OpenAI Weighs Purchase of Google TPUs to Cut Nvidia Dependence

OpenAI plans to buy AI chips from Google, according to a report by The Information carried by Reuters, in a move that would extend the ChatGPT maker's hardware options beyond Nvidia's processors. The financial terms remain undisclosed. That absence of detail matters more than the headline suggests.

The proposed purchase would deepen commercial ties between OpenAI and Google's parent Alphabet. The two companies compete directly in AI products and in search. A chip deal makes them supplier and customer at the same time.

What would the deal change?

OpenAI ranks among the largest buyers of Nvidia's graphics processing units (GPUs), the chips it uses to train AI models and to run inference workloads (the process of generating answers from a trained model). Buying Google's silicon would give the company a second hardware path.

Google developed its own AI processors, known as tensor processing units (TPUs), to handle machine-learning workloads. These chips are emerging as an alternative to Nvidia's GPUs, with companies exploring options that could deliver cost and performance advantages for specific AI tasks. No benchmark data, pricing, or volume figures appear in the reporting. The claimed advantages remain unquantified.

How is Google selling TPUs now?

Google has traditionally made its TPUs available only through its own cloud infrastructure. That model kept the chips inside Google's walls. The company has now started selling the chips directly to select customers, and the expansion of its chip business could help Google capitalise on growing demand for AI computing. Which customers qualify, and at what price, is not stated.

OpenAI is not starting from zero. The company had already begun renting Google's AI chips in 2025, according to earlier Reuters reporting. The reported plan to purchase them outright would mark another step in its engagement with Google's hardware ecosystem, moving from tenant to buyer.

What is not being said?

Three gaps stand out. First, no financial terms have been disclosed, so the scale of the commitment cannot be assessed. Second, the reporting does not specify which TPU generations or configurations OpenAI would acquire. Third, nothing in the report addresses how Nvidia, the incumbent supplier, has responded to its largest customer diversifying.

The deal is reported, not signed. The Information's report and Reuters' relay describe an intention, and intentions in the AI supply market have a record of shifting before contracts close. Buyers seeking alternatives to Nvidia's dominant processors face a market where the alternative's pricing, availability, and measured performance all remain, for now, outside public view.