OpenAI plans to buy Google TPUs, The Information reports

Neither company has confirmed the plan, and the price, the number of chips and the delivery dates have not been disclosed. Bloomberg reported on 27 October that OpenAI was reviewing adoption of Google's chips and considering a purchase.

The reported deal would deepen a relationship that began with renting. Reuters reported that OpenAI started renting Google's TPUs in 2025, work the company described as early testing with no plans to deploy them at scale. An OpenAI spokesperson previously said the company "has no plans to use TPUs specifically." Buying the hardware would mean owning it instead of paying Google by the hour.

The job listing behind the report

The clearest public evidence sits in OpenAI's hiring. The company advertised a job in San Francisco for an engineer to lead optimization of its frontier models on TPUs. The work involves compilers, the software that translates a model into instructions a specific chip can run, which is why switching chips is harder than swapping parts. Richard Ho, who leads OpenAI's hardware program, spent nearly nine years at Google and helped lead its Cloud TPU project.

OpenAI already buys from several suppliers

Google would join a supplier list that is already crowded. Nvidia signed a letter of intent in September 2025 to invest up to $100 billion in OpenAI alongside at least 10 gigawatts of its systems, then put $30 billion into the funding round. Two weeks later, AMD gave OpenAI the right to buy up to 160 million of its shares, roughly 10% of the company, tied to OpenAI deploying 6 gigawatts of AMD chips. In June, OpenAI unveiled Jalapeño, its first custom chip, built with Broadcom for inference, with initial deployment planned for the end of 2026. IJR reported that OpenAI has committed to spend about $600 billion on Nvidia servers through 2030.

Google's TPU business is widening

Oracle said on 27 October that it will offer Google TPUs through its cloud, the first time the chips have been offered on another hyperscaler's platform. Oracle is buying the TPUs directly. Anthropic, OpenAI's main rival, agreed with Google and Broadcom in April for multiple gigawatts of next-generation TPU capacity, which it expects to begin operating in 2027. Its IPO-related filing says it plans to spend $114.5 billion on Google computing resources from April 2026 to July 2033, and estimates $125.0 billion to $150.0 billion on TPU compute across providers.

Analysts expect the business to grow quickly from a small base. Mizuho forecasts Google TPU revenue of $28.0 billion in 2026 and $67.0 billion in 2027, rising to as much as $100.0 billion in 2028. TD Cowen projects $138.0 billion for 2028, or 34% of Google Cloud revenue. Alphabet's finance chief said TPU systems began shipping for customers' own data centers in the second quarter, with most of that revenue landing in 2027. For scale, Nvidia posted revenue of $96.2 billion in its latest quarter alone.

What it means for Nvidia

Nvidia still supplies most of the hardware behind the AI boom, and OpenAI ranks among its biggest customers. Some analysts expect the spread of multi-chip strategies to weaken Nvidia's pricing power, because buyers who will accept Google or AMD chips have more to negotiate with. Nvidia chief executive Jensen Huang has played down the threat. "If there is no Anthropic, is there any reason for TPUs to grow?" he said earlier this year in an interview with podcaster Dwarkesh Patel.

What happens next

Whether any OpenAI purchase of Google chips closes, and on what terms, remains unconfirmed. Jalapeño's initial deployment is planned for the end of 2026, and Anthropic expects its contracted TPU capacity to come online in 2027.