Nvidia, Microsoft and Meta push US to back open-weight AI models

This letter marks an unusually broad alignment of interests across the AI stack, from chipmakers and cloud vendors to model labs and open‑source foundations. The signatories are effectively telling Washington that if it clamps down too hard on downloadable model weights, it risks locking AI innovation inside a few closed labs and hyperscalers, and ceding ground to fast‑moving Chinese players like Moonshot and DeepSeek. By rooting their argument in economic competitiveness, national security and cybersecurity rather than “openness for its own sake,” they are trying to reframe open‑weight AI as a strategic asset rather than a regulatory threat.([indianexpress.com](https://indianexpress.com/article/technology/artificial-intelligence/nvidia-microsoft-meta-push-for-open-weight-ai-10803122/))

For the race to AGI, this is really about who gets to experiment at the frontier. Open‑weight models make it cheaper for startups, universities and mid‑tier countries to run powerful systems on their own infrastructure, which spreads cutting‑edge capabilities well beyond a handful of US labs. That could dramatically expand the pool of researchers and engineers pushing on alignment, multimodal reasoning and agentic systems, but it also multiplies the number of actors who can fine‑tune models for offensive cyber or information ops. The biggest competitive implication is geopolitical: if US policy tilts toward bans while China keeps shipping strong open‑weights, we may see a bifurcated ecosystem where leading open experimentation happens outside US jurisdiction.