OpenAI posts TPU engineer role as report points to Google chip purchases

OpenAI has posted a job opening for an engineer to lead work optimizing its frontier AI models for Google's Tensor Processing Units (TPUs), including workloads for training or inference. The listing follows a report by The Information on October 28, citing sources, that OpenAI plans to buy Google TPUs.

OpenAI declined to comment on the report, and no purchase or deployment schedule has been confirmed.

Google began selling TPUs directly to outside customers in the second quarter of this year. On the 27th, Google announced that Oracle plans to offer TPUs through its cloud infrastructure, with service starting in 2027 — the first case of another major cloud provider offering TPUs outside Google's own infrastructure, according to the announcement.

OpenAI has recently said it would offer its latest AI models through Google Cloud, and it has rented Nvidia GPUs there. Nvidia is both a GPU supplier to OpenAI and an investor in the company.

In August last year, Nvidia agreed to provide more than $100 billion in credit support for a large data center project OpenAI would build in Ohio, and Nvidia CEO Jensen Huang said OpenAI committed to spending about $600 billion on Nvidia servers through 2030. OpenAI also plans to deploy custom AI chips designed with Broadcom starting later this year.

Google has developed TPUs since 2013 and has provided them through its own cloud. Last year, Anthropic signed a contract with Google for up to 1 million TPUs.