AWS and Nvidia announced on August 26 that AWS will deploy 2 million additional Nvidia GPUs, including Blackwell Ultra and Rubin chips, across its data centers in 2027 and 2028, doubling the scale of their prior 1 million GPU commitment.

Read the full story: AWS and Nvidia to Add 2 Million More AI GPUs by 2028 →

Transcript

AWS and Nvidia just announced AWS will deploy two million more Nvidia GPUs in 2027 and 2028. That's on top of chips already running today, and it includes three generations at once: Blackwell Ultra, Rubin, and Rubin Ultra. Here's the part that matters. This expansion happened because AWS's previous commitment, around one million GPUs, ran out early. AI compute demand keeps outrunning even aggressive supply plans. But there's a twist. The deal also gives AWS's own Trainium chips, the ones built specifically to reduce Nvidia dependence, access to Nvidia's new custom memory technology. So AWS is buying record volumes of Nvidia GPUs while also asking Nvidia to help its Nvidia-competing chips get faster. That's not a contradiction. It's coopetition. For Nvidia, it's a multi-year revenue lock-in before those chips even fully ship. For Amazon, it's a bet that AI compute demand keeps climbing through 2028.