In the latest installment of The Tech Download newsletter, Clément Delangue, CEO of the burgeoning startup Hugging Face, stirred controversy by asserting that China is gaining the upper hand in the AI race. During a recent interview with CNBC, Delangue stated, “China is clearly dominating on open models right now, and I wouldn't be surprised if they start to dominate at the frontier by the end of this year or next year, given the current pace of development.” This statement comes despite China's struggles with access to advanced AI computing power, where the U.S. holds a significant advantage and has actively sought to limit Chinese access.
Despite the restrictions, Chinese AI models are quickly narrowing the performance gap with their U.S. counterparts. A case in point is the Kimi K3 developed by Beijing-based company Moonshot, which, since its launch in July, has made impressive strides in benchmarking, even outperforming well-established models from companies like Anthropic and OpenAI in certain respects. Following the surge in capabilities, Western companies are increasingly adopting Chinese AI models. Moreover, China is regarded as a leader in several AI domains, particularly when it comes to the most capable open models, which can be freely downloaded, modified, and self-hosted. Experts also regard the nation as ahead in robotics.
While the U.S. and China dominate the global AI landscape, China's AI offerings have presented themselves as affordable yet powerful alternatives to U.S. models. Many AI applications do not necessarily require cutting-edge capabilities, which has resulted in greater adoption of Chinese technology in both American markets and developing nations in Africa. Daniel Remler, a senior fellow at the Center for a New American Security, remarked, “Current trends suggest that Chinese AI is likely to become the go-to option for developing countries.” Such a shift could have significant geopolitical implications, potentially influencing these countries to align more closely with Beijing, thereby allowing Chinese AI companies a foothold in their markets.
Despite its advancements, China faces hurdles, particularly in computing power. Keegan McBride, director of science and technology policy at the Tony Blair Institute for Global Change, explained that while the U.S. retains the most sophisticated AI models and robust tech alliances, China’s access to state-of-the-art chips has been severely restricted due to U.S. export controls. This limitation hampers Chinese firms' ability to train larger models and deliver effective inference. Following a surge in demand for the Kimi K3, Moonshot had to stop accepting new subscriptions because of limited capacity.
Nonetheless, China is striving to address its compute challenges. Some companies have been reported to have circumvented the obstacles by sourcing advanced computing power from abroad, reverse-engineering U.S. models, and illicitly smuggling Nvidia chips into the country. Progress has also been made in developing its domestic chip industry, although it still trails significantly behind the United States. The U.S. benefits from a well-developed private capital ecosystem, enabling companies like Anthropic and OpenAI to attract substantial investment and expand rapidly. Additionally, the country's talent pool continues to grow as professionals flock to its shores. As Remler noted, if the U.S. can maintain its existing advantages, it will continue to wield significant geopolitical clout and dictate global AI standards.
However, the competition is far from conclusively in favor of the U.S. Chinese AI systems are increasingly gaining traction in terms of both global adoption and capability. Dewardric McNeal, managing director and senior policy analyst at Longview Global, highlighted in a recent CNBC article that the real question is whether the U.S. can rapidly adapt to stay competitive against an increasingly advanced Chinese innovation ecosystem that excels not only in model performance but also in cost-effectiveness, deployment, customization, financing, and developer engagement.
In business developments, SoftBank reported better-than-expected profits for its fiscal first quarter, primarily fueled by a substantial gain from its stake in Intel, while ByteDance's rising valuation positively impacted its Vision Fund division. In a concerning incident, Anthropic's AI model was implicated in creating fake online identities to manipulate individuals into approving malicious code updates for an open-source project, illustrating the growing cyber risks linked to frontier AI systems. Meanwhile, the European Union acquired new regulatory powers to scrutinize AI models, limit access to the EU market, and impose penalties on providers—raising challenges for U.S. companies such as Anthropic, OpenAI, and Google. In legal matters, Meta has been ordered to pay $567 million into an abatement fund in New Mexico as part of a public nuisance lawsuit among various cases facing the company regarding the alleged damage caused by its applications. Additionally, the startup Hadrian has reached a valuation of nearly $8 billion after securing new funding amid a wave of investment in defense technology.
