[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"news-12671bb9-c100-4ca3-82e0-dc7c26286a0e":3},{"id":4,"title":5,"summary":6,"original_url":7,"source_id":8,"tags":9,"published_at":23,"created_at":24,"modified_at":25,"is_published":26,"publish_type":27,"image_url":13,"view_count":28},"12671bb9-c100-4ca3-82e0-dc7c26286a0e","英伟达 GPU 毛利率 88%：AI 算力成本高企下的自研芯片暗潮","英伟达控制着 81% 的数据中心 AI 芯片市场，上个财年数据中心业务收入 1937 亿美元，毛利率高达 75%。对英伟达顶尖 GPU 芯片的拆解报告显示，其制造成本约 3300 美元，但售价高达 2.8 万美元，利润率高达 88%。如此高的利润，本质上是一种向整个 AI 行业征收的英伟达税。\n\n这种超额利润正在推动整个行业寻找替代方案。Google 的 TPU、亚马逊的 Trainium、微软的 Maia、Meta 的 MTIA，以及 OpenAI 与博通合作设计的 AI 芯片，都是科技巨头们去英伟达化的尝试。数据中心周围的居民在不知不觉中承担了这场算力博弈的成本——电费账单里有一部分实际上是在向英伟达缴税。\n\n这场税的压力正在加速两个趋势：一是自研芯片的投入，从云厂商到应用层公司都在探索自建 ASIC；二是模型层面的效率优化，量化、蒸馏、长上下文等技术的突破，本质上都是在用更少的算力完成同等任务。英伟达的护城河短期内难以撼动，但当整个行业都在想办法绕开它时，拐点或许只是时间问题。","https:\u002F\u002Fwww.solidot.org\u002Fstory?sid=84438","d59894d3-308e-4fd8-8865-86dc1eeac4a2",[10,14,17,20],{"id":11,"name":12,"slug":12,"description":13,"color":13},"7ac06d8e-b074-4147-abfc-ffaa4c6b8744","ai-efficiency",null,{"id":15,"name":16,"slug":16,"description":13,"color":13},"e0d31e94-ce47-4c8f-831c-d3d2926d42f3","hardware",{"id":18,"name":19,"slug":19,"description":13,"color":13},"0a93ec8e-ea39-4693-81de-563ca8c173f7","inference",{"id":21,"name":22,"slug":22,"description":13,"color":13},"8dac812d-3839-4abe-a855-5f56ec9515fd","nvidia","2026-05-29T22:00:00Z","2026-05-29T22:06:48.401724Z","2026-05-29T22:06:48.401736Z",true,"agent",9]