Amazon (AMZN) said Wednesday that it has completed a massive data center project and that Anthropic is set to use 1 million of its custom chips by the end of 2025.
The news comes as AI chip competition heats up among Big Tech firms. Anthropic said last week that it would use 1 million of Google’s (GOOG, GOOGL) custom chips called TPUs, or tensor processing units — a multibillion-dollar deal that had raised questions about Amazon and Anthropic’s relationship going forward.
Amazon said in a press release Wednesday that Anthropic is using 500,000 of its Trainium2 chips to run its Claude AI models as part of the tech giant’s Project Rainier, a massive AI compute cluster spread across multiple data centers in Mississippi and Indiana. AI compute clusters link thousands of specialized chips across servers — sometimes even across data centers — so they can work together as a single system capable of training and running AI models.
In total, Anthropic’s models will scale to use more than 1 million of Amazon’s Trainium2 chips from Project Rainier and other Amazon data centers this year, the tech giant said.
“It’s a big deal for Amazon because there’s this market perception and fear around … whether or not they have the same infrastructure AI advantage Google has,” Hedgeye Risk Management analyst Andrew Freedman said. He added that the deal likely represents multiple billions of dollars in annual revenue to AWS.
An Amazon spokesperson declined to comment on the estimate.
Meanwhile, Amazon’s and Google’s expanding partnerships with AI developers illustrate rising competitive pressures on Nvidia (NVDA), which has the biggest share of the AI chip market. Leading AI cloud providers, including Amazon and Google, have ramped up their in-house chip efforts rather than relying on Nvidia.
“Nvidia — it’s the place where most customers start today when they think about AI,” Amazon Web Services vice president of compute and machine learning Dave Brown told Yahoo Finance. “But there’s a growing number of chips out there, Trainium being one of them, that are competing for that workload. And I think it’s very important that we have an environment where customers get to choose.”
The company said Project Rainier, which is one of the biggest AI compute clusters in the world and represents the largest deployment of Amazon’s custom chips to date, is now fully operational, less than one year after it was first announced in December 2024.
“We’re going to keep scaling for many years to come,” Brown said.
