OpenAI Google TPU Purchase Report: What It Means for Nvidia

OpenAI is reportedly planning to buy Google's Tensor Processing Units (TPUs) rather than just rent them, according to a report summarised by Hardware Asylum on November 2. Neither company has confirmed the arrangement, so treat it as unconfirmed. If true, it would be a notable step in OpenAI's effort to reduce its dependence on Nvidia.

Quick summary

• Status: reported, not confirmed by OpenAI or Google.
• Shift: from renting Google TPUs (started June 2025) to owning them.
• Not a break with Nvidia: OpenAI's Nvidia commitments are about 12 GW, possibly rising to 16 GW.
• Other legs: its in-house Jalapeño inference chip with Broadcom and an AMD deal.

What exactly is being reported?

The report says OpenAI wants to move from paying hourly cloud rates for Google TPUs to owning the silicon. OpenAI began renting TPUs in June 2025, and described that at the time as early testing rather than a large deployment. Supporting hints include a recent OpenAI job listing for an engineer to optimise frontier models on TPUs and the presence of hardware lead Richard Ho, who spent nearly nine years at Google leading its Cloud TPU project.

Is OpenAI leaving Nvidia?

No. The report describes a multi-vendor strategy. OpenAI remains one of Nvidia's biggest customers, with current and planned commitments of roughly 12 gigawatts, potentially 16 gigawatts, and some estimates of up to $600 billion in spending through 2030. Nvidia has also provided up to $105 billion in credit support for an OpenAI data centre campus in Ohio that is largely restricted to Nvidia hardware.

What are OpenAI's three hardware bets?

Why would Google sell TPUs?

Alphabet has begun selling its processors to outside customers instead of using them only internally. Current offerings include the eighth-generation TPU 8t for training and TPU 8i for inference. Alphabet's finance chief Anat Ashkenazi has indicated that some TPU sales revenue arrives by end of 2026, with the bulk expected in 2027. Anthropic has also secured multi-gigawatt TPU capacity from Google and Broadcom.

Why does this matter for gamers and PC buyers?

Nvidia says demand far exceeds its production capacity, a bottleneck expected to last through fiscal 2028, and AI data centres are pulling memory and components away from consumer products. That is part of why graphics cards cost more. Read how it plays out in our GPU price rise report and our look at the GPT-6.1 Terra release, which these chips ultimately serve. More chip suppliers could ease pressure over time, but nothing here promises cheaper GPUs soon.

What is the verdict?

This is a credible-sounding but unconfirmed report. The real story is leverage: credible alternatives give OpenAI better protection against shortages and stronger bargaining power on price. Whether it works depends on whether TPU software can close the gap with Nvidia's ecosystem.

Related articles

Claude Code Blocks Game Clones: What the Opus 5.5 Rule MeansGPT-6.1 Terra: Price, Context and How It Compares to SolGemini 4 Carbon: Access, Price and How It ComparesRTX 5080 24GB Launch in Doubt: No Design Kits Yet (Report)