OpenAI plans to buy Google AI chips, report says

The Information reported on October 27, 2026, that OpenAI plans to buy Google’s AI chips.[1] The report attributed the planned purchase to a person familiar with the matter.[1] Google’s TPUs are specialized processors developed for AI workloads, and they compete with Nvidia GPUs for those workloads.[2] The evidence does not give a purchase amount, price, delivery schedule or other transaction details; the financial terms and other details were not known.[2] That leaves the size and practical timing of the reported plan unclear.

OpenAI’s job listing seeks an engineer to lead optimization of its frontier models on TPUs, with the role based in San Francisco.[3] The engineer would be responsible for performance across the full forward pass, from model code through kernels to hardware execution.[3] The listing identifies latency, throughput and hardware utilization as optimization targets.[3] It also says the engineer may explore quantization to improve inference efficiency while preserving model quality.[3] The posting describes technical work OpenAI wants done; by itself, it does not establish how many chips the company plans to buy.

OpenAI began renting Google TPUs through Google Cloud in 2025, according to The Information.[2] The reported plan to buy Google AI chips would therefore be distinct from that existing rental arrangement, although the evidence does not explain how the two would relate.[1][2] Google has also begun selling TPUs directly to selected customers.[2] OpenAI has been one of Nvidia’s largest customers, using Nvidia GPUs to train models and run inference workloads; Nvidia has invested billions of dollars in OpenAI.[2]

OpenAI recently announced that its frontier models would be available through Google Cloud.[2] Google said the models were rolling out in private preview to select Google Cloud enterprise customers.[4] Google also said customers could access the models through its APIs and connect Codex CLI and Desktop workflows to its Gemini platform for inference.[4] Separately, Oracle said Google TPUs would become available on OCI in calendar year 2027, with capacity increasing over multiple years.[5] These cloud announcements describe model access and chip availability, not the terms of OpenAI’s reported purchase.