Top-tier chips given away to the U.S. military, NVIDIA’s strategy is crystal clear

NVIDIA has made a highly symbolic move by donating its most advanced current-generation GB300 AI supercomputer—free of charge—to a nonprofit foundation affiliated with the U.S. Naval Postgraduate School. This hardware represents the absolute peak of NVIDIA’s performance capabilities today, and this deployment marks the first time the U.S. military has directly adopted top-end servers based on the Blackwell architecture.

The school falls under direct jurisdiction of the U.S. Navy, and all its master’s and doctoral programs are centered around war-related applications—drone tactics, maritime target recognition, battlefield AI simulations are core research areas. The Commander of the Indo-Pacific Command was present at the event. The equipment has been installed at the school’s coastal campus in California, providing technical support specifically for the Pacific theater—the military significance is unmistakable.

Huang Renxun openly stated that AI will become a cornerstone of U.S. national defense. The U.S. military plans to use this supercomputer to rapidly compile strike lists and simulate joint operations among unmanned systems, accelerating the practical application of military AI.

What’s most striking about this entire incident is NVIDIA’s starkly dualistic approach. To curry favor with the U.S. military, it freely gifts top-tier computing power; yet when it comes to the Chinese market, constrained by U.S. regulations, NVIDIA never allows the GB300 model—instead offering only downgraded, specially tailored chips with reduced performance. On one hand, it fully supports strengthening America’s military computing advantage and maintains a solid hardware gap; on the other, it refuses domestic enterprises equal access to high-end hardware.

Interestingly, Huang Renxun recently publicly praised the outstanding capabilities of China’s open-source AI models and urged American companies to actively adopt them. While rigorously blocking the flow of advanced technology into China—afraid that China might gain powerful computing resources to catch up—it still covets China’s mature open-source ecosystem, hoping to leverage domestic AI achievements to enhance NVIDIA’s own software stack. It wants to restrict hardware exports while also profiting from China’s AI development dividends.

NVIDIA didn’t sell the equipment directly to the military but chose to transfer it through a non-profit organization—a maneuver that is both tactful and effective. This approach simplifies deployment procedures while enabling deep integration into the U.S. military’s research system. NVIDIA gains access to vast amounts of real-world military computing data to refine its chips, while the military benefits from hardware-optimized combat strategies. Both sides achieve their goals. Moreover, the institution trains thousands of U.S. military officers annually, who grow accustomed to using NVIDIA equipment—ensuring that future large-scale procurement of computing power by the U.S. military is effectively locked in.

Technology knows no borders, but corporations have clear affiliations. Every decision NVIDIA makes prioritizes U.S. interests: donating chips strengthens America’s military edge, while selling downgraded chips maximizes profits in the Chinese market. This dual-pronged strategy serves as a timely reminder that high-end computing power cannot be reliant on external supply—developing indigenous AI chips with complete software and hardware independence is the only long-term, stable path forward.

Original article: toutiao.com/article/1871496413351948/

Disclaimer: This article represents the personal views of the author