Nvidia(US:NVDA) 程序员的那些事·2025-06-29 11:50Group 1 - DeepSeek has not confirmed the release date for its R2 model due to dissatisfaction with its performance from Liang Wenfeng [1] - The R2 model was initially planned for release in May, aimed at enhancing programming capabilities and enabling multilingual reasoning [1] - Engineers have been optimizing the R2 model while awaiting approval from Liang Wenfeng for its release [1] Group 2 - The shortage of NVIDIA server chips in China, caused by U.S. export controls, will hinder the rapid deployment of the R2 model even if it is released [1] - The demand for DeepSeek R2 could overwhelm Chinese cloud service providers, as running AI models requires advanced NVIDIA chips [1] - DeepSeek has engaged with several Chinese cloud companies to provide technical specifications for hosting and distributing the R2 model [1] Group 3 - Most cloud customers currently using the R1 model operate it on NVIDIA H20 chips [2] - New U.S. export controls implemented in April have prohibited NVIDIA from selling H20 chips to the Chinese market, which was previously the only AI processor that could be legally exported [2]