Moonshot AI temporarily stopped new sign-ups for its Kimi K3 model due to high demand and compute constraints. Anni Sen, managing partner of BluBird Capital, told MarketWatch that the surge in demand for Kimi K3 indicates a potential long-term increase in chip demand. She stated that this development is "a positive tailwind for the memory trade." While Kimi K3 is efficient to run, the model's need to hold 2.8 trillion parameters in active memory could make it challenging for enterprises to operate on their servers, according to Sen. This could lead to an increase in AI workloads and further boost demand for chips. Wedbush analyst Matt Bryson noted that if Chinese AI models continue to gain traction, it would be "arguably good for memory vendors," as demand for high-performance memory is expected to rise. Over the past 5 days, SK Hynix stock declined 10.05%, while Micron stock dropped 12.29% on the NASDAQ Meanwhile, SK Hynix's Chairman Chey Tae-won has argued for expanding memory supply rather than maximizing profits from the current shortage. He expects overall memory demand to rise by more than 50% to 60% next year, with AI-specific demand potentially climbing by 60% to 100%. Disclaimer: This content was partially produced with the help of AI tools and was reviewed and published by Benzinga editors. Image via Shutterstock Market News and Data brought to you by Benzinga APIs To add Benzinga News as your preferred source on Google, click here.

Kimi K3: China's Moonshot AI pauses new Kimi K3 subscriptions amid surge in demand

In a social media post, Kimi AI stated the surge in demand over the last 48 hours has pushed the company's current capacity close to its limits. Hence, to ensure a stable experience for existing subscribers, new subscriptions have been temporarily paused. China's Moonshot AI has temporarily paused new subscriptions for its latest AI model Kimi K3 amid a sharp surge in demand, as per a statement by the company. In a social media post, Kimi AI stated the surge in demand over the last 48 hours has pushed the company's current capacity close to its limits. Hence, to ensure a stable experience for existing subscribers, new subscriptions have been temporarily paused. "Over the past 48 hours, demand has pushed close to the limits of our current capacity. To protect the experience of existing subscribers, we're temporarily pausing new subscriptions and prioritising compute for current members," it said. The company further noted, while compute resources are being prioritised for current members, "Existing subscribers will not be affected." Apart from this, the company is adding further capacity and will reopen new subscription spots in batches. "We're adding capacity as fast as we can and will reopen new subscription spots in batches," it said. Furthermore, the company will also introduce two more focused membership plans to enable more precise allocation of computing resources and help maintain a stable user experience. "Going forward, we'll also split membership into two more focused plans: Kimi Membership for Kimi Web, App, and Work; and Kimi Code Membership for coding workflows. This will help us match compute more precisely and keep the experience stable," it said. Developed by the Chinese startup Moonshot AI, Kimi K3 is a 2.8 trillion-parameter model with a 1-million-token context window. "It is the world's first open 3T-class model, designed for frontier intelligence across long-horizon coding, knowledge work, and reasoning," the company said in its blog. "Kimi K3 is available today on Kimi.com, Kimi Work, Kimi Code, and the Kimi API. At launch, Kimi K3 will use max thinking effort by default, with low- and high-effort modes to be introduced in subsequent updates," the company added. The full model weights will come out by July 27, 2026. The startup claimed that its model delivered frontier-level performance across its evaluation suite, although it still lagged behind Claude Fable 5 and GPT 5.6 SoI. (ANI)

Nvidia, Micron just got hit by Kimi K3 AI model from Beijing

Every market boom eventually runs into the same problem, and it rarely shows up where investors expect it. You ride a simple story - in this case, artificial intelligence needs endless chips and data centers - straight up the chart until the numbers stop feeling real. Then something jolts the narrative, not by breaking the old leaders, but by proving they aren't alone anymore. Right now, that jolt is coming from Beijing. Chinese startup Moonshot AI just introduced Kimi K3, an open‑weight artificial intelligence model built with roughly two point eight trillion parameters, a scale that rivals the largest systems in the world. Independent benchmarks cited by outlets such as the BBC say Kimi K3 performs in the same neighborhood as top models from OpenAI and Anthropic on many reasoning and coding tasks. Kimi K3 is not about stealing Nvidia's customers overnight. Still, my take is that it attacks the assumption that U.S. companies will always enjoy a comfortable technological lead, and that is exactly where the market decided to hit the brakes. Chip investors just got a wake‑up call Nvidia's stock, along with Micron Technology and other chipmakers, slid as traders digested the idea that a Chinese lab could make top‑tier artificial intelligence models cheaper and more accessible, The Wall Street Journal reported. The PHLX Semiconductor Index fell about 10% in the week of Kimi K3's launch, its steepest weekly drop since April 2025, as investors dumped AI‑linked names, The Journal reported. More Artificial Intelligence: Tech stocks broadly sold off, with the Nasdaq down roughly one point four percent, but chipmakers took the brunt because their valuations are built directly on artificial intelligence demand. David Sacks and Bill Ackman have been warning that China's new model narrows America's lead in artificial intelligence and heightens policy and national‑security risks around data centers and cloud infrastructure, according to Benzinga. Their argument is simple, and it matters. If you own these stocks or work in the sector, cheaper high‑end models from China could force corporations and governments to rethink how much they spend on Western chips and cloud capacity. VCG / Getty Images What is the Kimi K3 AI model? Kimi K3 is Moonshot AI's latest flagship model, built as an open‑weight system that developers can download, inspect, and modify, unlike the closed models most U.S. users rely on. The model clocks in at around 2.8 trillion parameters, making it the largest open system yet disclosed from China and placing it firmly in the same size class as elite Western models, according to the BBC. Artificial Analysis and other benchmarking firms have found that Kimi K3 competes closely with leading reasoning models, ranking near the top on tasks like web interface engineering and complex coding. On paper, this does two things investors care about. First, it proves that China's labs can match or nearly match state‑of‑the‑art performance in core commercial tasks such as software development and data analysis. Second, by making such a system open and, in some configurations, cheaper than Western closed alternatives, it threatens the idea that only Silicon Valley giants can deliver cutting‑edge artificial intelligence at scale. At‑a‑glance numbers behind the Kimi K3 shock * Kimi K3's parameter count is about 2.8 trillion, according to BBC and Inc. * Weekly drop in PHLX Semiconductor Index, roughly 10%, The Wall Street Journal reported. * Nasdaq declined on the main sell-off day, around 1.4%, The New York Times confirmed. * Nvidia and other chipmakers' intraday moves were between 2% and 4% percent down. When I compare those numbers against how fast Nvidia and its peers ran up this year, I believe that this is less a crash and more a sentiment reset around how durable the AI spending story really is. Why Kimi K3 matters for your wallet If you own Nvidia or Micron, you are basically betting that artificial intelligence workloads keep growing faster than anyone can build cheap alternatives. Kimi K3 does not demolish that thesis, but it introduces real competition in what used to be a one‑way narrative. TheStreet has covered how investors hope strong Nvidia earnings can give the broader rally more life, and how analysts are still raising price targets ahead of big quarters, but days like this show that the path will not be smooth. Cheaper or open models from China could push some companies to experiment with lower‑cost infrastructure or shift workloads, which would chip away at the premium multiples that data‑center suppliers enjoy. For workers, especially in tech and cloud‑related roles, this is a reminder that if artificial intelligence becomes more globally commoditized, the pricing power and hiring power concentrated in a few U.S. giants could spread out, or in some areas, shrink. For your kid growing up into this market, the story might not be about one or two American companies owning the future, but about whether they can stay ahead of a crowded field where China, and other countries, release powerful tools for anyone to build on. How to think about Nvidia and Micron now Short‑term, this kind of shock tends to pass once investors see hard earnings data. Part of the volatility in chip stocks has been driven by traders locking in profits ahead of major quarterly reports from Nvidia and other megacap technology names, The Journal noted. If those numbers show that demand for U.S. chips and cloud capacity is still growing, the Kimi K3 headlines may fade into the background, at least until the next competitive threat emerges. Long‑term, though, you should be asking different questions about your exposure to AI‑linked names. Do you own them because you believe they can stay ahead of global competition, including open systems from Beijing, or because you assume their lead is guaranteed by politics and hype? If it is the latter, this week is a warning to rethink that assumption. I would treat this sell-off as an opportunity to revisit whether your portfolio is overly concentrated in a single story (i.e., artificial intelligence needs endless Western chips). It's clear that AI technology itself is becoming cheaper, more open, and more evenly distributed around the world. The Arena Media Brands, LLC THESTREET is a registered trademark of TheStreet, Inc. This story was originally published July 21, 2026 at 2:47 PM.

Another 'DeepSeek moment'? What China's Kimi K3 means for the global AI industry - The Korea Times

Visitors at the booth for Moonshot's Kimi K3 during World AI Conference in Shanghai, July 17. AP-Yonhap The launch of Moonshot AI's Kimi K3 has revived an intense debate that has raged in Silicon Valley ever since DeepSeek's shock breakthrough last year: whether China can overcome its limited access to advanced chips to match the performance of the United States' frontier artificial intelligence models. Trillions of U.S. dollars might rest on the answer, as some see Chinese developers' success in increasing performance through architectural innovation as weakening the rationale for America's vast spending on AI infrastructure. The arrival of K3 has sent shock waves across the global AI industry. The 2.8-trillion-parameter open-weight model performed close to the latest frontier systems from OpenAI and Anthropic, fuelling concerns in Silicon Valley that China had closed its model-development gap with the U.S. to weeks rather than months. For some analysts, the launch is about more than a shift in the benchmark contest between U.S. and Chinese models. It offers a test of how China's advances in AI software could affect an industry built around US dominance in frontier models - and the computing hardware that powers them. "For much of the past three years, the global artificial intelligence story has been framed as an American one. OpenAI, Google and Anthropic have dominated headlines, while investors have poured money into the companies supplying the chips, memory and data centres needed to power the AI boom," said Sunil Tirumalai, head of emerging markets and Asia equity strategy at UBS, in a research note. "But a new question is beginning to emerge: what happens if Chinese AI models become much better - and much cheaper?" Several investment banks have described K3 as evidence that Chinese laboratories are moving beyond their earlier reputation for producing cheaper but less capable alternatives to Silicon Valley's top systems. Morgan Stanley said K3 represented an "all-round catch-up" in model scale, performance and pricing. Goldman Sachs described China's progression as moving from the cost efficiency demonstrated by DeepSeek to stronger intelligence and, with K3, a greater ability to charge for frontier-level performance. Bernstein said the model showed that China's leading laboratories could continue to keep pace with the U.S. frontier. Meanwhile, Chinese AI models are increasingly finding users outside China. Their appeal is not that they are the smartest models in the world. Rather, they are often "good enough" for many everyday tasks while costing a fraction of leading U.S. alternatives, Tirumalai said. But the emergence of cheaper, more efficient models from China does not necessarily mean that demand for underlying hardware such as semiconductors and data centres will fall. As some analysts have pointed out, when technologies become cheaper, people often use them more, not less. Moreover, there is evidence to suggest that K3 does not offer the dramatic cost reductions that some assume. Nomura, citing benchmark provider Artificial Analysis, estimated K3's average cost at about $0.94 per task, close to OpenAI's GPT-5.6 Sol. It is also worth looking back to what happened following the "DeepSeek moment" in January 2025. After the Chinese start-up's R1 model shocked global markets by matching top-tier U.S. systems at a fraction of the cost, there were widespread fears that Silicon Valley might have massively overinvested in AI compute. But after the shock died down, the world's technology ecosystem doubled down with huge capital expenditure on AI hardware, underscoring the idea that cost-effective models do not necessarily reduce compute demand. Washington is also seeking to reinforce America's existing advantage in computing resources. U.S. Treasury Secretary Scott Bessent said in an interview released this month that the U.S. could soon account for 80 per cent of the world's computing power, although he did not explain the methodology or provide a timetable. Tirumalai said total infrastructure demand depended on the computing required for each AI task multiplied by the number of tasks performed. Efficiency should reduce the first figure, but cheaper and more capable models could cause the second to rise more quickly as adoption expands. He compared the process with mobile communications. Successive generations of networks sharply reduced the cost of transmitting data, but overall consumption surged as video streaming, gaming and social media created new uses. "The rise of Chinese AI should not be viewed simply as a threat to today's winners," he said. "It may instead reshape where value is created and who captures it." Lei Meng, China equity strategist at UBS, said technology and AI would remain the main investment themes in 2026, with sectors linked to AI capital expenditure benefiting from spillover beyond narrowly defined AI spending. The more immediate pressure from K3 may therefore fall on the margins of model providers rather than on demand for computing infrastructure. Technology investor Gavin Baker argued that an industry dominated by only a few frontier laboratories would allow those companies to preserve high inference margins and gradually expand into infrastructure and software. But greater competition from Chinese and other open-weight models could make it harder for OpenAI and Anthropic to capture such a large share of AI spending. That could leave more spending for chipmakers, cloud providers, data-centre operators and application developers. Models developed in China would still require processors, memory and electricity wherever overseas users choose to deploy them, according to Baker. Moonshot's own roll-out offers an early test of that argument. Despite claiming a sharp improvement in development efficiency, the company is still racing to bring more graphics processing units online after demand for K3 overwhelmed its existing capacity. Read the article at SCMP.

China's Moonshot AI pauses subscriptions for powerful Kimi K3 model due to surging demand

Chinese AI firm Moonshot has paused new subscriptions for its powerful new Kimi K3 model due to high demand - days after it was released with capabilities rivaling those of OpenAI and Anthropic. The Beijing-based Moonshot said Sunday it was grappling with "unprecedented compute challenges" and would temporarily focus on ensuring it could serve its existing paid users. The large-language model was trained on 2.8 trillion parameters, making it the largest open-source model ever released. "Kimi K3 has received far more love than we expected, and our GPUs are feeling it," Moonshot wrote in an X post. "Over the past 48 hours, demand has pushed close to the limits of our current capacity." Moonshot was founded in 2023 by AI researcher Yang Zhilin, who once studied at Carnegie Mellon University in Pittsburgh. His firm is one of several Chinese AI startups that have gained steam in recent months while releasing powerful open-source models that are available for a fraction of the cost of Anthropic and OpenAI offerings. Benchmark tests showed Kimi K3 outperforming Anthropic's Opus 4.8 model and OpenAI's ChatGPT 5.5 on most coding tasks, though it is still less powerful that Anthropic's cutting-edge Fable model. Meanwhile, Moonshot has begun talks with Goldman Sachs and other firms about a potential initial public offering in Hong Kong, Reuters reported, citing sources with knowledge of the matter. The company, which has raised more than $5.5 billion from investors to date, has seeking an injection of $2 billion in new capital, the outlet reported. Moonshot's valuation was $30 billion as of June. Some US officials are fretting that China is catching up to American firms in AI development despite tough export controls that have limited the country's access to the best computer chips offered by Nvidia. As The Post reported, Anthropic and OpenAI have each accused Chinese rivals in recent months of using unauthorized distillation - in which a more advanced AI model is used to train a new one - to rip off their technology. Last February, Anthropic directly accused Moonshot and two other Chinese firms, DeepSeek and Minimax, of distilling its models. So far, Anthropic has not said whether it suspects distillation was used to help build Kimi K3.