A congressman claims that Nvidia assisted DeepSeek in refining AI models that were subsequently employed by China's military

U.S. Lawmaker Says Nvidia Helped China’s DeepSeek Develop AI Later Used by Military

U.S. chipmaker Nvidia provided technical assistance that helped China’s DeepSeek develop artificial intelligence models later used by the Chinese military, the chairman of a U.S. House of Representatives committee said in a letter seen by Reuters on Wednesday.

The allegation comes amid heightened concern in Washington over China’s rapid progress in AI, despite U.S. restrictions on the export of advanced computing chips.

DeepSeek rattled global markets early last year after releasing AI models that rivalled leading U.S. systems but were developed using significantly less computing power, fuelling fears that China could narrow the AI gap with the United States even under export controls.

In a letter to U.S. Commerce Secretary Howard Lutnick, Representative John Moolenaar, a Michigan Republican who chairs the House Select Committee on China, said documents obtained by the committee showed DeepSeek’s breakthroughs followed extensive technical assistance from Nvidia.

“According to NVIDIA records, NVIDIA technology development personnel helped DeepSeek achieve major training efficiency gains through an ‘optimized co-design of algorithms, frameworks, and hardware,’” Moolenaar wrote.

Internal Nvidia reporting cited in the letter said DeepSeek-V3 required just 2.788 million H800 GPU hours for full training, far fewer than those typically needed by U.S. developers to train frontier-scale models, such as those produced by OpenAI, Anthropic or Google.

GPU hours measure the time AI chips must run to train a model, while frontier-scale models refer to the most advanced systems at the cutting edge of AI development.

The documents reviewed by the committee cover Nvidia’s activities in 2024. At the time the assistance was provided, Moolenaar said, there was no public indication that DeepSeek’s technology was being used by China’s military.

“Nvidia treated DeepSeek accordingly—as a legitimate commercial partner deserving of standard technical support,” he wrote.

Nvidia and China respond

Nvidia’s H800 chip, which DeepSeek used, was designed specifically for the Chinese market and sold there before it was brought under U.S. export controls in 2023.

Reuters reported last year that U.S. officials believe DeepSeek is aiding China’s military.

In response, Nvidia said China does not rely on U.S. chips for military applications. “China has more than enough domestic chips for all of its military applications, with millions to spare,” the company said. “Just like it would be nonsensical for the American military to use Chinese technology, it makes no sense for the Chinese military to depend on American technology.”

Liu Pengyu, a spokesperson for the Chinese Embassy in Washington, said China opposes efforts to “overstretch the concept of national security” or politicise trade and technology, and urged the U.S. to take steps to keep global supply chains stable.

The U.S. Commerce Department did not immediately respond to a request for comment. DeepSeek did not respond to requests for comment outside business hours in China.

Export controls under scrutiny

Earlier this month, the administration of U.S. President Donald Trump approved sales of Nvidia’s more powerful H200 chips to China under restrictions barring sales to entities that assist the Chinese military. The decision drew criticism from China hawks across the political spectrum, who warned the chips could strengthen Beijing’s military capabilities and undermine the U.S. lead in AI.

“If even the world’s most valuable company cannot rule out the military use of its products when sold to Chinese entities, rigorous licensing restrictions and enforcement are essential,” Moolenaar wrote.

“Chip sales to ostensibly non-military end users in China will inevitably result in violations of military end-use restrictions,” he added.