Amazon expands Nvidia chip partnership for AWS AI infrastructure

Traders Union
All News
Financial News
Amazon expands Nvidia AI partnership

Highlights

Amazon and Nvidia expanded their partnership, with AWS ordering over 2 million additional Nvidia GPUs for data centers by 2027–2028, exceeding previous demand forecasts.
Nvidia networking hardware, AI models, CPUs, software, and robotics platforms will be integrated across AWS, while Amazon's Bedrock and SageMaker will offer Nvidia Nemotron open models.
Nvidia reported second-quarter sales of $96.2 billion, driven by data center revenue up 117% to $89 billion, and has committed $279 billion to supply and manufacturing for AI projects.

AWS scales orders as AI demand rises

As reported by TechCrunch, citing Nvidia, the companies announce an expanded partnership that adds another 2 million Nvidia GPUs to Amazon Web Services data centers after Amazon had already agreed five months earlier to deploy more than 1 million GPUs starting this year. Nvidia says demand has exceeded those earlier expectations, with startups, enterprises, AI labs and governments all contributing to stronger appetite for AI computing capacity.

The new deployment includes Blackwell Ultra, Rubin and Rubin Ultra GPUs, which are built for training and running AI models, and the chips are scheduled for AWS data centers in 2027 and 2028. Financial terms are not disclosed, but the scale of the order suggests a deal worth tens of billions of dollars based on GPU pricing.

Nvidia CFO Colette Kress says the company also plans to send an unspecified number of Vera CPUs, with some integrated with Rubin and others delivered as standalone products. The announcement comes during Nvidia's quarterly earnings call, where investors are also watching for signs that next-generation Rubin products can sustain the company's growth trajectory.

Partnership widens beyond chips

The agreement extends beyond semiconductor supply even as Amazon continues investing in its own custom AI processors. Nvidia says its networking hardware, open models, CPUs, data processing software and robotics platform will be integrated across AWS, while Amazon's Bedrock and SageMaker services will offer Nvidia's Nemotron family of open models.

Amazon has been building Trainium chips as an alternative to Nvidia hardware for deep learning workloads and has promoted Graviton as a challenger in the server CPU market. On its latest earnings call, Amazon says its custom chip business reached a $25 billion annualized revenue run rate, supported by $225 billion in total commitments from AI labs including Anthropic and OpenAI.

The partnership also reaches Amazon's warehouse automation operations. Kress says Amazon plans to adopt Nvidia's physical AI stack for robots, including Omniverse, Cosmos, Isaac and Jetson, widening Nvidia's role inside Amazon beyond cloud infrastructure.

Nvidia separately reports second-quarter sales of $96.2 billion, above analyst estimates, with data center revenue rising 117% from a year earlier to $89 billion. The chipmaker expects third-quarter revenue of $108 billion and says it has committed $279 billion to secure supply and manufacturing capacity for current and future data center projects as industry spending on AI infrastructure accelerates.

In our earlier coverage of Nvidia’s upcoming fiscal second-quarter earnings, we outlined what investors were watching for in the report—whether AI infrastructure demand remained strong enough to sustain the company’s rapid growth. We also noted close attention on the Vera Rubin product ramp, potential supply constraints and cost pressures, alongside intensifying competition in AI chips.

This material may contain third-party opinions, none of the data and information on this webpage constitutes investment advice according to our Disclaimer. While we adhere to strict Editorial Integrity, this post may contain references to products from our partners.