Open-weight AI companies are the Valley’s hottest acquisition targets

Details

The open-weight AI ecosystem is undergoing rapid consolidation as deep-pocketed incumbents race to own the infrastructure layer. Nvidia is in talks to acquire Hugging Face — the dominant platform for sharing, hosting, and benchmarking open-weight models — for a reported $13 billion. This follows Nvidia's $6 billion acquisition of Poolside, an open-weight model builder whose staff will integrate into the chip maker. Separately, Stripe closed a $7 billion deal for OpenRouter, the leading API gateway for accessing open-weight models in production. Nvidia's strategic calculus is clear: as OpenAI and Google develop their own inference silicon (including OpenAI's Jalapeño chip), Nvidia's dominance of AI training may erode at inference time. Owning Hugging Face gives Nvidia direct access to the developer community and the ability to steer adoption toward its own chips and standards. Despite the deal volume, open-weight model adoption remains narrow. Just 6% of companies use them according to Ramp spending data, and only 2% of software engineers according to Jellyfish. The primary use case today is high-volume, repetitive workloads like customer service chatbots — where a fine-tuned open model beats paying frontier model rates per token. Coding and agentic tasks still favor proprietary models. The broader signal is competitive: hyperscalers building chips, frontier labs vertically integrating, and now hardware players buying into open-source platforms. The AI stack is being carved up fast.