The US Military's Top Graduate School Just Switched On One of the World's Most Powerful AI Supercomputers
Jensen Huang flew to Monterey to personally commission an AI supercomputer at the Naval Postgraduate School. Here is what it does, and why it reaches well beyond the military.
Key points
- NVIDIA CEO Jensen Huang commissioned the DGX GB300 AI supercomputer at the Naval Postgraduate School (NPS) in Monterey, California, in May 2025.
- More than 1,500 resident students and 600 faculty at NPS now have on-site access to the machine for research covering weather prediction, cybersecurity and disaster planning.
- Admiral Samuel Paparo, commander of US Pacific Command, called AI literacy essential for future military leaders.
- Infrastructure partners DDN, VAST Data and Vertiv built the storage, data management and power systems that keep the supercomputer running.
- The school already uses AI to build digital twins, simulated copies of real-world environments used to practise navigation and decision-making without real-world risk.
Picture a library that can read every book it owns, cross-reference them all in seconds, and write a custom report just for you. That is roughly what an AI supercomputer does for researchers. The books are enormous data sets; the report might be a typhoon simulation or a cyberattack scenario.
That machine is now live at the Naval Postgraduate School, the US military's flagship graduate university.
What exactly got switched on?
The system is the DGX GB300, a purpose-built AI supercomputer from NVIDIA, the chip company best known for making graphics cards. NVIDIA announced the commissioning during NPS's three-day Converge @ NPS event, and Huang was there in person to mark it. We first covered the DGX GB300 on 26 July 2026.
The machine lets NPS train foundation models, meaning large AI systems built from scratch on the school's own data rather than borrowed from a commercial provider. It can also run high-fidelity simulations, detailed virtual recreations of real conditions.
Researchers are already modelling sea conditions and predicting atmospheric changes. The school has partnered with MITRE, a US nonprofit research organisation, to build digital twin environments, detailed virtual copies of physical places used to rehearse navigation decisions before anyone acts in the real world.
Why does this matter outside the military?
Much of what NPS produces eventually reaches civilian life. Disaster resilience tools developed here inform how communities prepare for floods and hurricanes. Cybersecurity frameworks can protect infrastructure well beyond military networks.
Huang made the broader point at the event. "Leadership is about being in service of something else," he told attendees. "All of this technology is just tools to help you." He also said that non-computer-science experts shouldn't be put off: "The most important advice that I would give to someone is to engage the technology. It's not as hard as you think it is."
Paparo framed it in operational terms. "Information will move at lightning speeds, compressing response times," he said. "Our advantage will come from leaders who can use the technology to see, understand, decide and act faster while exercising the judgment that machines cannot provide."
What else is the school building?
NPS runs regular hackathons, intensive problem-solving competitions, that have produced advances in ocean research and autonomous systems. The DGX GB300 now powers those workloads on campus.
NVIDIA has also expanded an instructor programme through its Deep Learning Institute, putting training toolkits in the hands of NPS faculty so AI runs through graduate courses from space operations to ocean science.
The hardware partners matter here. DDN built high-performance data storage. VAST Data supplied a unified platform for secure data access across computing environments. Vertiv handled the physical racks, cooling and power infrastructure, plus installation, testing and fluid monitoring.
For anyone outside the military, the practical point is this: the tools that serious researchers get first tend to become the civilian tools of the next decade. Putting that computing power in the hands of people trained to think hard about consequences is a reasonable place to begin.
