Google and Oracle will offer Google's custom AI chips to Oracle cloud customers next year, giving Nvidia more competition and AI buyers more choice.
Google's custom AI chips are about to leave Google's own data centers. On Tuesday, Google and Oracle announced that Oracle cloud customers will be able to use Google's chips starting next year. The companies said it is the first time these chips have been offered on another big cloud company's platform.
The chips are called TPUs, short for Tensor Processing Units. Google began building them more than a decade ago to power search and other products. Earlier this year, Nvidia CEO Jensen Huang downplayed the threat, saying TPU growth was "100% Anthropic," the AI company behind the Claude assistant. The Oracle deal shows demand reaching beyond one customer.
Anthropic itself is a huge buyer. In its IPO paperwork filed this week, it said it plans to spend $114.5 billion on Google computing from April 2026 to July 2033. It also uses chips from Amazon, AMD and Nvidia, so it is spreading its bets.
OpenAI, the maker of ChatGPT, has a job listing for a software engineer to optimize its models for TPUs. The Information also reported Tuesday that OpenAI plans to buy TPUs from Google. OpenAI and Nvidia did not respond to requests for comment, and Google declined to comment.
These are analyst forecasts, not results. Nvidia is still far ahead. It reported $96.2 billion in revenue in its previous quarter alone, and its share price stayed flat on Wednesday. Some analysts think the market is underestimating Google.
Most people will never buy an AI chip. But the cost of running AI shows up in the price of the tools you use or pay for. If more companies mix chips from several suppliers, it could weaken Nvidia's ability to set prices. Over time, that competition could help keep AI costs in check.
That is a possibility, not a promise. Anthropic did not say how much of its Google spending will go to TPUs specifically. Nothing in the reporting says prices for AI tools will fall, or when.
More companies are buying AI chips from Google as well as Nvidia, and that competition could eventually help keep the cost of AI tools down.
Visa tested 22 AI models on realistic shopping tasks. They passed just over half, and often bought the wrong thing anyway.
Read more: Visa tested 22 AI shoppers, and most still make costly mistakesDARPA wants AI to hunt for design flaws in stock, bond, energy and cloud markets, so enemies cannot use them to cause economic damage that reaches households.
Read more: Pentagon wants AI to find weak spots in stock and energy marketsTuring Award winner Richard Sutton says more computing power alone will not make AI truly smart. Big AI companies disagree, which matters for anyone betting on AI.
Read more: AI pioneer says piling on computing power won't make machines smart
