Robert Way
The next large language model from Chinese artificial intelligence company DeepSeek (DEEPSEEK) could be impacted by the lack of access to Nvidia’s (NASDAQ:NVDA) high-end GPUs due to the ongoing U.S. export control curbs, The Information reported.
The next model, known as R2, is likely to see slower adoption, given how optimized DeepSeek’s models are to Nvidia’s hardware, the news outlet added, citing people with knowledge of the matter.
In conjunction, DeepSeek has not yet a release date of R2, as CEO Liang Wenfeng is not satisfied with the performance of the model, two of the people added.
The majority of cloud customers that are running DeepSeek’s R1 model use Nvidia’s H20 Hopper GPUs, some of the people added.
DeepSeek did not immediately respond to a request for comment from Seeking Alpha.
The H20 is part of Nvidia’s Hopper line of GPUs and was released in 2023. Nvidia has since released its Blackwell line of GPUs. This past March, Nvidia took the wraps off its next two generations of AI accelerators, including Blackwell Ultra and the widely speculated Rubin.
Nvidia shares rose fractionally in early trading on Thursday, one day after hitting a record high.