WSJ reports Nvidia plans a $6 billion investment aimed at competing with Chinese AI models DeepSeek and Kimi K3.

Nvidia has signed a $6 billion agreement with artificial intelligence startup Poolside to develop a powerful open-weight model capable of competing with Chinese large language models such as DeepSeek and Kimi K3, according to a Wall Street Journal report citing its own sources.

On August 21, Bloomberg reported that the chipmaker agreed to pay Poolside $6 billion for AI model licenses, offer employment to more than 100 of the startup’s personnel, and invest an additional $1 billion. According to the report, Nvidia has recently shifted its strategic focus toward supporting open-source models that developers can freely use and customize. This approach aims to broaden technological accessibility and prevent market concentration among a handful of dominant players. Nvidia also ties these initiatives to anticipating and sustaining future demand for its chips and AI infrastructure.

The structure of Nvidia’s agreement with Poolside mirrors its recent acquisition of chip developer Groq. Reports indicate that Nvidia paid $20 billion for Groq’s licenses and retained most of its founders and employees.

Per the WSJ, Nvidia’s arrangement with Poolside will directly challenge leading U.S. AI firms, including OpenAI and Anthropic. Open-weight models are generally significantly cheaper to operate and permit straightforward customization, granting them a distinct competitive edge in the current landscape.

In July, reports emerged that the administration of U.S. President Donald Trump resumed discussions on measures that could restrict the deployment of Chinese open-weight models. Twenty-five American technology companies, including Nvidia, Meta, and Microsoft, urged federal authorities against imposing such restrictions. In an open letter, the companies warned that bans would undermine U.S. technological leadership and instead consolidate market power among a few closed-source developers.

The recent release of Kimi K3, whose creators the White House accused of distilling Anthropic’s Fable model, has further intensified debates over potential regulatory restrictions. Earlier, ForkLog examined the rapid rise of China’s Moonshot AI in a dedicated industry analysis.