Everyone's eyes are on Nvidia's potential $13 billion acquisition of Hugging Face, a platform for sharing open-source AI models and benchmarks. Rumours suggest this deal follows Nvidia's $6 billion deal with Poolside, an open-source model builder, and Stripe's acquisition of OpenRouter for over $7 billion. These deals reflect a sector where giving stuff away is making big money.

Nvidia aims to reduce its reliance on major hyperscalers and frontier labs. With companies like OpenAI and Google developing their own inference chips, Nvidia wants to capture a bigger slice of the model-making business by acquiring Hugging Face, the largest U.S. developer space for open models.

Open-source models are mainly used by companies with high-volume tasks like customer service chats, where they can be tuned to perform tasks efficiently. However, for coding and agentic tasks, proprietary models still dominate due to easier access and token subsidies. As AI workflows mature, more companies may turn to open-source models for control and configurability, but the price is a key factor.

Fireworks, a leading open-source model router, processes over 40 trillion tokens a day, more than Gemini's or OpenAI's APIs. CEO Lin Qiao believes every app company should consider hiring an in-house researcher to build their own model, as the future is specialized intelligence.