- The Naval Postgraduate School commissioned the first Nvidia DGX GB300 deployed within the U.S. military on July 22, 2026. [1] - The three-rack system has 72 Blackwell Ultra GPUs, 36 Grace CPUs, 20 terabytes of high-bandwidth memory and peak compute power consumption of about 135 kilowatts. [2] - Nvidia donated the computer, but neither Nvidia nor NPS disclosed its value or the amount specifically allocated to projects using it. [1][4]
The Naval Postgraduate School has activated an Nvidia DGX GB300 at its Monterey, California, campus, giving the military graduate university an on-premises system for large-scale AI training, inference and simulation. NPS formally commissioned the machine on July 22 and called it the first DGX GB300 deployed within the U.S. military. [1]
The liquid-cooled system combines 72 Nvidia Blackwell Ultra GPUs and 36 Grace CPUs across 18 compute trays. It carries 20 terabytes of high-bandwidth GPU memory, 37 terabytes of total fast memory and up to 576 terabytes per second of memory bandwidth. NPS says its compute components draw roughly 135 kilowatts at peak use and operate within the campus’s existing power and water capacity. [2]
Approved access for defense-focused programs #
Nvidia donated the system to the nonprofit NPS Foundation, which installed it with support from Vertiv, DDN and VAST Data. The announcements did not disclose the machine’s value or quantify the accompanying research funding. The foundation separately says it has committed up to $2 million to broader NPS AI initiatives, including personnel, infrastructure, hardware, software and laboratory resources. [1][4]
NPS will manage the GB300 as a shared research capability. Access extends to approved collaborations involving its faculty and students, government organizations, operational commands, laboratories and appropriate industry partners. Nvidia said the platform is intended to support more than 1,500 resident students and 600 faculty, but that does not indicate unrestricted individual access. [2][3]
Planned work includes foundation-model training, adapting language models for naval planning and intelligence, multi-agent training, cybersecurity, weather and ocean modeling, digital twins, synthetic data generation and adversarial-resilience testing. NPS describes the platform as available for “secure experimentation,” but has not publicly detailed classification approvals, network segregation or the types of operational data permitted on the system. Initial demand will be managed as utilization increases. [1][2]
Companies mentioned #
Further sources #
[1] Naval Postgraduate School, “NPS Launches First NVIDIA DGX GB300 AI Supercompute… ↗
[[2] Naval Postgraduate School AI Task Force, “NVIDIA DGX GB300,” system specificati… ↗](https://nps.edu/web/ai/gb300)
[[3] Nvidia, “AI Supercomputer Comes Online at Naval Postgraduate School,” July 22, … ↗](https://blogs.nvidia.com/blog/naval-postgraduate-school-dgx-ai-supercomputer/)
[[4] Naval Postgraduate School Foundation, “Accelerating AI at NPS,” funding and pro… ↗](https://www.npsfoundation.org/support/accelerating-ai)
The stories that matter, in one email. Free — unsubscribe anytime.