Nvidia, Microsoft back open-source AI push as industry shifts

On July 24, Nvidia, Microsoft and more than 20 other companies and organizations urged U.S. lawmakers to back open-weight AI models, putting some of the industry’s biggest names behind systems that can be downloaded and adapted more easily than tightly controlled proprietary tools. The signers included Meta, IBM, Palantir, Hugging Face and Andreessen Horowitz, while OpenAI and Anthropic were not on the letter.

Open models can be copied, modified and integrated by developers, startups and enterprise customers with far less friction than software locked inside one company’s platform. For buyers, that can mean lower experimentation costs, faster product development and more flexibility to run models on private infrastructure. For Nvidia and Microsoft, the economics still work even when the model itself is more open, because the chips, storage and deployment services needed to run AI at scale remain expensive and profitable.

Microsoft published a July 24 post titled Open Weights and American AI Leadership. The post said open-source software helped power the internet and pointed to open-source pioneers in the 1980s challenging the idea that software could advance only if companies kept tight control over code.

The letter warned against premature restrictions on open-weight models. Broad limits could slow competition and innovation, while pushing developers toward Chinese AI systems that are often lower-cost and attractive to researchers and startups. Supporters of open models argue wider access improves transparency and trust. Critics warn that widely distributed models can be harder to secure and easier to misuse once they leave a single company’s control.