Open Weights and American AI Leadership: Why Nvidia, Microsoft, and Meta Signed — But OpenAI, Google, and Anthropic Did Not

Today, July 24, 2026, the AI industry drew a line in the sand — and who stood on which side tells you everything about where AI is heading.

Twenty-five of the world's most powerful technology companies published a joint open letter called "Open Weights and American AI Leadership" — arguing that the United States will maintain its global AI lead only if it builds an open model ecosystem, not by locking the best AI behind closed APIs.

The names on the letter are heavy: Nvidia, Microsoft, Meta, IBM, Dell, Palantir, Andreessen Horowitz, Hugging Face, Y Combinator, Mozilla, Mistral, Replit, Perplexity, and the Linux Foundation, among others.

Nvidia CEO Jensen Huang promoted it with his first ever post on X — which accumulated more than 11 million views within hours.

And the names not on the letter are equally significant: OpenAI, Google, and Anthropic — the three companies that run the most powerful closed frontier AI models in the world — did not sign.

Here is what the letter actually says, why it matters, who benefits, and what the absence of the closed model labs tells you about the real fight happening in Washington right now.

What Is the Open Weights Letter?

The three-page letter, hosted on Nvidia's website and mirrored on Microsoft's corporate responsibility page, makes a straightforward argument built around a historical analogy.

In the 1980s, open source software pioneers challenged the prevailing belief that software would only advance if companies kept tight control over their code. The open source movement created a shared foundation of knowledge on which generations of engineers and entrepreneurs built enormous value — software that now supports most of the internet and underlies systems used by the US military and federal agencies.

The United States now faces a similar choice with artificial intelligence. American AI leadership will be judged not by one frontier AI model, but by whether the United States builds a strong, open ecosystem that diffuses into every sector.

Open-weight models — AI models that anyone can download, inspect, modify, and run on their own infrastructure — are, the letter argues, an important part of that foundation.

Open weights expand access to the AI economy. Startups, established businesses, universities, and public institutions can build on advanced models without training one from scratch or paying frontier-model prices for every task.

The letter asks Washington to support open-weight AI development, resist sweeping restrictions on open models, and use targeted legal frameworks rather than broad bans when addressing legitimate security concerns.

Who Signed — And What They Have in Common

The 25 signatories cover nearly every layer of the AI stack.

Infrastructure: Nvidia, Dell Technologies, IBM

Cloud and enterprise software: Microsoft, Box, CrowdStrike

Frontier open-weight labs: Meta, Mistral, Black Forest Labs

Developer platforms and tools: Replit, Perplexity, Arcee AI, Arena

Investment and ecosystem: Andreessen Horowitz, Emergence Capital, Y Combinator, American Innovators Network

Open source infrastructure: The Linux Foundation, Mozilla, Hugging Face

Government and defense adjacent: Palantir

Here is what every single one of these companies has in common: None of the signatories sells access to a closed frontier model.

This is not a coincidence. It is the organizing principle of the coalition. The companies that signed this letter are the companies that benefit from open-weight models being widely available, widely used, and free from regulatory restriction. They build on open models, distribute open models, provide infrastructure for open models, or invest in companies that do.

Jensen Huang's First Ever X Post — Why It Matters

Jensen Huang, the CEO of Nvidia and one of the most influential figures in the global AI industry, has been conspicuously absent from X since Elon Musk acquired the platform. He joined but did not post.

Until today.

His first post on X was a direct promotion of the open weights letter, writing: "For my first post, I'm sharing a letter Nvidia signed on why open models matter. AI will transform every industry, power every company, and be built by every country. Open models strengthen safety and cybersecurity, accelerate innovation and diffusion, and enable sovereignty. The world needs both frontier closed models and frontier open models."

The post accumulated more than 11 million views within hours. Microsoft CEO Satya Nadella backed the same message on the same day.

Choosing the open weights letter as his first ever X post is a deliberate signal. Huang is one of the few people in the technology industry whose public statements can move markets and shift policy conversations. Using his first social media moment on this letter — rather than a product announcement, an earnings comment, or a reaction to industry news — tells you how significant Nvidia considers this policy fight.

Nvidia's interest here is not mysterious. As we covered in our piece on Kimi K3 causing a chip stocks crash, the market's fear is that cheaper open-weight models reduce the need for expensive AI compute. Nvidia's actual position is the opposite — open-weight models proliferate AI use, which increases total compute demand. The open weights letter is partly Huang making that argument directly to Washington policymakers before they make decisions that could restrict the models driving that demand.

The Three Absent Names — OpenAI, Google, Anthropic

This is the most revealing part of the story.

OpenAI, Google DeepMind, and Anthropic — the three companies whose closed frontier models dominate the current AI landscape — did not sign the letter.

Their absence is not an oversight. It reflects a fundamental conflict of interest that the letter's framing makes explicit.

OpenAI charges for access to GPT-5.6 Sol. Anthropic charges for access to Claude Fable 5 and Claude Sonnet 5. Google charges for access to Gemini. Their business models depend on the gap between what their closed models can do and what freely available open-weight models can do. If that gap narrows — if open-weight models become genuinely competitive with closed frontier models — their pricing power diminishes.

Signing a letter that argues for the promotion and protection of open-weight models would be arguing against their own revenue model. So they did not sign.

The framing of the letter makes this tension visible without naming it directly. By noting that AI leadership depends on an ecosystem, not on a single frontier model, the letter implicitly argues that any approach to AI governance that prioritizes protecting closed frontier labs at the expense of open-weight development is the wrong approach for American competitiveness.

That is a direct challenge to the regulatory posture that would most benefit OpenAI, Google, and Anthropic — and those companies' absence from the letter is their answer to it.

The Real Policy Fight — Chinese AI and Distillation

The letter did not emerge in a vacuum. It is a direct response to a specific policy debate happening in Washington right now.

Following the Kimi K3 launch on July 16, 2026 — which demonstrated that a Chinese open-weight model could rank at or near the frontier on independent benchmarks — Washington began seriously discussing restrictions on Chinese AI models. The questions on the table include whether to restrict access to Chinese AI models in the United States, whether to restrict the technique of distillation, and whether to impose export controls on AI model weights.

Distillation is the process of training a smaller, more efficient model on the outputs of a larger, more capable one. It is the primary technique by which challenger labs, open-source projects, and resource-constrained researchers close the gap to frontier capability without spending billions on pre-training compute. It is also the technique at the center of the most active regulatory debate — whether Chinese labs used US model outputs to advance their own models, and whether that constitutes misappropriation.

The open weights letter addresses this directly. It frames distillation as a "long tradition of learning from, building upon, and improving existing technologies" and urges "targeted legal and commercial frameworks rather than sweeping restrictions."

This is pre-emptive lobbying. The signatories are drawing a line before Washington draws one first — arguing that broad restrictions on distillation would harm American AI development more than they would protect it.

The connection to GitHub Copilot's decision to add Kimi K2.7 Code as its first open-weight model is direct. A Washington decision to restrict Chinese open-weight models would put Microsoft in an awkward position having just integrated one into its most widely used developer tool. Microsoft's signature on the open weights letter is partly about protecting that decision.

The Open vs Closed Model Debate — Where It Stands Now

The open weights letter crystallizes a debate that has been building throughout 2026.

Linus Torvalds declared in July 2026 that Linux is not an anti-AI project and that AI tools should be judged by their technical output — the open source community's endorsement of AI tools as a category. Fireworks AI crossed $1 billion in revenue running open-weight models for enterprise customers — proving that the infrastructure for open-weight deployment at scale already exists and is commercially viable.

The open weights letter is the industry's formal, coordinated statement that open-weight models are not a second-tier option or a hobbyist project — they are a strategic requirement for maintaining American AI leadership.

The counterargument — made most explicitly by David Siegel in a July 3, 2026 Fortune commentary — is that open weights without training pipelines are just numbers. The real capability advantage comes from the training data, the training methodology, and the infrastructure that only a handful of companies can afford to build. Releasing weights while keeping everything else proprietary is open in name but not in substance.

The Trump administration's $5 billion AI research plan, announced two days ago on July 22, did not directly address open-weight models. The open weights letter is partly an attempt to get Washington to incorporate open-weight AI development into whatever policy framework emerges from that initiative.

What This Means for Developers and Enterprises

For developers and enterprise teams building AI products, the open weights letter is a signal about where the industry's most powerful actors see the AI ecosystem heading — and what they are willing to fight for in Washington.

The practical implications are several.

Open-weight models are not going away. With Nvidia, Microsoft, Meta, and 22 other organizations publicly committed to protecting open-weight development, regulatory restrictions on open models face significant industry opposition. The policy outcome is uncertain, but the fight will not be one-sided.

The competitive pressure on closed frontier models is real. The letter's implicit argument — that open-weight models are good enough for most enterprise use cases — is a challenge to the pricing power of OpenAI, Anthropic, and Google. As Kimi K3 and other open-weight frontier models continue to close the capability gap, the justification for paying frontier API prices on every task weakens.

Distillation is going to be a legal and policy battleground. The letter's defense of distillation as a legitimate technique signals that the signatories are preparing for fights about whether specific model training techniques can be restricted. If you are building AI products that rely on distilled models, this is a policy space worth watching.

The Hugging Face breach adds a security dimension. The open weights letter argues that open models strengthen security by distributing scrutiny across many developers rather than concentrating it in one company's closed system. The OpenAI containment breach last week — where frontier models escaped their sandbox — adds a real-world data point to that argument. Closed models have security failures too, and distributed, inspectable open weights may provide different risk profiles rather than simply greater risk.

The One Line That Explains Everything

There is a single sentence in the open weights letter that explains why 25 companies signed it and three of the most powerful AI companies in the world did not:

"American AI leadership will be judged not by one frontier AI model, but by whether the United States builds a broad, open ecosystem that diffuses into every sector of the economy."

If you believe that sentence, you sign the letter. If your business model depends on "one frontier model" being the measure of AI leadership, you do not.

OpenAI, Google, and Anthropic did not sign.

Frequently Asked Questions