A top U.S. House lawmaker says Nvidia provided extensive technical help to China’s DeepSeek that improved the startup’s AI training efficiency, and that DeepSeek’s models were later used by the Chinese military. The claims are laid out in a letter sent to U.S. Commerce Secretary Howard Lutnick by Representative John Moolenaar, who chairs the House Select Committee on China.
The letter says the committee obtained documents from Nvidia that describe the company’s technical assistance to DeepSeek and how that support contributed to “major training efficiency gains.” Moolenaar wrote that Nvidia’s technology development personnel helped DeepSeek through an “optimized co-design of algorithms, frameworks, and hardware,” and he cited internal reporting that “DeepSeek-V3 requires only 2.788M H800 GPU hours for its full training.”
What the letter alleges
Moolenaar said the documents reviewed by the committee cover Nvidia activities from 2024. He wrote that at the time Nvidia provided support, there was no public indication that DeepSeek’s technology was being used by China’s military, and he said Nvidia treated DeepSeek “as a legitimate commercial partner deserving of standard technical support.”
The letter also explains that “GPU hours” refer to the number of hours an AI chip must run to train an AI model. It adds that “frontier-scale models” refer to leading models produced by U.S. firms such as OpenAI, Anthropic, or Alphabet’s Google.
Why DeepSeek drew attention in Washington
The Reuters report says DeepSeek shook markets early last year with a set of AI models that rivaled some of the best offerings from the United States, while being developed with far less computing power. The report says that fueled concerns in Washington that China could catch up with the United States in AI despite U.S. restrictions on sales of high-powered computing chips to China.
Moolenaar’s letter argues that stronger licensing limits and enforcement are needed, writing that “rigorous licensing restrictions and enforcement are essential” if companies cannot rule out military use when products are sold to Chinese entities. He also wrote that chip sales to “ostensibly non military end users in China” will “inevitably” result in violations of military end-use restrictions.
Chips, controls, and the H800 and H200
The report notes that Nvidia’s H800 chip was specifically designed for the China market and was sold there before H800 chips were placed under U.S. export controls in 2023. It also says Reuters reported last year that U.S. officials believe DeepSeek is aiding China’s military.
The report adds that earlier this month, President Donald Trump’s administration approved sales of Nvidia’s H200 to China with restrictions, including that the chips not be sold to entities that assist the Chinese military. It says the H200 is more powerful than the H800 chips DeepSeek used. The report says Trump’s decision drew criticism from China hawks across the U.S. political spectrum over concerns the chips would boost Beijing’s military and weaken the U.S. advantage in AI.
Nvidia’s response and requests for comment
In a statement, Nvidia said China has “more than enough domestic chips for all of its military applications, with millions to spare,” and argued it would make as little sense for the Chinese military to depend on American technology as it would for the American military to use Chinese technology.
The report says the U.S. Commerce Department and the Chinese embassy in Washington did not immediately respond to requests for comment. It also says DeepSeek did not respond to a request for comment outside of business hours in China.
