By Stephen Nellis
SAN FRANCISCO, Jan 28 (Reuters) - U.S. chipmaker Nvidia helped China's DeepSeek hone artificial intelligence models that were 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.
DeepSeek shook markets early last year with a set of AI models that rivaled some of the best offerings from the United States but were developed with far less computing power, fuelling concerns in Washington that China could catch up with the U.S. in AI despite U.S. restrictions on the sale of high-powered computing chips to China.
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 from Nvidia showed the achievement came after 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,' with internal reporting boasting that 'DeepSeek-V3 requires only 2.788M H800 GPU hours for its full training' - less than what U.S. developers typically require for frontier-scale models," Moolenaar wrote in the letter.
GPU hours are the number of hours an AI chip must run to train an AI model, while frontier-scale models refer to leading models produced by U.S. firms such as OpenAI, Anthropic or Alphabet's Google.
The documents cover Nvidia activities from 2024. At the time Nvidia provided DeepSeek help, Moolenaar wrote, 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," Moolenaar wrote.
Nvidia's H800 chip was specifically designed for the China market and sold there before H800s were put under U.S. export controls in 2023. Reuters reported last year that U.S. officials believe DeepSeek is aiding China's military.
"China has more than enough domestic chips for all of its military applications, with millions to spare. 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," Nvidia said in a statement.
The U.S. Commerce Department and the Chinese embassy in Washington did not immediately respond to a request for comment. DeepSeek did not respond to a request for comment outside of business hours in China.
Earlier this month, U.S. President Donald Trump's administration approved sales of Nvidia's H200 to China with some restrictions, including that the chips not be sold to entities that assist the Chinese military. The H200 is more powerful than the H800 chips DeepSeek used.
Trump's decision drew fire from China hawks across the U.S. political spectrum over concerns the chips would supercharge Beijing's military and erode the U.S. advantage 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 to prevent such assurances from becoming superficial formalities," Moolenaar wrote.
"Chips sales to ostensibly non military end users in China will inevitably result in a violation of the military end use restrictions," he added.
(Reporting by Stephen Nellis in San Francisco; Editing by Nia Williams)
