Oracle plans Google TPU access through its cloud in 2027
Oracle and Google plan to bring Google’s Tensor Processing Units to Oracle Cloud Infrastructure in 2027, followed by a multiyear capacity increase. The agreement covers training, post-training and inference workloads. Oracle will provide power, cooling and networking. The planned service adds an accelerator option; chip quantities have not been disclosed.
Artificial Intelligence··Evening
Google’s accelerators are planned for Oracle’s cloud
Oracle and Google announced an agreement on 27 October to offer Google’s Tensor Processing Units through Oracle Cloud Infrastructure, Oracle’s cloud platform. These purpose-built AI accelerators are planned to become available in 2027, with capacity increasing over several years. Oracle describes the deployment as the first offering of Google TPUs on a hyperscale cloud outside Google. The Register independently reported the plan and its announced starting year.[1], [2]
Oracle will provide power, cooling and networking
Oracle will provide the power, cooling and data-centre networking for the planned deployment. The partners identify model training, post-training and production inference among the intended workloads. Inference is the use of a trained model to generate responses. The announcement also names reasoning, agent workflows and long-context applications. Oracle Cloud Infrastructure president Mahesh Thiagarajan described the agreement as giving customers a choice among computing architectures for differing workload requirements.[1]
Capacity will grow after the initial service starts
The Register notes that Oracle did not disclose how many TPUs would be provided in the first year or later. Google infrastructure executive Amin Vahdat said the partnership would give more organizations access to the technology and greater flexibility in building and running complex AI applications. The capacity increase is planned to extend over multiple years after the service starts.[1], [2]
