Key Takeaways:
Nvidia CEO Jensen Huang made his first-ever X post to defend open-weight AI models, warning that restrictive regulations could drive innovation overseas.
Nvidia and 24 other tech companies pushed back against proposed US curbs on open-weight AI models, warning that restrictive rules could cede AI leadership to Chinese rivals already outpacing American models in daily use.
"Policymakers have an important opportunity to act, keeping the frontier plural by avoiding premature restrictions on open models that stifle competition or drive innovation overseas," the 25 signatories wrote in a letter shared Friday by Huang on X, his first-ever post on the platform.
The signatories include Microsoft, Meta Platforms, Hugging Face and IBM. The pushback comes as White House officials weigh curbs on Chinese models such as Moonshot AI's Kimi K3. Launched July 16 with 2.8 trillion parameters, Kimi K3 ranks among the best-performing models globally, open or closed. White House adviser Michael Kratsios has said Moonshot copied a US model to build it.
The stakes are high for Nvidia, which lost nearly $600 billion in market value in a single day in January 2025 when China's DeepSeek demonstrated that competitive AI could be built with fewer chips. Nvidia shares, trading at roughly 35 times forward earnings, face additional regulatory uncertainty as Washington decides how to balance national security concerns against the industry's push for openness.
About a year ago, the US government endorsed open models in its own AI Action Plan, calling them a strategic asset for the country. The shift in tone reflects growing anxiety about intellectual property theft as Chinese models close the gap. The letter draws a distinction between distillation — training one model using another's output — and outright theft, calling the former normal research that has driven progress across the field.
The signatories argue that open models let many teams inspect code, spot flaws and fix them faster than closed models held by a few firms, which they call a single point of failure for security. OpenAI and Anthropic, two of the most prominent closed-model developers, did not sign. They have instead warned Washington that strong Chinese open models pose a security risk, creating a split in the industry over how open AI should be.
The $600 billion question for Nvidia
The debate cuts to the core of Nvidia's business model. Open models expand the market for AI inference, driving demand for Nvidia's GPUs across a wider range of applications and customers. But if Washington restricts open-weight distribution, it could slow adoption and dampen the GPU demand that has propelled Nvidia's revenue to record levels. The company's data center revenue reached $47.5 billion in its most recent fiscal year, more than quadruple the level two years earlier, as hyperscalers raced to build out AI infrastructure.
The next few weeks will determine the regulatory path. The letter's signers include some of the largest buyers of Nvidia's chips, giving their views weight in Washington. A restrictive outcome could reshape the competitive dynamics of the $200 billion-plus AI chip market, where Nvidia commands an estimated 80 percent share of training chips.
This article is for informational purposes only and does not constitute investment advice.
