On July 24, 2026, Nvidia CEO Jensen Huang broke his silence on X to share a letter titled 'Open Weights and American AI Leadership,' urging Washington not to ban open-source AI models. The letter, hosted on Nvidia's server and co-signed by 25 companies initially, quickly gained traction, with over 70 signatories by July 26, including Microsoft, Google, and OpenAI. Notably absent were Anthropic and Amazon.
Background and Trigger
The letter was a response to escalating tensions between closed-source and open-source AI camps. On July 17, Dean Ball, OpenAI's Head of Strategic Futures and former White House AI policy advisor, called an open-source-dominated world a 'dystopian nightmare' and suggested using regulatory uncertainty to deter US companies from using Chinese open-source models. This followed the release of Kimi K3 by Moonshot AI, a 2.8 trillion parameter open-source model with top-tier performance at low cost, threatening the scarcity of closed-source frontier models.
Divisions Within the US Government
Ball's remarks sparked backlash from within the US government. David Sacks, chair of the President's Council of Advisors on Science and Technology, accused closed-source labs of 'regulatory capture.' Deputy Secretary of Defense Michael called Ball the 'supreme village idiot' of AI. Meanwhile, Treasury Secretary Bessent called for investigations and sanctions on Chinese AI models, highlighting deep internal splits.
Industry Response
On July 22, nearly 200 startups via the 'Small Tech Association' urged Trump not to ban Chinese open-source models. Particle founder Doshi warned that a ban would 'instantly kill hundreds of US companies.' The subsequent letter from major firms solidified industry opposition, driven by economic necessity: Chinese open-source models offer high performance at a fraction of the cost, enabling US small businesses, developers, and institutions to afford AI without vendor lock-in.
Impact of Chinese Open-Source Models
Chinese open-source models now account for over 40% of downloads on Hugging Face, surpassing US models for the first time. Alibaba's Qwen family leads globally, with over 200,000 derivative models. On OpenRouter, the top five most-used models in June 2026 were all Chinese. These models provide cost-effective, deployable AI, with performance gaps to closed-source leaders narrowing to just 3.3% (Mozilla report, July 2026).
Corporate Stances
Each signatory had pragmatic reasons: Nvidia sells GPUs and benefits from AI adoption; Microsoft hosts models on Azure; Meta, despite shifting its flagship Muse Spark to closed-source, supported the letter to legitimize distillation used in its training; OpenAI signed to avoid isolation, though its strategic head had advocated for regulation. Anthropic, the only holdout, relies entirely on closed-source revenue and released Claude Opus 5 on the same day.
Broader Implications
The letter reflects a fundamental shift: Chinese AI firms are no longer mere imitators but innovators, with papers like K3's KDA architecture published months before release. The US faces a dilemma: banning Chinese models would harm its own ecosystem, while allowing them deepens dependence. This mirrors past trade dynamics, where the US both relies on and tries to contain Chinese manufacturing. Meanwhile, China's policy focus on building AI infrastructure contrasts with US regulatory debates.
In conclusion, the letter is a testament to China's rise in AI competitiveness, forcing the US to publicly grapple with its reliance on Chinese innovation. The 70 signatures confirm that Chinese open-source models have become indispensable to the US AI ecosystem.
