Sources: Nvidia plans to use its $6B deal with Poolside to build an open-weight AI model to compete with Chinese models like DeepSeek and Kimi
Context & Ripple Effects
This closes an arc that started with Nvidia weighing a $500M–$1B bet on Poolside last fall and hardened into last week's non-exclusive $6B licensing deal plus a $1B check at a $12B pre-money valuation. What changed today is the stated purpose: the license isn't defensive IP hoarding — Nvidia intends to fold Poolside's technology into an open-weight model aimed squarely at DeepSeek and Kimi.
First-order effects
- Poolside gets $6B in licensing revenue and a $12B valuation mark overnight, while 109 of its staffers receive Nvidia job offers — a talent-and-technology transfer disguised as a commercial deal.
- DeepSeek and Kimi gain a well-funded Western open-weight rival whose distribution can be bundled with the GPUs most inference buyers already run.
Second-order effects
- The move echoes Nvidia's own framing from when it called DeepSeek "an excellent AI advancement" while stressing that inference still requires significant numbers of Nvidia GPUs — open weights are a way to keep that inference demand flowing through Nvidia silicon rather than ceding it to cheaper Chinese stacks.
- Because the license is non-exclusive, Poolside remains free to strike similar deals with other chipmakers, so rivals now have a template for buying model capability without acquiring a lab outright.
Third-order effects
- If this fits the pattern of March's Nemotron 3 Super launch and Nvidia's disclosed $26B five-year open-model budget, Nvidia is no longer just selling the shovel layer — it is vertically integrating into the model layer itself, with open weights as both product and demand engine for its hardware.
The trend: Nvidia is responding to efficient Chinese open-weight models by buying model capability outright and giving weights away, turning openness into a moat for its GPU business.