Jensen Huang has never posted on X.
Not a product tease, not a meme, not a reply to a rival. The Nvidia CEO’s entire social media presence on the platform formerly known as Twitter has been exactly zero posts for the entirety of its existence. So when he finally broke that silence on Friday, he didn’t waste it on small talk. He used it to tell Washington that restricting open-weight AI models would be the kind of catastrophic policy mistake the software industry almost made in the 1980s and has spent four decades being grateful it didn’t.
Twenty-Five Companies, One Message
The post shared an open letter signed by 25 companies, including Nvidia, Microsoft, Meta, and Palantir, arguing that open-weight AI models are essential infrastructure for American technological leadership. “Open models strengthen safety and cybersecurity, accelerate innovation and diffusion, and enable sovereignty,” Huang wrote in what may be the most consequential debut post in the platform’s history.
The letter’s core argument is straightforward: the open-source software movement that powered everything from Linux to Android to the modern internet almost didn’t happen because policymakers in the 1980s considered restricting source code distribution. The companies are drawing a direct line between that near-miss and the current debate over whether Washington should restrict the distribution of open-weight AI models, particularly Chinese ones like Moonshot AI’s Kimi K3, which launched on July 16 and already ranks among the most capable models anywhere.
The Conspicuous Absences
What makes the letter as interesting for who didn’t sign it as for who did: OpenAI and Anthropic are nowhere on the list. Both companies have instead been warning Washington that powerful Chinese open-weight models represent a national security risk, a position that conveniently aligns with their business model of keeping their own models proprietary and behind API paywalls.
This is one of those moments in tech where the AI industry’s most significant actors have officially split into two camps with irreconcilable positions. The open-weight coalition, led by companies whose businesses benefit from AI being widely distributed, is arguing that restriction kills innovation. The closed-model camp, led by companies whose businesses depend on scarcity, is arguing that openness creates risk. Both camps are right about the other’s incentives and wrong about their own purity of motive, which is exactly why this is a genuinely hard policy problem rather than a simple good-versus-evil narrative.
Why Huang Chose This Moment
The timing here is not accidental. The White House has been accelerating its AI policy framework, and congressional hearings on AI regulation have shifted from “should we regulate?” to “how aggressively?” in the past six months. The Kimi K3 release lit a particular fire under Washington because it demonstrated that restricting American open-weight models doesn’t actually prevent other countries from building capable ones. It just ensures American developers don’t benefit from the ecosystem.
Huang’s argument carries weight precisely because Nvidia profits regardless of whether AI models are open or closed. The company sells the GPUs that train all of them. His advocacy for openness is harder to dismiss as pure self-interest than Meta’s, which needs open models to compete with Google and OpenAI, or Microsoft’s, which hedges its bets across both open and closed ecosystems.
The 1980s Parallel Is More Apt Than It Looks
The letter’s historical comparison deserves more credit than the usual “this is just like that other thing” tech-lobby framing. In the early 1980s, AT&T and IBM genuinely tried to restrict the distribution of Unix source code, and the U.S. government considered export controls on software that could have functionally killed open-source development. The reason those restrictions didn’t happen had less to do with enlightened policymaking and more to do with the technology moving faster than regulators could act.
The same dynamic is playing out with AI. By the time Washington settles on a restriction framework, the models it’s trying to restrict will already be two generations behind the state of the art and freely available from researchers in a dozen countries. The letter is essentially asking Congress to learn from a near-miss rather than repeat the mistake at scale.
What Comes Next
The letter doesn’t exist in a vacuum. Congress is actively drafting AI legislation, and the question of how to handle open-weight models is one of the genuinely unresolved fault lines. Export controls on chips have already reshaped the competitive landscape. Export controls on model weights would reshape it further, but in directions that even the letter’s critics acknowledge might backfire.
What Huang didn’t say, but the subtext makes clear: the real risk isn’t that open-weight models enable bad actors. It’s that restricting them hands the open AI ecosystem to China, which has no intention of restricting its own. The Kimi K3 release already proved the point. The model is capable, it’s open, and no amount of American restriction prevented it from existing.
Whether Washington reads the letter as a warning or a lobbying document will say a lot about where American AI policy lands by the end of the year. But the fact that Jensen Huang, a man who has never once felt the need to post on X, chose this issue for his debut tells you exactly how high the stakes feel inside the companies building the infrastructure that everything else runs on.