The US is investigating whether DeepSeek may have imported embargoed NVIDIA high-performance GPUs via Singapore



On January 20, 2025, Chinese AI company DeepSeek released its inference model, '

DeepSeek-R1 ,' as open source. Developed amid strict US semiconductor export restrictions, DeepSeek-R1 is said to offer performance on par with OpenAI's 'o1' inference model, and has attracted much attention, including reaching number one in the App Store's free app rankings . However, it has been reported that the US Department of Commerce is investigating DeepSeek over concerns that it may have illegally obtained high-performance AI chips subject to US export restrictions through a Singaporean intermediary.

US Probing If DeepSeek Got Nvidia Chips From Firms in Singapore - Bloomberg
https://www.bloomberg.com/news/articles/2025-01-31/us-probing-whether-deepseek-got-nvidia-chips-through-singapore



US reported investigating whether China's DeepSeek used restricted AI chips | Fox Business
https://www.foxbusiness.com/technology/us-reportedly-investigating-whether-chinas-deepseek-used-restricted-ai-chips

US investigates whether DeepSeek smuggled Nvidia AI GPUs via Singapore | Tom's Hardware
https://www.tomshardware.com/tech-industry/artificial-intelligence/us-investigates-whether-deepseek-smuggled-nvidia-ai-gpus-via-singapore

DeepSeek claims that it developed DeepSeek-R1 using a 'reduced-performance NVIDIA GPU' that complies with U.S. semiconductor export restrictions. DeepSeek has not disclosed the specific model of GPU used for training DeepSeek-R1, but when it released 'DeepSeek-V3' with 671 billion parameters in December 2024, it stated that 'training took approximately 2.8 million GPU hours on 2048 NVIDIA H800 GPUs.'

Chinese AI company DeepSeek releases AI model 'DeepSeek-V3' comparable to GPT-4o, with a threatening 671 billion parameters - GIGAZINE



In contrast, Meta reported that Llama 3 training had 405 billion parameters using a supercomputer equipped with 16,384 NVIDIA high-performance GPUs and H100s . It took 30.8 million GPU hours to complete the training.

Because DeepSeek-R1 shows better performance than Llama 3, suspicions have arisen that it may have been trained on a more powerful GPU cluster than the one used for Llama 3, raising speculation that it may have been trained using high-performance NVIDIA GPUs, which are subject to import restrictions.



The U.S. Department of Commerce and the Federal Bureau of Investigation (FBI) are investigating whether DeepSeek may have acquired high-performance NVIDIA GPUs through a third-party company in Singapore. Singapore is not subject to export restrictions on high-performance GPUs, and NVIDIA's revenue from Singapore has been increasing rapidly since the third quarter of 2023, when China imposed restrictions on AI chips.




NVIDIA also issued a statement saying, 'Revenue related to Singapore does not indicate a diversion to China.' 'Many of our customers have entities in Singapore and use those entities for products targeted at the US and Western countries. We insist that our partners comply with all applicable laws and take immediate action when we receive information to the contrary.'

'It appears that DeepSeek has successfully circumvented U.S. export restrictions. China must compete fairly without using U.S. hardware, and if approved, we will strengthen enforcement of restrictions on the sale of AI chips,' Secretary of Commerce Howard Lutnick said. DeepSeek has not commented on the report.

Related Posts:

in AI,   Hardware,   Software, Posted by log1r_ut