OpenAI is adding a major new supplier to the infrastructure behind its AI products. According to The Information, the company has started using Google's Tensor Processing Units, or TPUs, through Google Cloud to run ChatGPT and other services.
The decision matters because OpenAI has been one of Nvidia's biggest customers and has relied on Nvidia's GPUs for both training and deploying large language models. Using Google TPUs at scale changes that pattern and puts another major cloud provider inside OpenAI's operating stack.
What OpenAI is changing
The TPUs are being rented through Google Cloud. Their role is tied to inference, the process of running an already trained model so it can answer new prompts.
That distinction is important. Training is the stage where a model is built. Inference is the ongoing work that happens every time a user asks a system such as ChatGPT for an answer. For a product with heavy usage, inference can become a major operating cost.
The Information reports that the Google TPU arrangement is meant to lower those inference costs. The source article also says this is the first time OpenAI is relying on chips beyond Nvidia's graphics processors at scale.
The shift does not mean OpenAI is abandoning Nvidia. The source states that OpenAI has used Nvidia GPUs for both training and deployment, and that OpenAI has been one of Nvidia's biggest customers. The new development is that Google hardware is now part of the picture for running AI products.
Why Microsoft will notice
The move sends a strategic message to Microsoft. Microsoft is OpenAI's largest investor and has provided much of the infrastructure used by OpenAI's products.
By moving some workloads onto Google infrastructure, OpenAI is showing that it can work with a direct Microsoft competitor. Google Cloud competes directly with Microsoft Azure, so this is not a neutral supplier change inside the cloud market.
The timing also matters because OpenAI CEO Sam Altman and Microsoft CEO Satya Nadella are reportedly in ongoing talks about the companies' partnership. The source article also notes that OpenAI has expanded compute capacity through a deal with Oracle.
Taken together, those facts point to a company trying to broaden its compute options. More infrastructure sources can give OpenAI flexibility as it runs products, manages costs, and negotiates with the companies that provide the capacity behind its systems.
Google's TPU business opens wider
Google originally kept its TPUs for internal use. The source article says Google is now opening them to more outside partners.
OpenAI is not the only customer named. Other customers include Apple and the startups Anthropic and Safe Superintelligence, both founded by former OpenAI executives.
Still, the OpenAI partnership has boundaries. According to The Information, Google is not giving OpenAI access to its most powerful TPU models. A Google Cloud employee confirmed that restriction.
That limit shows how carefully Google is handling access. Google can rent TPU capacity to outside companies while still keeping its strongest models of the hardware away from a high-profile customer that is also connected to a cloud rival.
What it means for Nvidia, Google Cloud, and Azure
The OpenAI decision touches more than one market. In AI chips, Google's TPU cloud goes head to head with Nvidia's GPUs, especially for running large models.
For Nvidia, the point is not just that another chip supplier exists. The point is that OpenAI, one of Nvidia's biggest customers, is now using a different type of processor at scale for parts of its work.
For Google Cloud, the deal strengthens the role of TPUs in its AI strategy. Google Cloud is described in the source article as a major growth driver for Google, and the TPU cloud is a core part of Google's plan in artificial intelligence.
For Microsoft, the significance runs through Azure and through its relationship with OpenAI. The source article notes that the cloud sector has been central to Microsoft's stock performance in recent years. If OpenAI can place more work outside Microsoft's infrastructure, that becomes relevant to both AI competition and cloud competition.
The larger compute signal
According to research firm Epoch AI, Google's infrastructure gives it the world's largest AI computing capacity. That makes Google a meaningful infrastructure option for companies trying to run large models at scale.
OpenAI's use of Google TPUs is therefore not just a technical adjustment. It is a sign that the biggest AI companies are weighing cost, capacity, hardware access, and cloud leverage at the same time.
The immediate use case is clear: OpenAI is renting Google TPUs through Google Cloud to reduce inference costs for ChatGPT and other AI products. The broader message is just as clear: OpenAI wants more than one path to the compute it needs, even when that path runs through a major Microsoft competitor.