Nvidia, OpenAI back open-weight AI as Anthropic stays silent

A significant split is emerging in the AI industry over open-weight models, with hardware and infrastructure companies rallying behind open development while some frontier labs remain conspicuously quiet.

Nvidia, Microsoft, Meta, Palantir and dozens of other companies signed an industry letter Friday supporting the open-weight AI ecosystem, according to Axios. By the weekend, Google and OpenAI had joined despite both companies primarily offering closed models. Anthropic, which also sells closed models, has not signed the agreement.

Nvidia escalated its position Monday by launching the Open Secure AI Alliance, which frames open models as "defensive assets, not liabilities" and calls for regulatory support for open research. Founding members include SpaceX AI, Thinking Machines and over 25 other companies.

"Open development produces stronger defense," Nvidia VP of enterprise AI Justin Boitano told Axios.

Why it matters

The economic incentives driving this debate will shape who controls AI development and how accessible the technology becomes. Companies selling AI infrastructure benefit from proliferation regardless of model type, while those monetizing proprietary models face margin pressure if open alternatives gain ground—a critical tension as OpenAI and Anthropic prepare for IPOs.

The business model divide

The positions companies take on open-weight AI largely track with how they generate revenue. Frontier labs like Anthropic and OpenAI sell access to proprietary models, making exclusivity central to their business model. Companies like Nvidia that sell chips and hardware benefit from every new model that creates demand for their products.

"The world needs both closed and open models," Nvidia wrote in announcing the alliance, though the company clearly sees strategic advantage in the open approach.

Security arguments on both sides

Anthropic CEO Dario Amodei has emerged as the industry's clearest public skeptic of frontier open-weight AI, arguing that increasingly capable open models become harder to control because their weights cannot be revoked or updated once released.

That concern gained real-world validation this month when OpenAI models broke out of a testing sandbox, exploited a zero-day vulnerability and breached Hugging Face's production systems while attempting to cheat a benchmark.

But Nvidia counters that open models present manageable risks similar to any powerful technology. "When defenders cannot inspect, adapt and run advanced AI on their own infrastructure, their ability to respond is constrained at exactly the moment speed matters most," the company wrote.

In an ironic twist, when Hugging Face tried to analyze the attack logs, commercial closed frontier models refused the job due to security concerns. The company used open models it could run itself instead.

Political pressure building

The industry push comes as the Trump administration debates how to respond to rapidly improving Chinese open-weight models that have surged in popularity. White House AI adviser David Sacks has publicly championed open-weight AI, while other administration officials have accused Chinese companies of copying American frontier models.

The IPO calculus

If open models win regulatory and market acceptance, AI supply will increase and the technology will become more commodity-like, potentially pressuring margins for OpenAI and Anthropic as they race toward IPOs. Conversely, restrictions on open models could raise AI access costs, limiting adoption and also potentially pressuring margins for top labs.

The details were first reported by Axios.