Nvidia CEO Jensen Huang has urged Washington to preserve open access to AI models while also telling investors that the artificial intelligence infrastructure boom is not close to ending.

Huang made his first post on X on July 24, using it to support a letter titled “Open Weights and American AI Leadership.” The letter calls on Washington not to restrict freely downloadable AI models. Huang’s post appeared at https://x.com/JensenHuang/status/2080643682408321103.

The debate centers on “open weights,” a term for model parameters that can be downloaded and run outside the company that built them. Supporters argue that access to those weights helps startups, researchers and developers build on frontier AI without depending only on closed application programming interfaces controlled by a small number of providers.

By the following afternoon, the number of companies backing the letter had increased from 25 to 50. OpenAI, Google, AMD, Cisco, Cloudflare, GitHub, Block and Ollama were among the listed supporters. Two notable names, however, were absent: Anthropic and Amazon.

Anthropic and Amazon were absent from the open AI letter

The letter initially circulated with 25 backers. After Huang’s post drew 11 million views, additional companies signed on, explaining why two versions of the letter showed different supporter counts on the same afternoon.

Amazon is Anthropic’s largest financial backer, and Anthropic uses Amazon’s Trainium chips, among other hardware. Anthropic has also recently moved into the top position in enterprise AI. Google, which has also invested in Anthropic, signed the letter, making Amazon’s connection to Anthropic the more specific point of interest.

Neither Anthropic nor Amazon has publicly explained why it did not sign the letter. Possible explanations include direct business considerations. Anthropic sells closed, frontier-level AI access and has for years made a public safety argument that once model weights are released, they cannot be withdrawn.

Cryptopolitan previously reported that Anthropic has also been conducting its own lobbying effort in Washington, spending $1.97 million in Q2 2026 alone.

Nvidia’s open AI position contrasts with CUDA’s closed role

The letter argues that keeping AI models open is important to prevent any single company from dominating the field. Nvidia’s own position in AI infrastructure, however, highlights a tension in that argument. The company controls CUDA, the software platform widely used to run AI workloads on Nvidia chips. CUDA is deeply embedded across the AI industry, broadly relied upon, and not open.

That makes Huang’s intervention relevant beyond model policy. Nvidia sells the chips and software stack used by much of the AI sector, so rules that affect how models are distributed can also shape demand for the hardware and developer tools needed to train, customize and run them.

During the same week that Huang backed the open-access letter, he spoke with Axios co-founder Mike Allen in an interview focused on investor concerns about AI infrastructure spending.

Chip stocks have declined sharply in recent weeks, even as chipmakers reported strong earnings and continued to face supply shortages. The concern is that major cloud and technology companies are spending hundreds of billions of dollars per year on AI infrastructure, and that spending is no longer being funded entirely from internal cash. Alphabet has moved into negative cash flow, and large technology companies have begun borrowing to keep pace.

Huang says AI buildout is still in its early stages

Asked directly whether the sector is heading toward a bust, Huang replied, “No, not for a while.” When Allen asked whether “this time is different,” Huang accepted the framing.

“This time is different because this is not demand-driven,” Huang said. “This is industrially driven, meaning the fundamental technology of computers is changing.”

The phrase “this time is different” has a long history in financial markets. It was used during the dot-com boom to argue that the expansion would continue, before that boom ended. The expression is often treated as a warning sign when it appears in optimistic forecasts.

Huang acknowledged that the bubble will eventually pop, but argued that the current buildout remains in an early phase. He also said supply constraints affecting chips, land, power and construction labor are helping slow the pace of expansion enough to prevent supply from outstripping demand too quickly.

“We basically are constrained in every single direction, in every single way,” Huang said. “That constraint is good. That constraint is what holds the system back.”

Huang pointed to companies such as Anthropic as evidence that AI is already generating real profits, particularly as businesses identify practical uses for AI agents. The next policy signal to watch is whether Washington treats open model weights as a competitiveness issue, a safety issue, or both, because that framing will determine how much room companies have to release powerful models outside closed platforms.