Nvidia urges Washington to keep AI open as Huang calms investors

Nvidia’s Jensen Huang told Washington to keep AI open and told Wall Street the boom is far from over. All while running the most indispensable company in AI. Jensen Huang made his first post ever on X on July 24.

He used it to back a letter called “Open Weights and American AI Leadership,” which asks Washington to leave freely downloadable AI models alone. By the following afternoon, the number of companies backing that letter had gone from 25 to 50. OpenAI, Google, AMD, Cisco, Cloudflare, GitHub, Block and Ollama are also on the list.

However, two major companies are missing: Anthropic and Amazon. Why Anthropic skip Jensen Huang’s open AI letter? The letter went out with 25 backers initially.

When Huang’s post got 11 million views, new signatures came in. This is why two copies of the letter showed different counts on the same afternoon. Still, the two missing names are the most telling part of the list.

Amazon is Anthropic’s biggest financial backer, and Anthropic runs on Amazon’s own Trainium chips, among other hardware. Anthropic recently moved into the top spot in enterprise AI. Google, which has also put money into Anthropic, signed anyway, which makes the Amazon connection the more specific one.

Neither company has said why it stayed off the list. The explanations range from straightforward business logic. Anthropic sells closed, frontier-level access.

Anthropic has made a safety argument publicly for years, which is that once model weights are released, there is no pulling them back. As previously reported by Cryptopolitan, Anthropic has been running its own parallel lobbying operation in Washington, spending $1.97 million in Q2 2026 alone, which makes its absence from the letter a position, not an oversight.

Nvidia’s open AI message comes with a catch While the letter says that keeping AI models open is necessary for keeping any one company from having to rule, Nvidia shows a contradiction. The company controls CUDA which is the software needed to run AI on