DeepSeek Snubs Nvidia by Withholding Latest AI Model

DeepSeek, the Chinese artificial intelligence lab, has reportedly withheld its upcoming V4 flagship model from U.S. chipmakers, including Nvidia and AMD ^[1]. This decision breaks standard industry practice, where software developers typically provide early access to major hardware manufacturers for performance optimization ^[1][2]. By restricting these pre-release versions, DeepSeek limits the ability of American companies to fine-tune their chips for the new software ^[2][4]. Instead, the lab has granted early access to domestic Chinese suppliers such as Huawei Technologies ^[1][3]. Industry analysts view this as a strategic effort by the Chinese government to ensure domestic hardware remains competitive while disadvantaging Western tech in the region ^[1][4]. The move comes shortly after reports that DeepSeek may have used Nvidia’s restricted Blackwell chips for previous training, a potential violation of U.S. export controls ^[1][2]. U.S. officials are currently examining these developments as tensions over AI supremacy escalate ^[2]. While the immediate impact on Nvidia’s broader data center revenue is expected to be minimal, the shift highlights a growing divide in the global AI supply chain ^[1][3].