Marketing is reported that due to the influence of US AI chip export control, the new AI founder DeepSeek, which was born in China, is unlikely to obtain NVIDIA (Nvidia Corp.) high-level picture processor (GPU), and the next-generation AI model may...
Marketing is reported that due to the influence of US AI chip export control, the new AI founder DeepSeek, which was born in China, is unlikely to obtain NVIDIA (Nvidia Corp.) high-level picture processor (GPU), and the next-generation AI model may be affected.
Seeking Alpha and Tom’s Hardware quoted The Information report on the 26th. Unnamed sources revealed that the specific release time of DeepSeek's next-generation AI model "R2" has not been determined, because executive chief Liang Wenfeng is not satisfied with it yet, and the company is still improving the model performance.
DeepSeek's "R1" model has been quickly and widely adopted by China's private new, large enterprises and government-related organizations. Sources revealed that these cloud customers use NVIDIA's "H20" Hopper architecture GPU when executing R1.
Reporting said that the shipping of H20 is now subject to the control of the Washington Bureau, which has begun to restrict the use of R1 and also put the release of R2 on the side.
DeepSeek internal staff believe that if R2 performance exceeds all open alternative platforms on the market, the usage forecast will surge, and the Chinese cloud platform may be difficult to match.
According to the report, DeepSeek's AI model is to optimize NVIDIAGPUs, which makes the company particularly vulnerable to US political decisions.
