Nvidia is reportedly pursuing a $13 billion acquisition of Hugging Face, a platform for sharing open-weight AI models and benchmarks.

Hugging Face serves as a central hub for developers building and deploying large language models that are not owned by frontier AI labs, functioning similarly to a GitHub for the AI era.

The potential deal follows Nvidia's $6 billion agreement with Poolside, an open-weight model builder, and Stripe's recent acquisition of OpenRouter for more than $7 billion, indicating a significant influx of capital into the open-weight AI sector.

Nvidia aims to reduce dependence on deals with major hyperscalers and frontier labs as companies such as OpenAI and Google develop proprietary inference chips, like OpenAI's Jalapeño. By acquiring Hugging Face, Nvidia would gain access to a large developer base to promote its chips and standards, while open-weight models currently see limited adoption, used by roughly 6% of companies and 2% of software engineers according to surveys by Ramp and Jellyfish.

Analysts note that open-weight models are primarily employed in high-volume, repetitive tasks such as customer service chat, where cost efficiency is critical. Stripe's CEO Patrick Collison highlighted that tokens are a central currency for AI companies and that efficient use of compute resources will determine real-world economic potential.

Fireworks CEO Lin Qiao said her company processes 40 trillion tokens daily, more than the APIs of Gemini or OpenAI, and advocates for model diversity, suggesting that companies should develop specialized in-house models for specific use cases.

The trend reflects broader industry efforts to hedge against concentration in a few dominant AI labs and to explore open technology as an alternative, though widespread adoption of open-weight models remains limited.