On July 24, 2026, the AI industry fractured along a clear, structural fault line. A coalition of 25-plus companies, including heavyweights like Nvidia, Microsoft, Meta, and Palantir, alongside venture capital pillars Andreessen Horowitz and Y Combinator, signed a letter titled Open Weights and American AI Leadership. The document is a defense of the open-weight ecosystem, arguing that national leadership depends on diffusion into every sector rather than the concentration of power in a single frontier model. Yet, the most revealing aspect of the letter is not who signed it, but who did not: OpenAI and Anthropic.
While their competitors aligned to protect the open-weight distribution channel, OpenAI and Anthropic remained silent. This is not merely a difference of opinion; it is a divergence in business survival strategies. The signatories are defending an ecosystem that allows organizations to match the right model to the right job at the right cost. The absentees, meanwhile, are protecting a moat that relies on keeping frontier-scale capability behind a proprietary wall.
The letter arrived on the same day the DeepSeek V4 model was retired, just three days before the scheduled release of the Kimi K3 open-weights, and only eight days before the August 1 deadline for the White House framework under EO 14409. The signatories are clearly positioning themselves ahead of this regulatory window. They explicitly distinguish between legitimate distillation and unlawful extraction, a direct response to recent allegations from figures like Michael Kratsios regarding IP theft by Chinese labs. The letter argues that restricting open weights would only drive innovation overseas, effectively ceding the ground the US is trying to secure.
DeepSeek V4 Flash is priced at $0.14/$0.28 per million tokens, compared to $10/$50 for Claude Fable 5—a 57x difference in output cost. OpenRouter data confirms this shift in market behavior: Chinese-origin models have surged from under 2% of token consumption in late 2024 to more than 50% by June 2026. The question is whether the market will continue to pay a premium for closed-frontier models when efficient, specialized alternatives are increasingly available.
Jensen Huang’s involvement underscores the stakes. He made his first-ever post on X to share the letter, stating, “For my first post, I’m sharing a letter @NVIDIA signed on why open models matter.” When the primary supplier of the compute that powers the entire industry takes such an unambiguous position, it signals that the current trajectory of closed-model dominance is not a foregone conclusion. Satya Nadella also shared the letter, further aligning Microsoft with the open-weight camp despite its heavy investment in OpenAI.
The absentees have their own incentives. Both OpenAI and Anthropic are reportedly gearing up for massive IPOs. Protecting closed-model revenue is a financial imperative for companies whose valuations are tied to the exclusivity of their technology. OpenAI’s head of strategic futures has previously argued that open-weight models slow AI capex, while Anthropic has publicly characterized AI distillation by Chinese companies as IP theft. These are not just policy stances; they are efforts to maintain pricing power that is currently being squeezed from two sides: from below by Chinese open-weight commoditization and from above by US export-control actions.
The letter also addresses the security dimension, arguing that in a world where attackers use advanced AI, defenders need access to models with comparable capabilities. This reframes the debate from one of “safety through restriction” to “security through parity.” If the government follows the path of restriction, it will fundamentally reshape compute demand and model pricing, likely accelerating the flight of innovation to jurisdictions where open-weight development remains unencumbered.
The next phase of the AI industry will be defined by the tension between those who profit from the scarcity of intelligence and those who profit from its ubiquity. The signatories are betting that the future of American AI leadership lies in the widespread adoption of efficient, specialized models. The absentees are betting that they can keep the frontier closed long enough to capture the market. The White House framework forces a binary choice on model distribution that will determine the future of the American AI stack.