OpenAI Plans Purchase of Google TPUs to Diversify Beyond Nvidia Silicon

OpenAI plans to purchase artificial intelligence processors directly from Google, diversifying its hardware base beyond an exclusive reliance on Nvidia systems, according to reporting from The Information cited by TechRadar Pro.

The transaction centers on Google's proprietary tensor processing units (TPUs). While chip quantities, contract valuations, and delivery schedules have not been disclosed, the agreement signals how leading AI developers are broadening their hardware supply lines to meet the compute requirements of training and serving advanced models.

The move deepens an operational bridge between two direct commercial competitors. OpenAI already ranks among Nvidia's largest GPU customers, but it began renting Google AI chips in 2025 and previously announced that its models would be made available on Google Cloud. The prospective direct hardware purchase indicates an expansion from cloud instance rental to a broader silicon procurement strategy.

Alphabet has gradually opened commercial access to its custom silicon. The search company previously restricted TPUs largely to internal workloads and managed cloud infrastructure before opening commercial silicon sales to select third-party enterprise customers. Alphabet Chief Financial Officer Anat Ashkenazi previously noted that the company anticipates recognizing initial revenue from TPU commercial agreements late in 2026, with the primary financial realization occurring across 2027.

OpenAI is simultaneously pursuing proprietary silicon alongside third-party merchant hardware. In June 2026, the lab announced a collaboration with semiconductor designer Broadcom to develop custom processors, targeting an initial deployment before the end of 2026.

That multi-vendor approach unfolds against record capital outlays across the sector. Combined infrastructure spending by Alphabet, Amazon, and Microsoft is projected to reach approximately $600 billion during 2026, intensifying market attention on hardware efficiency, delivery timelines, and revenue realization.