NVIDIA Invests $7 Billion in Poolside to Build Open-Weight Nemotron Models

NVIDIA is shifting its AI strategy by investing $7 billion in startup Poolside to build open-weight Nemotron models, a move that signals the chipmaker is preparing for a potential slowdown in external GPU demand. This pivot allows NVIDIA to retain hardware revenue by redirecting chips to internal workloads if customer adoption stalls. Buyers monitoring data center trends should note that this internal hedging could impact the availability of high-end GPUs for third-party cloud providers.

NVIDIA invests $7 billion in Poolside to build open-weight Nemotron models

The investment splits into $6 billion for technology licensing and a $1 billion equity stake in Poolside. NVIDIA plans to hire most of Poolside's engineering team to accelerate the development of Nemotron-class models. This acquisition targets the software layer that runs on NVIDIA's Grace Blackwell and Vera Rubin architectures, aiming to create a viable ecosystem for open-weight AI.

Nemotron 3.5 Lightning represents the technical core of this push, quadrupling token output speed compared to previous iterations. Despite this speed increase, the model only accelerates actual agentic tasks by 30 percent, indicating diminishing returns on complex reasoning workloads. This performance gap suggests that raw generation speed does not linearly translate to utility in autonomous agent workflows.

Sam Altman recently admitted that the economic integration of AI has been slower than he predicted when GPT-4 launched in 2023. His comment highlights a broader industry realization that software disruption may not arrive as quickly as early benchmarks suggested. Meanwhile, NVIDIA is raising prices for Grace Blackwell and Vera Rubin systems by approximately 15 to 17 percent starting early next year. This price hike occurs even as the company bets heavily on internal model development to sustain hardware demand.

NVIDIA's dual approach of raising hardware prices while subsidizing open-weight model development reflects a strategy to lock in long-term ecosystem value. The company is effectively buying its own future demand by making its software stack more attractive to independent developers. This move stabilizes NVIDIA's position in the AI market regardless of how quickly enterprise customers adopt new GPU clusters.