Amazon and Nvidia announced an expanded agreement that will place an additional 2 million Nvidia GPU chips into Amazon Web Services data centers in 2027 and 2028. The GPUs named for the deployment include Nvidia Blackwell Ultra, Rubin, and Rubin Ultra models. The announcement was made during Nvidia’s quarterly earnings call.
Scale and commercial details
The new order follows a pact made five months earlier for more than 1 million Nvidia GPUs across AWS starting this year. Nvidia said that since that deal, “demand has exceeded those expectations.” Neither company disclosed financial terms. Reporting noted that, given typical GPU unit costs, the purchase could amount to tens of billions of dollars, though the exact return for Nvidia was not specified.
Technology and integration
Nvidia said its networking hardware that links thousands of GPUs, its open models, CPUs, data-processing software and robotics platforms will be integrated across AWS. Nvidia CFO Colette Kress said the company also plans to deliver an unspecified number of Vera CPUs, “some integrated with Rubin, others standalone.” CEO Jensen Huang said on the call that AI is generating “profitable tokens,” and argued that more compute would produce more profit for related services.
The partnership extends beyond server chips: Nvidia products will be used in Amazon’s warehouse robotics stack, including Omniverse, Cosmos, Isaac and Jetson, and Nvidia said it is introducing a new Jetson variant for entry-level edge robotics. On the enterprise side, Nvidia’s Nemotron family of open models will be offered on AWS services Amazon Bedrock and SageMaker.
The announcement comes as Amazon continues to develop its own silicon. Amazon’s AI chief Peter DeSantis has said AWS is in talks to sell Trainium chips to other data centers, and Amazon’s Arm-based Graviton CPUs remain an alternative to traditional server processors. Amazon has reported its custom chip business reached a $25 billion annualized revenue run rate, supported by about $225 billion in commitments from AI labs including Anthropic and OpenAI.
Nvidia reported $96.2 billion in sales for the second quarter, with data center revenue of $89 billion (up 117% year over year), and projected $108 billion in third-quarter revenue. Nvidia also said production shipments of Rubin began this quarter and that it has committed $279 billion to secure supply and manufacturing capacity for current and future data-center projects.
Original source: TechCrunch AI