Nearly 200 Silicon Valley companies have petitioned the White House to block China’s open-source models—America is the one that will be hurt.

Nearly 200 Silicon Valley companies, including Proton and Y Combinator, jointly sent a letter to the White House opposing a comprehensive shutdown of China’s open-source models. At the same time, the Trump administration is stepping up accusations against Moonshot AI over distilling Anthropic models and allegedly obtaining Nvidia chips in violation of rules, and neither side can seem to move the other.
(Background recap: U.S. companies are accelerating their shift to China’s AI models to cut costs, Congress is becoming tense, and investigations have begun.)
(Additional context: Huang Renxun powerfully backs China’s AI—saying the U.S. is in “Kimi panic,” “getting the point wrong”: free open source will ignite demand for Nvidia chips.)

Table of contents

Toggle

• More than 200 companies sign on
• Cheap—that’s the key
• Big companies and small companies, two different ledgers

Nearly 200 Silicon Valley companies, including Proton and Y Combinator, jointly wrote to the White House with a request that boils down to a single sentence: don’t shut the door on China’s open-source models. When Playground AI founder Suhail Doshi was interviewed by POLITICO, he put it even more directly: if there’s a one-size-fits-all ban, “hundreds of companies will die immediately.”

The ones that will die aren’t Chinese companies—they’re U.S. startups that can’t afford the subscription tiers of Anthropic and OpenAI.

More than 200 companies sign on

This joint letter was sent out on Wednesday by the newly established Little Tech Association. Its recipients are President Trump and Commerce Secretary Howard Lutnick, with copies sent to Michael Kratsios, director of the White House Office of Technology Policy. The content contains only one request: U.S. leadership needs two things—top U.S. open-weight models for the world, and the right for U.S. developers to continue obtaining the already-existing global open-source models.

The reason this erupted at this point in time is that Moonshot AI has just released its powerful Kimi K3 model, and immediately afterward reports emerged that the Trump administration is considering disabling China’s AI models. In response, a White House spokesperson Liz Huston said in a statement that the U.S. will continue to widen its AI gap with other countries, but did not directly deny that the discussion is taking place.

An unnamed White House official said that any policy announcements would be made directly by the government; before that, all reports are merely “groundless speculation.” This didn’t deny that the discussion exists—instead, it made the outside world even more convinced that the option of a ban is indeed on the table.

Cheap—that’s the key

Why are nearly 200 companies willing to sign on under their own names? The answer is very real: money.

Most startups don’t have the capital to purchase Anthropic or OpenAI usage tiers over the long term. Yet they can download Moonshot AI’s Kimi K3, Alibaba’s Qwen, or DeepSeek for almost free, then fine-tune them on their own servers—or even run them offline.

You can tell in one sentence why the scales of this campaign are tilting so sharply by looking at the difference between free downloadable China open-weight models and U.S. APIs billed by usage. Once an open-weight model is downloaded, it spreads into countless private data centers and mirror sites around the world. Trying to “fully claw it back” is, technically speaking, almost impossible—this is the core argument of the joint letter: a ban can’t stop the spread of the model; it can only block the lifeline for U.S. startups. Doshi put it even more directly: “This is good for Anthropic—we all have to pay for Anthropic.” One sentence lays bare the most ironic part of the ban: it claims to be “to prevent China,” but in the end, it benefits U.S. giants instead.

Big companies and small companies, two different ledgers

This debate also reflects a widening rift within the AI industry. Giants like Anthropic tend to support stricter regulation, citing cybersecurity risks; the startup camp argues that bans provide limited help for security, yet would push them into a competitive disadvantage. Nvidia CEO Jensen Huang previously publicly criticized the U.S.’s “Kimi panic,” saying the country is “getting the point wrong.”

According to two people familiar with the matter, the White House and top cabinet officials did discuss the issue on Monday night, but the option of a “complete prohibition” was not seriously considered. As of Wednesday, the Department of Commerce also had not added any Chinese AI companies to the list of entities subject to import and export control.

In an interview, Harry Godfrey, executive director of Little Tech Association, offered what may be the most practical line in this tug-of-war: policy tools should be “surgical knives, not big sledgehammers.” The real question is which approach uses the lightest touch—one that addresses legitimate cybersecurity concerns without raising costs, without restricting access, and without crushing American innovation.