Amazon triples its Nvidia GPU order amid 'surging demand'

Amazon and Nvidia announced Wednesday an expanded partnership that includes a deal to add another 2 million Nvidia GPU chips to Amazon’s data centers. The GPUs — Nvidia Blackwell Ultra, Rubin, and Rubin Ultra chips built to handle training and running AI models — will ship to Amazon Web Services data centers in 2027 and 2028.

The announcement, made during Nvidia’s quarterly earnings call, comes just five months after Amazon agreed to deploy more than 1 million Nvidia GPUs across AWS starting this year; Nvidia said demand has since “exceeded those expectations.” Neither company disclosed financial terms, but considering typical GPU unit costs, the deal is worth tens of billions of dollars. Beyond chips, Nvidia’s networking hardware, open models, CPUs, data-processing software, and robotics platform will also be integrated across AWS, with “surging demand” from startups, enterprises, AI labs, and governments cited as the driver.

The expanded deal comes even as Amazon builds its own AI chips to reduce its Nvidia dependence — including Trainium chips, a direct alternative to Nvidia’s H100 or Blackwell chips, which Amazon’s AI chief Peter DeSantis has said AWS is in talks to sell to other companies for use in data centers, and its Arm-built Graviton CPU. Amazon said its custom chip business has crossed a $25 billion annualized revenue run rate, driven by $225 billion in total commitments from AI labs including Anthropic and OpenAI.

Nvidia also plans to send an unspecified number of Vera CPUs — “some integrated with Rubin, others standalone,” according to Nvidia CFO Colette Kress — with Vera expected to be deployed by every major hyperscaler, neocloud, AI lab, and system OEM, including Oracle and SpaceXAI. Nvidia CEO Jensen Huang has pointed to a “brand-new $200 billion TAM” for Vera. The partnership also extends to Amazon’s warehouse robots, with AWS adopting Nvidia’s physical-AI stack — Omniverse, Cosmos, Isaac, and Jetson — plus serving Nvidia’s Nemotron open models on Amazon Bedrock and SageMaker.

Nvidia reported $96.2 billion in second-quarter sales, beating analyst estimates, with data-center revenue at $89 billion, up 117% from a year earlier; it expects $108 billion in third-quarter revenue, including early Rubin GPU shipments. The company has now committed $279 billion to secure supply and manufacturing capacity, up from $119 billion last quarter, including $92 billion in planned spending for the rest of this fiscal year and $87 billion in fiscal year 2028.

“AI is generating profitable tokens … If we had more compute, we could generate more profitable tokens, which results in more profit for all of the services,” Huang said on the earnings call.

Why it matters

The scale of Nvidia’s new supply commitments shows the chipmaker betting that AI infrastructure demand keeps compounding, even as a major customer like Amazon builds competing in-house silicon. Investors will be watching whether that additional compute keeps translating into profit, as Huang argues, or whether the industry is over-building ahead of demand.