TechCapsules

AI-summarised brief · reviewed before publication

OpenAI has recently started renting Google's artificial intelligence chips to power its products, including ChatGPT, according to a source close to the matter. The company, one of the largest purchasers of Nvidia's graphics processing units (GPUs), uses these AI chips to train models and for inference computing, a process in which an AI model uses its trained knowledge to make predictions or decisions based on new information. OpenAI's decision to rent Google's AI chips marks a significant shift away from relying on Nvidia's GPUs and Microsoft's data centers. This move is part of OpenAI's plan to add Google Cloud service to meet its growing needs for computing capacity, which was exclusively reported earlier this month. Google, on the other hand, is expanding the external availability of its in-house tensor processing units (TPUs), which were historically reserved for internal use. The company has already won customers such as Apple, Anthropic, and Safe Superintelligence, two startups launched by former OpenAI leaders. The move to rent Google's TPUs is significant for OpenAI as it marks the first time the company has used non-Nvidia chips meaningfully. According to The Information, this could potentially boost TPUs as a cheaper alternative to Nvidia's GPUs. OpenAI hopes that the TPUs, which it rents through Google Cloud, will help lower the cost of inference. However, Google is not renting its most powerful TPUs to OpenAI, a rival in the AI race, according to a Google Cloud employee. Google declined to comment on the matter, while OpenAI did not immediately respond when contacted.