Amazon and Nvidia are getting even closer. The two companies announced Wednesday an expanded partnership that includes a deal to add an additional 2 million Nvidia GPU chips to Amazon’s data centers.
These GPUs are designed to handle the heavy computing demands of training and running AI models and include Nvidia Blackwell Ultra, Rubin, and Rubin Ultra GPUs. The chips will be shipped to Amazon Web Services data centers in 2027 and 2028.
The announcement was made during Nvidia’s quarterly earnings call and comes just five months after Amazon agreed to deploy more than 1 million Nvidia GPUs across its AWS infrastructure starting this year. Nvidia said in a statement that “demand has exceeded expectations” since then.
The companies did not share financial terms. It’s unclear what the exact benefit to Nvidia will be. But when you factor in the unit cost of GPUs, this deal is worth tens of billions of dollars.
This announcement is notable not only for its size and speed of growth, but also because it goes beyond Amazon buying more Nvidia chips. And it’s happening even as Amazon invests in its own AI chips that could potentially compete.
Nvidia announced Wednesday that its technology, including open models, CPUs, data processing software and robotics platforms, as well as networking hardware that connects thousands of GPUs into a single system, will be integrated across AWS.
The companies said a “surge in demand” from startups, corporations, AI labs and even governments influenced their decision to work more closely together.
The expanded partnership comes as Amazon ramps up its efforts on its AI chips, particularly CPUs, the general-purpose processors that are at the heart of servers.
Amazon is developing its own chips to reduce its dependence on Nvidia and even compete with chip giants. Peter DeSantis, Amazon’s head of AI, said AWS is in talks to sell its Trainium chips, a direct replacement for Nvidia’s H100 or Blackwell chips for deep learning workloads, to other companies for use in data centers. Amazon’s Arm Graviton CPU is also seen as a challenger to traditional server chips from Intel and AMD.
Amazon said its custom chip business is growing, noting in its last earnings call that annual sales exceeded $25 billion, with commitments totaling $225 billion from AI research institutes such as Anthropic and OpenAI.
But Nvidia still seems to be the GOAT in the world of AI chips.
With Amazon adding 2 million GPU chips to AWS starting in the third quarter, Nvidia plans to ship an unspecified number of Vera CPUs, “some integrated with Rubin, some standalone,” said Nvidia CFO Colette Kress.
Nvidia CEO Jensen Huang has big plans for his company’s Vera CPUs, boasting in May that he had found “an entirely new $200 billion TAM” for the company.
Kress said Wednesday that apart from AWS, NVIDIA expects Vera to be deployed in “all major hyperscalers, neoclouds, AI labs, and system OEMs, and has already begun shipping to key partners including Oracle and SpaceXAI.”
The partnership will also extend to Amazon’s warehouse robots and enterprise products.
Kress said Amazon plans to use Nvidia’s complete physical AI stack to power its fleet of robots. The stack includes Omniverse, its simulation and digital twin platform. Cosmos (its world modeling platform); Isaac (robot development platform); Jetson (computing hardware for robots and edge AI). This week, Nvidia also announced a new version of Jetson, designed as a more accessible robotic computer for “entry-level edge AI.”
On the enterprise side, AWS offers Nvidia’s Nemotron family of open models on Amazon Bedrock, the company’s managed foundation model platform, and SageMaker, its managed cloud service.
Nvidia on Wednesday announced second-quarter revenue of $96.2 billion, beating analysts’ expectations. Data center revenue accounted for the bulk of Nvidia’s revenue in the quarter, increasing 117% year over year to $89 billion.
Nvidia said it expects third-quarter revenue to reach $108 billion, with some of that coming from its next-generation Rubin GPUs. Nvidia said it began mass production shipments this quarter. Investors will be watching Rubin’s first sales numbers for the third quarter for signs of continued demand for Nvidia’s next-generation hardware.
Nvidia has pledged $279 billion to secure supply and manufacturing capacity for current and future data center projects, up sharply from $119 billion last quarter, as the chipmaker looks to secure the memory and manufacturing capacity to meet AI demand for years to come. This commitment includes $92 billion in planned spending for the remainder of the fiscal year and an additional $87 billion in fiscal year 2028.
“What’s important for the industry is that AI is doing productive and useful work,” Huang said on a conference call Wednesday. “AI is generating profitable tokens… If we have more computational power, we can generate more profitable tokens, which in turn brings more profits to all services. This is exactly the stage we are at and that is why everyone is leaning in.”
As AI companies pour hundreds of billions of dollars into infrastructure, investors are watching to see whether the additional compute actually translates into additional profits.
If you buy through links in our articles, we may earn a small commission. This does not affect editorial independence.
