The Amazon logo. [Reuters]
The Amazon logo. [Reuters]

Nvidia will supply an additional 2 million GPUs to Amazon Web Services' global infrastructure between 2027 and 2028.

Nvidia said Wednesday (local time) that it would "significantly expand strategic cooperation to meet surging global demand for AI infrastructure."

Amazon had earlier said it would install 1 million Nvidia chips in AWS data centers starting this year; the newly announced order comes on top of that commitment.

The chips covered by the deal include Nvidia's Blackwell Ultra, Rubin and Rubin Ultra GPUs.

Neither Amazon nor Nvidia disclosed the financial terms, but each chip carries a list price of at least tens of thousands of dollars, putting the total at an estimated tens of billions of dollars. Discounts typically apply for large customers.

Nvidia GPUs have become a cornerstone of the AI boom, underpinning the development and operation of most large language models.

Amazon, Microsoft and Google — all Nvidia customers — are simultaneously developing their own chips to compete with Nvidia. These hyperscalers believe that building hardware optimized for their own data centers will allow them to pass efficiency gains on to customers.

Amazon has invested heavily in its in-house semiconductor division, Annapurna Labs, for years. The unit produces AI accelerators designed to rival Nvidia GPUs, alongside central processing units.

Amazon has said it plans to incorporate Nvidia's networking technology into the next version of its AI chip, Trainium.


yckim6452@heraldcorp.com