Nvidia Spends $7B, Supercharges Nemotron

Nvidia is turning open models into a demand engine for the hardware and infrastructure underneath them—not a charitable giveaway.

• –Model Factory automates foundation-model training, evaluation, reinforcement learning, and data experimentation across GPU clusters. [Poolside](https://poolside.ai/blog/introducing-the-model-factory)
• –The deal gives Nvidia model-building know-how and experienced researchers while Poolside remains legally independent.
• –Free Nemotron weights, datasets, and training techniques can pull developers toward NVIDIA’s ecosystem, echoing CUDA’s platform strategy. [NVIDIA](https://www.nvidia.com/en-us/ai-data-science/foundation-models/nemotron/)
• –Token providers selling merely “good enough” models face a brutal new competitor with hardware margins to subsidize zero-price software.
• –Poolside’s lost 40,000-chip cluster shows that frontier AI is increasingly gated by capital and physical compute, not just research talent. [The Next Web](https://thenextweb.com/news/nvidia-poolside-6bn-model-factory-licence)