Nvidia has struck a deal with AI startup Poolside for $6 billion to develop a powerful model with open weights capable of competing with Chinese LLMs like DeepSeek and Kimi K3. This was reported by WSJ citing its own sources.
On August 21, Bloomberg reported that the chipmaker agreed to pay Poolside $6 billion for licenses to AI models, offer jobs to over 100 startup employees, and additionally invest $1 billion. Recently, Nvidia has focused its efforts on supporting open-source models that can be freely used and customized by developers. The company aims to make technology more accessible and avoid concentration in the hands of a few major players, linking these efforts to supporting future demand for its chips and AI systems.
The deal with Poolside is structured similarly to Nvidia's recent acquisition of chip developer Groq, for which the company paid $20 billion for licenses, hiring most employees and founders. According to WSJ, the agreement with Poolside will also pose a direct challenge to leading American AI companies, including OpenAI and Anthropic, as models with open weights are much cheaper to operate and allow for easy customization.
In July, it became known that discussions in the U.S. administration resumed regarding measures that could complicate the use of Chinese open-weight models. 25 American companies, including Nvidia, Meta, and Microsoft, urged authorities not to impose restrictions. In an open letter, they warned that bans would not strengthen the technological leadership of the U.S., but rather shift the market towards a few closed developers.