OpenAI is hiring an engineer to tune its models for Google’s TPUs

Updated October 28, 2026: a job listing described by The Information shows OpenAI is recruiting an engineer to work on its models on Google’s TPUs.

OpenAI plans to buy Google’s AI chips, according to The Information’s report, and a job listing now points the same way: the company seeks an engineer to “lead the optimization” of its models on tensor processing units (TPUs). The purchase itself is still unconfirmed.

A job post that backs up the report

The Information, as summarized by Techmeme, attributes the purchase plan to a person familiar with the matter. It adds that a listing shows OpenAI looking for someone to run model optimization on TPUs. The outlet frames the move as a boost for a rising competitor to Nvidia, which backs OpenAI.

That is the first piece of evidence in the story that does not depend on an unnamed source. A hiring post is something OpenAI wrote itself, and it points to engineering work on Google’s chips rather than a vague interest. But it is not an announcement. A listing for one role does not say how many chips OpenAI would buy, what they would cost, or when they would arrive.

As far as we could find, neither OpenAI nor Google has commented publicly on the plan, and Nvidia has not either.

What the report and the listing leave open

Several questions decide how big this is:

• Scale: no figure for the number of chips or the size of the order has been given.
• Buy or rent: the report speaks of buying, but it does not say whether that means owning hardware or renting capacity in Google’s cloud.
• Workload: the listing mentions optimizing models on TPUs, but not whether that means training, serving users, or both.
• Nvidia exposure: nothing in the reporting says OpenAI will reduce its Nvidia purchases. A TPU hire is consistent with adding a second supplier, not replacing the first.

Optimization work also hints at the effort involved. Models tuned around one vendor’s chips usually need engineering time before they run efficiently on another’s, which is a reason to hire a dedicated lead before buying at scale. That is an inference from the listing, not something either company has said.

Google’s chips are showing up in other stacks

The timing fits a pattern. As we reported, Oracle is adding Google TPUs, OpenAI’s own Jalapeño chip and Cerebras hardware to its cloud lineup, according to CNBC. We covered Oracle’s Nvidia side of that news in Oracle puts Vera Rubin and Vera CPUs on sale and its software side in Oracle’s business apps head into ChatGPT.

For Google, selling TPUs to OpenAI would put its hardware in the hands of a direct rival in chatbots and AI models. For Nvidia, it would show that even a customer it backs is keeping more than one source of compute.

Until OpenAI, Google or Nvidia say something on the record, the size, price and timing of any purchase remain unknown. We will update this article if any of them confirm it.