Summary
Citing sources, The Information reported late on October 27, 2026 (UTC 23:59) that OpenAI plans to buy Google's AI chips (TPUs), with financial terms undisclosed; meanwhile Oracle announced TPUs will be available on OCI starting 2027, leaving OpenAI a cloud-based TPU option as well.
What the Report Says
What is the source, and what does the report say?
Citing sources, The Information (reporter Ryan Barwick) reported at 23:59 UTC on October 27, 2026 that OpenAI plans to buy Google's AI chips. Per FourWeekMBA's summary and verification of the report, the original article is paywalled, and financial terms and other details could not be learned.
Have purchase amount, chip quantity and delivery timing been disclosed?
No. Per The Information's source-cited report, the purchase amount, number of chips and delivery timeline have not been disclosed; FourWeekMBA's summary contains no figures either.
Reasons and Signals
Why does OpenAI want Google chips in addition to Nvidia?
The report does not specify commercial terms, but the backdrop is OpenAI's push on both owned data centers and multi-cloud compute. Per Nvidia's official press release, OpenAI is a long-time Nvidia customer: its GPT-6 Astra is trained and served on the Abilene OCI Supercluster (over 100,000 Nvidia Blackwell GPUs). Adding Google TPUs alongside Nvidia diversifies compute sources.
What does OpenAI's TPU engineer job listing signal?
Per OpenAI's careers page (read on October 28, 2026), OpenAI lists a "Software Engineer, TPU Compilers Workload" role in San Francisco on the Inference team. The job covers optimizing the full-forward-pass performance of frontier models on TPUs, with compensation of $380,000 to $500,000 plus equity. The listing shows OpenAI is already building an engineering team for TPU inference workloads.
Will the chips deploy in OpenAI's own data centers or on public clouds?
Per The Information's source-cited report, OpenAI plans to buy Google chips directly; the deployment location is undisclosed. The backdrop: Google is delivering TPU systems into customer data centers (see next question), and Oracle on the same day announced TPUs on OCI from 2027 — so OpenAI has both paths open: self-purchased chips and cloud-based TPUs.
Industry Impact
How far has Google's external TPU sales progressed?
Per Alphabet's July 2026 Q2 earnings call (verified via FourWeekMBA's summary), Alphabet has for the first time delivered TPU systems into customer data centers. Alphabet said most related contract revenue will be recognized in 2027; Alphabet reported Q3 earnings on October 28, 2026.
What does this mean for Nvidia and Google?
Per The Information's source-cited report, if the deal closes, Google gains a top-tier customer in OpenAI, while Nvidia's single share of OpenAI's compute dilutes. Per Nvidia's press release, OpenAI currently trains and serves GPT-6 Astra on over 100,000 Blackwell GPUs in the Abilene OCI Supercluster; Nvidia the same day also announced expanding Vera Rubin NVL72 with Oracle to multi-gigawatt scale, showing its public-cloud capacity tied to OpenAI is still growing.