JPMorgan analyst says Nvidia's GPU share of finalized compute procurement by frontier AI labs has fallen to 39%
English text is being prepared.
Michael Cembalest, Chairman of Market and Investment Strategy at JPMorgan Asset Management, said in a report that among the compute infrastructure that frontier AI labs such as OpenAI and Anthropic have finalized and will procure in the future, Nvidia GPUs account for only 39%, while from 2023 to 2025 Nvidia's share of overall AI chip revenue reached 83%. Broadcom custom chips ranked second at 24%, Google TPUs accounted for 16%, AMD GPUs for 15%, and Amazon Trainium for 6%. In October 2025, OpenAI announced a partnership with Broadcom to develop 10 gigawatts of custom AI accelerators, while also finalizing the procurement of 6 gigawatts of GPU compute from AMD.
Why read this
AI score 58
The JPMorgan analyst laid out the chip share structure of finalized compute procurement by frontier labs, which can be compared with Nvidia's previous 83% revenue share to see changes in customer structure.
Topics
Body
文章主体 EM_StockImg_Start
EM_StockImg_End
摩根大通资产管理市场与投资策略主席迈克尔·切巴莱斯特在前一日发布的报告中称,OpenAI、Anthropic这类前沿AI实验室已经敲定、未来将采购的算力基础设施里,英伟达GPU占比仅39%。而2023至2025年,英伟达在AI芯片整体收入中的份额曾达到83%,这预示着它在下一代算力基础设施中的占比可能大幅萎缩。
竞争对手博通的定制芯片份额位居第二,达24%;谷歌张量处理器(TPU)占16%,AMD GPU占15%,亚马逊Trainium芯片占6%。这一结构说明,谷歌、亚马逊、微软等曾经的英伟达客户,或是和博通合作自研芯片,或是选择性价比更高的AMD处理器。OpenAI在2025年10月宣布与博通合作,开发10吉瓦的定制AI加速器,同时敲定从AMD采购6吉瓦算力的GPU。
(文章来源:财联社) 正文中部 内嵌广告