Amazon is sharply expanding its use of Nvidia hardware and software as demand for AI computing capacity continues to climb. The companies announced Wednesday that Amazon will add another 2 million Nvidia GPU chips to its data centers, deepening a partnership that had already grown significantly this year.
The chips are headed to Amazon Web Services data centers in 2027 and 2028. They include Nvidia Blackwell Ultra, Rubin, and Rubin Ultra GPUs, which are built for the compute-heavy work of training and running AI models.
A bigger AWS commitment to Nvidia GPUs
The new order follows an earlier Amazon commitment, made five months ago, to deploy more than 1 million Nvidia GPUs across AWS infrastructure starting this year. Nvidia said that since then, “demand has exceeded those expectations.”
Neither Amazon nor Nvidia disclosed the financial terms of the expanded deal. The source article notes that, given the cost of GPU units, the agreement is worth tens of billions of dollars.
The size of the order is one reason the announcement stands out. The speed of the expansion is another. In a short span, Amazon moved from a major Nvidia deployment to an even larger commitment, suggesting that AWS sees continued pressure from customers that need more AI infrastructure.
The companies pointed to “surging demand” from startups, enterprises, AI labs, and governments as a reason for moving closer together. For AWS, the practical issue is capacity: customers want access to systems powerful enough to support demanding AI workloads. For Nvidia, the deal reinforces the central role its chips continue to play in that market.
The partnership goes beyond chips
This is not only a larger hardware purchase. Nvidia said its technology will be integrated more broadly across AWS, including networking hardware, open models, CPUs, data processing software, and its robotics platform.
That matters because modern AI infrastructure depends on more than individual processors. Nvidia’s networking hardware connects thousands of GPUs into one system, allowing large workloads to run across many chips. By bringing more of Nvidia’s surrounding technology into AWS, the companies are aligning around a wider stack of AI computing tools.
Nvidia CFO Colette Kress said Nvidia also plans to send an unspecified number of Vera CPUs as part of the broader AWS expansion, with “some integrated with Rubin, others standalone.” Nvidia CEO Jensen Huang has been emphasizing Vera CPUs as a new opportunity for the company, saying in May that he had found a “brand new $200 billion TAM.”
Kress also said Wednesday that Nvidia expects Vera to be deployed by “every major hyperscaler, neocloud, AI lab, and system OEM, with shipments already underway to our lead partners,” including Oracle and SpaceX AI.
Amazon still has its own chip ambitions
The expanded Nvidia relationship comes while Amazon continues to invest in its own AI chip work. That makes the deal more complex than a simple supplier agreement.
Amazon has been building custom chips to reduce its dependence on Nvidia and to compete in parts of the AI hardware market. Amazon’s AI chief Peter DeSantis has said AWS is in talks to sell its Trainium chips to other companies for use in data centers. Trainium is described in the source as a direct alternative to Nvidia’s H100 or Blackwell chips for deep learning workloads.
Amazon also has its Arm-built Graviton CPU, which is seen as a challenger to traditional server chips from Intel and AMD. On its last earnings call, Amazon said its custom chip business crossed a $25 billion annualized revenue run rate, driven by $225 billion in total commitments from AI labs like Anthropic and OpenAI.
Even with that internal momentum, Amazon is adding millions of Nvidia GPUs to AWS. The logic is straightforward: building a competing chip business does not remove the immediate need for Nvidia hardware, especially when customers are asking for more AI compute now and over the next few years.
Robots, open models and enterprise AI
The partnership is also extending into Amazon’s warehouse robots and enterprise AI services. Kress said Amazon plans to adopt Nvidia’s full physical AI stack to power its robot fleet.
That stack includes Omniverse, Nvidia’s simulation and digital twin platform; Cosmos, its world model platform; Isaac, its robotics development platform; and Jetson, its computing hardware for robots and edge AI. Nvidia also introduced a new version of Jetson this week for “entry-level edge AI.”
On the enterprise side, AWS will serve Nvidia’s Nemotron family of open models on Amazon Bedrock and SageMaker. Bedrock is Amazon’s managed foundation model platform, while SageMaker is its managed cloud service.
For AWS customers, the result is a broader Nvidia presence across cloud infrastructure, model access, and robotics tools. For Amazon, it creates a wider set of services tied to Nvidia technology while AWS continues to develop its own chips and platforms.
Nvidia’s results show why compute demand matters
Nvidia also reported Wednesday that second-quarter sales reached $96.2 billion, beating analyst estimates. Data center revenue accounted for most of that total at $89 billion, up 117% from a year ago.
The company said it expects revenue to reach $108 billion in the third quarter, with some of that coming from its next-generation Rubin GPUs. Nvidia said it began production shipments this quarter. Investors have been watching Rubin’s initial Q3 sales for signals about whether demand will continue into Nvidia’s next hardware generation.
Nvidia has committed $279 billion to secure supply and manufacturing capacity for current and future data-center projects, up from $119 billion last quarter. That includes $92 billion in projected spending for the rest of the fiscal year and another $87 billion in fiscal year 2028.
Huang framed the moment around useful AI output, saying, “The thing that matters for the industry is that AI is now doing productive and useful work,” and “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. This is the exact phase where we're at, which is the reason why everybody's leaning in.”
The open question is whether the industry’s infrastructure spending will translate into profits as directly as companies hope. For now, Amazon’s expanded Nvidia order shows that major cloud providers are still preparing for an AI market where demand for compute keeps rising.