The Information reported on the night of October 27, 2026, citing one person familiar with the matter, that OpenAI plans to buy Google's custom tensor processing units (TPUs), the chips Google built specifically for AI work.

Reuters picked up the story hours later, but neither company has confirmed it: OpenAI declined to comment and Google did not immediately respond. Sixteen months later, buying the chips outright would mark a structural bet on Google silicon alongside Nvidia.

Until this week, Nvidia was effectively the only road for OpenAI's compute, and now the most talked-about company in AI is openly discussing chips from its closest rival in consumer AI products.

The pairing broke out on Google Trends in the United States within 24 hours, according to Pasquale Pillitteri's report.

The business logic is notable too: Google competes with OpenAI in search and AI products, yet last week OpenAI said its frontier models would become available on Google Cloud, after a similar move on Amazon Web Services.

What was once a purely adversarial relationship is quietly becoming a supply-chain one, which few predicted when OpenAI dismissed TPU plans a year ago.

The strongest public clue is not a press release but a verified job post: OpenAI has published an official listing for a Software Engineer, TPU Compilers, based in San Francisco, seeking someone to lead optimization of its frontier models on TPUs, from kernel writing to inference quantization, per Pasquale Pillitteri's report.

Companies do not hire for years-long compiler work over occasional capacity needs. The same report notes OpenAI previously named Richard Ho, a former Google senior director of engineering and one of the people behind Cloud TPU, to head its hardware division.

OpenAI remains one of Nvidia's largest customers, and 24/7 Wall St. reports Nvidia agreed on August 17, 2026 to provide up to $105 billion in credit support for an OpenAI data center in Ohio starting at 4.25 gigawatts, with CEO Jensen Huang reportedly citing a commitment of about $600 billion in Nvidia server spending through 2030.

The risk is not immediate volume but perception: each credible alternative erodes Nvidia's leverage in negotiations, pressuring a stock that trades at a P/E of 44 and assumes its 71.1% gross margin holds.

For Alphabet, an OpenAI order would validate TPUs as an outside chip business; its CFO Anat Ashkenazi has said most TPU sales revenue will be recognized in 2027, though Anthropic, Apple, Citadel Securities and Meta already rent or buy capacity.

Significant caveats remain: no terms, volumes, timing or chip generation have been disclosed, and until either company confirms, this is a reported plan rather than a signed order. The first test comes at Alphabet's quarterly results after the close on October 28, 2026.

Per 24/7 Wall St., if management names outside TPU buyers or raises its 2027 TPU revenue outlook, the case strengthens; if the call skips TPUs and the stock closes below $320.58, the market will treat the report as noise.

The next six months may show whether ChatGPT inference quietly shifts, and whether OpenAI's story becomes a multi-vendor one spanning Nvidia, Google TPUs, its Broadcom-built custom chip and a roughly 6 gigawatt AMD deal signed in October 2025.