OpenAI Eyes Google Chips Beyond Nvidia

OpenAI is reportedly preparing to purchase artificial intelligence chips from Google, signalling a potential shift in its hardware procurement strategy as competition for computing capacity intensifies. The proposed arrangement, first reported by The Information on 27 October, would deepen an increasingly significant commercial relationship between two companies competing to develop advanced AI models and services.

According to Reuters, the planned purchase concerns Google's Tensor Processing Units (TPUs), specialised processors designed to handle demanding AI workloads. OpenAI already uses large quantities of Nvidia graphics processing units for model training and inference, the process through which trained models generate responses. The company began renting Google's AI chips in 2025 and recently announced plans to make its frontier models available through Google Cloud. Neither company has confirmed the proposed chip purchase, and financial terms have not been disclosed.

The development comes as technology companies seek alternatives to Nvidia's dominant hardware. Google has increasingly positioned its TPUs as commercially viable processors for external customers, while cloud providers compete to offer more flexible infrastructure. The search for alternative chips reflects growing concerns about supply, energy consumption and the financial demands of constructing AI data centres.

For OpenAI, diversifying its hardware could provide greater flexibility in meeting expanding computing requirements. For Google, supplying chips to a direct competitor could strengthen its position in the market for AI infrastructure. The reported arrangement illustrates how commercial partnerships are increasingly crossing competitive boundaries as companies race to secure the computing power behind advanced artificial intelligence.