Open-weight AI firms become tech's hottest acquisition targets
Nvidia, Stripe and other tech giants have struck or are reportedly pursuing multibillion-dollar deals for open-weight AI model platforms as companies seek alternatives to costly frontier-lab models. The wave includes a reported $13 billion Nvidia bid for Hugging Face.
A flurry of major acquisitions is reshaping the AI industry this week, as large tech companies rush to buy platforms that give access to "open-weight" AI models — systems whose underlying code and parameters are publicly available rather than controlled by top AI labs.
According to reports, Nvidia is preparing to confirm a roughly $13 billion acquisition of Hugging Face, a widely used platform for sharing open-weight models and benchmarking tools, often described as a kind of GitHub for the AI era. The deal follows Nvidia's earlier $6 billion agreement with open-weight model developer Poolside, under which most of that company's staff will move to Nvidia. Two weeks ago, payments company Stripe acquired OpenRouter, a leading provider of open-weight models to businesses, for more than $7 billion.
Why the rush
Nvidia's strategy appears aimed at reducing its dependence on deals with major cloud providers and frontier AI labs, some of which — including OpenAI and Google — are now developing their own AI inference chips. By gaining control of the largest U.S. hub for open-model development, Nvidia could steer a large user base toward its own chips and standards. The company already produces its own open-weight Nemotron models, though adoption has been limited.
Rising costs of AI inference are also pushing some companies to explore cheaper alternatives, including models from Chinese developers such as Moonshot, DeepSeek and Alibaba. Still, adoption of open-weight models remains modest overall — surveys cited put usage at roughly 2% to 6% among companies and software engineers, concentrated mainly in high-volume, repetitive tasks like customer service chatbots.
For more complex work such as coding and agentic tasks, proprietary frontier models still tend to dominate, partly because they are easier to access and sometimes come with subsidized pricing. Industry observers note that companies currently choose open models mainly for control and configurability rather than cost savings, though that calculus could shift if frontier-lab prices keep rising.
