Nvidia donates DGX GB300 AI supercomputer to US Navy school

This donation effectively puts a near-frontier AI factory inside the US military’s top graduate school, and that is strategically significant. A DGX GB300 with 72 Blackwell Ultra GPUs and 36 Grace CPUs is not just a research toy, it is serious training and inference infrastructure that can host large-scale models, multi agent simulations and high resolution forecasting workloads on premises. For the Pentagon, that means officers and researchers can prototype and test classified or sensitive AI concepts without relying on commercial clouds, while learning what it really takes to operate at that scale.

In the broader race to AGI, the move tightens the coupling between frontier commercial hardware and defense experimentation. Nvidia deepens its role as the default AI arms merchant, while US defense researchers gain hands on experience with the kind of agentic, simulation heavy systems that will shape next generation command and control, wargaming and decision support. It also shows an emerging model for public private collaboration: the system is donated through a foundation, with partners like Vertiv providing the power and cooling stack, which lets the military sidestep slower procurement cycles. That could accelerate how quickly new generations of AI compute reach operational testbeds.