OpenAI Plans to Buy Google's AI Chips as It Looks Beyond Nvidia: Report

What it claims

The Information reports, through one unnamed source, that OpenAI plans to buy Google's tensor processing units (TPUs) outright instead of renting them by the hour. Neither company has confirmed the plan. Price, chip count and delivery dates are not disclosed. In June 2025, OpenAI began renting TPUs and called it early testing with no plans for large-scale deployment. Decrypt treats two facts as signs that the plan is real:

• OpenAI has a San Francisco job listing for an engineer to lead optimization of its frontier models on TPUs. The work is compiler work, which the article gives as the reason that changing chips is harder than swapping parts.
• Richard Ho, who leads OpenAI's hardware program, spent nearly nine years at Google and was one of the leaders of its Cloud TPU project.

Evidence

The article sets the report against OpenAI's other compute commitments:

• Nvidia: In September 2025, Nvidia signed a letter of intent to invest up to $100 billion in OpenAI, tied to OpenAI deploying at least 10 gigawatts of Nvidia systems. Nvidia's actual investment in OpenAI's funding round was $30 billion. Nvidia CEO Jensen Huang said in March 2026 that it might be the last such investment before OpenAI goes public.
• AMD: Two weeks after the Nvidia letter, AMD gave OpenAI the right to buy up to 160 million AMD shares, roughly 10% of the company. The right is tied to OpenAI deploying 6 gigawatts of AMD chips.
• Jalapeño: OpenAI's own inference chip is built with Broadcom and was unveiled in June 2026. Initial deployment is planned for the end of 2026.

On Google's side:

• TPU 8t and 8i: Google announced two eighth-generation TPUs in April 2026, the 8t for training and the 8i for inference. Google states their performance only against its previous chip, Ironwood, not against Nvidia hardware.
• Customer data centers: Alphabet began delivering TPU systems to customers' own data centers in the second quarter. Its CFO expects most revenue from those system sales in 2027.
• Anthropic: In April 2026, Anthropic signed an agreement with Google and Broadcom for multiple gigawatts of next-generation TPU capacity. Anthropic expects that capacity to come online starting in 2027.
• GPT-6 on Google Cloud: Google Cloud began offering OpenAI's GPT-6 models to select customers in private preview in the week before publication.

Limits

The central claim rests on one anonymous source quoted secondhand, and neither company has confirmed it. The article gives no volume, price, delivery schedule, chip generation or workload split between training and inference for any OpenAI TPU purchase. It also gives no performance comparison between TPUs and Nvidia or AMD hardware. The only benchmark it mentions compares Google's new chips with Ironwood.