Bloomberg：Microsoft 停用从 Oracle 租用的 AI 服务器
Microsoft 宁可弃用 Oracle 租的 H100 老服务器也不续约，这篇文章讲清了新旧 GPU 代差对推理成本的实际影响。
据 Bloomberg 报道，Microsoft 决定不再续签 2023 年围绕 Bing 达成的租约，逐步关停从 Oracle 租用的 AI 服务器。原因在于 Oracle 机房以 Nvidia H100 等 Hopper/Ampere 旧代芯片为主，单卡仅 80GB 显存、只能高速连接 8 张 GPU，难以承载当前 MoE 架构大模型，而 Blackwell 机架可将 72 张 GPU（单卡 192-288GB）互联为一个单元。Microsoft 按 GPU 小时付费，同一小时在旧芯片上能完成的推理工作量明显更少。Oracle 则表示其 AI 基础设施需求依然强劲，9 月曾以高于原价 20% 的价格续签了已使用 4 年的 GPU 合同。
Bloomberg: Microsoft is switching off AI servers it rented from Oracle after declining to renew a 2023 contract built around Bing.
The capacity lost value over time because Oracle’s hardware ran older-generation Nvidia chips that no longer matched Microsoft’s specifications, according to people familiar with the matter.
Basically, Microsoft pays per GPU-hour. A Blackwell GPU produces several times more inference output than an H100 for similar power, so each Hopper or Ampere hour now buys far less work.
Current frontier models, often large mixture-of-experts designs, need lots of memory and fast links between GPUs. An H100 has 80GB of memory and talks quickly to only 8 GPUs in its server. Blackwell racks link 72 GPUs as one unit, with 192 to 288GB each. Big models either don’t fit on the old boxes or run slowly across their slower network.
Oracle says demand for its AI infrastructure remains strong, and in September it renewed contracts for 4-year-old GPUs at a 20% premium to the original price.
