Key facts
- Nvidia is reportedly close to acquiring Hugging Face for $13 billion.
- Nvidia also acquired Poolside for $6 billion, integrating its employees.
- Stripe purchased OpenRouter, a provider of open-weight models, for over $7 billion.
- Open-weight AI models are being adopted for cost-efficiency in repetitive inference tasks.
- Control and configurability are primary drivers for companies using open-weight models.
The artificial intelligence sector is witnessing a surge in acquisitions targeting companies focused on open-weight models, signaling a shift in investment strategies for major tech players. Nvidia is reportedly in advanced talks to acquire Hugging Face, a prominent platform for sharing open-weight AI models and benchmarks, for approximately $13 billion. This potential deal follows Nvidia's recent $6 billion agreement to integrate most of Poolside's employees, an open-weight model builder, into its operations.
Further underscoring this trend, Stripe recently acquired OpenRouter, a leading provider of open-weight models to businesses, for over $7 billion. These substantial investments highlight the growing importance of accessible AI development ecosystems, often compared to GitHub for the AI era.
For Nvidia, these moves appear to be a strategy to diversify its dependencies beyond major hyperscalers and frontier AI labs, especially as companies like OpenAI and Google develop their own inference chips. By acquiring Hugging Face, Nvidia could gain access to a large user base and drive adoption of its own chips and standards, complementing its existing Nemotron open-weight models.
The increasing cost of AI inference is also driving exploration of more economical models, including those developed by Chinese companies. While adoption of open-weight models is still relatively small, with surveys indicating usage by 6% of companies and 2% of software engineers, their appeal is growing, particularly for businesses with high-volume, repetitive inference workloads like customer service chatbots. These models can be tuned for efficiency, as highlighted by Stripe's framing of the OpenRouter acquisition around optimizing compute resources.
However, for complex coding and agentic tasks requiring varied reasoning, proprietary frontier models often remain preferred due to easier access and potential token subsidies. Experts suggest that as AI workflows mature, the cost-effectiveness and control offered by open-weight models will become more compelling. Lin Qiao, CEO of Fireworks, a major open-weight model host, emphasizes the future lies in specialized intelligence, with companies potentially developing unique models for specific use cases.