Oracle与Google宣布：Google TPU将入驻Oracle云基础设施，2027年上线，为Google之外首个超大规模云平台
Oracle与Google于10月27日在内华达州拉斯维加斯举行的Oracle AI World大会上宣布，Google的张量处理单元（TPU）将通过Oracle云基础设施（OCI）提供服务，这是Google TPU首次登陆Google之外的超大规模云平台；Google TPU在OCI上的服务定于2027年上线，并将在数年内逐步扩大规模。
两家公司介绍，Google第八代TPU包含两种专用架构：TPU 8t面向训练，可把前沿模型的开发周期从数月缩短到数周，每个舱（pod）的算力接近上一代的3倍；TPU 8i面向推理，专为处理大量专业智能体的迭代式工作而设计。TPU是Google与开放软件协同设计的定制加速器，覆盖从前沿训练、大规模推理到智能体多步推理的完整周期；OCI将提供在自身基础设施内规模化部署TPU所需的电力、冷却与数据中心网络。
Oracle云基础设施总裁马赫什·蒂亚加拉詹（Mahesh Thiagarajan）说：「不同的超级智能（SI）工作负载有不同的算力需求。把Google TPU引入OCI，让客户能在一个云平台上使用多种领先的算力架构，按每项工作的性能、吞吐量和经济性选择合适的基础设施。」Google高级副总裁兼首席技术官（超级智能与基础设施）阿明·瓦达特（Amin Vahdat）说：「下一代超级智能将需要多元化的算力。TPU为要求严苛的训练与推理工作负载提供了一个强大的专用选项；把TPU带到OCI，将让更多机构用上这项技术，更灵活地构建和运行其最复杂的应用。」
两家公司表示，随着模型规模增长和生产推理需求上升，每次训练与推理的性能和经济性愈发重要；客户将能在OCI内按工作负载、性能要求与成本目标选择Google TPU。Oracle称，OCI将成为首个提供Google TPU的Google之外超大规模云平台。
Oracle and Google Announce Google TPUs on Oracle Cloud Infrastructure: Coming in 2027, a First for a Hyperscale Cloud Outside Google
Oracle and Google announced at Oracle AI World in Las Vegas, Nevada, on October 27 that Google's Tensor Processing Units (TPUs) will be available through Oracle Cloud Infrastructure (OCI), marking the first time Google TPUs will be offered on a hyperscale cloud platform outside Google. Google TPUs on OCI will be available in calendar year 2027, ramping over a multi-year period.
Google's eighth-generation TPU portfolio introduces two purpose-built architectures: TPU 8t, for training, cuts frontier model development cycles from months to weeks and delivers nearly three times the compute performance per pod over the previous generation; TPU 8i, for inference, is designed to handle the iterative work of many specialized agents. Google TPUs are custom accelerators co-designed with open software to power the full lifecycle — from frontier training and large-scale inference to the multi-step reasoning demands of agentic systems — and OCI will provide the power, cooling, and datacenter networking to deploy them at scale within its infrastructure.
'Different AI workloads have different compute needs,' said Mahesh Thiagarajan, president of Oracle Cloud Infrastructure. 'Bringing Google TPUs to OCI gives customers access to multiple leading AI compute architectures in one cloud. That choice gives them the flexibility to choose the right infrastructure for the performance, throughput, and economics of each AI workload.' Amin Vahdat, Google's senior vice president and chief technologist, AI & Infrastructure, said: 'The next generation of AI will require a diverse range of compute. TPUs give customers a powerful, purpose-built option for demanding training and inference workloads. Bringing Google TPUs to OCI will give more organizations access to this technology and greater flexibility in how they build and run their most complex AI applications.' (Quotations are reproduced as issued in the companies' release.)
The companies said that as models grow and production inference demands increase, the performance and economics of every training run and inference request become increasingly important, and customers will be able to choose Google TPUs within OCI, optimized for their workloads, performance requirements, and cost objectives.
来源：GSM·美国（原文链接）
图片来源：图：Oracle（甲骨文）公司标志。图片：Wikimedia Commons，公共领域（文字与简单图形标志）。 / Photo: The Oracle Corporation logo. Source: Wikimedia Commons, public domain (text and simple geometric logo).
