Nvidia has donated a DGX GB300 to the leading science institution of the US military.
Jensen Huang visited Monterey this week to present the US Navy with a machine that most defense contractors would typically sell. The founder of Nvidia helped inaugurate a DGX GB300 AI supercomputer at the Naval Postgraduate School, marking the first installation of its kind within the American military, and the company chose to donate it instead of using a procurement contract.
This hardware consists of a single liquid-cooled rack containing 72 Blackwell Ultra GPUs and 36 Grace CPUs, interconnected as one memory domain and utilizing Nvidia’s Mission Control orchestration software. It represents the rack-scale configuration Nvidia offers for inference and reasoning tasks, and it is the top model in the DGX lineup. Currently, NPS operates what is the most powerful supercomputer the Pentagon possesses.
One estimate valued the system at about $15 million, although neither Nvidia nor the school has publicly confirmed any figures. The donation was made to the Naval Postgraduate School Foundation directly, bypassing the standard defense procurement process that involves competitive bidding. DDN, VAST Data, and Vertiv provided the storage, data infrastructure, and cooling systems for the rack.
NPS serves as the Navy’s graduate research institution, and the supercomputer is intended for educational and research purposes rather than active operations. Faculty and students aim to use it for weather forecasting, ocean modeling, cybersecurity, and disaster-response planning, as well as to create maritime digital twins through Nvidia’s Omniverse platform. The school also plans to develop its own foundational models, including an internal system it is reportedly dubbing NPS GPT for managing sensitive information.
The arrival of the rack coincides with the establishment of a dedicated AI research center on campus the previous year, ensuring the computational power has a clear purpose. The appeal lies partly in its resolution. Owen has noted the ability to forecast conditions for highly localized operations compared to the broader grids, approximately 50km across, that most existing models use.
For a Navy engaged in planning amphibious landings and small-boat movements in unpredictable weather, more accurate predictions are crucial. “The issue is not whether these technologies will be available," said Captain Michael Owen, the vice provost for warfare studies at the school. “The issue is whether future military leaders can assess their limitations, apply them ethically, and utilize them responsibly.”
This distinction is significant, as opposed to existing supercomputing centers at the Department of Defense that handle operational tasks, the NPS system focuses on research and innovation. The goal is to allow the school to refine models that the broader joint force might eventually rely on, rather than integrating the rack into everyday command operations.
Huang, who attended the commissioning event with Admiral Samuel Paparo, the head of US Pacific Command, characterized AI as a future “backbone of America’s defense.” This statement is not entirely neutral, considering that his company supplies much of this backbone.
The event concludes an arrangement that began discreetly in December 2024 when NPS and Nvidia signed a cooperative research and development agreement. The Pentagon's demand for Nvidia technology has notably increased over the past year, with the company spending a significant amount of that time selling similar systems to national governments.
The donation places the hardware within a military institution for the first time, representing a unique type of foothold. A school populated by active officers also serves as a training ground for those who will eventually establish the requirements.
Paparo addressed the audience, emphasizing that officers would increasingly operate in environments where response times contract, granting an advantage to those who can act more swiftly than adversaries. Vice Admiral Ann Rondeau, the school’s president, stated more directly, “Education remains our greatest strategic advantage.”
Whether a rack of GPUs qualifies as education or infrastructure is a question the Navy appears willing to leave unanswered for the time being.
Other articles
Nvidia has donated a DGX GB300 to the leading science institution of the US military.
Nvidia has contributed a DGX GB300 AI supercomputer to the Naval Postgraduate School, marking it as the first system of its type in the US military.
