AWS and Nvidia to Deploy 2 Million More GPUs in 2027-2028 as Demand Outpaces Prior 1-Million Commitment
AWS and Nvidia announced an expanded partnership to deploy 2 million additional GPUs across AWS data centers in 2027-2028, alongside new Vera CPU, networking, and robotics integrations.
Overview
Amazon Web Services and Nvidia announced on August 26 a major expansion of their partnership, under which AWS plans to deploy 2 million additional Nvidia GPUs across its global infrastructure in 2027 and 2028, according to a joint press release from Amazon. The companies said the expanded collaboration also spans CPUs, networking, open models, data processing and robotics.
What We Know
The new commitment covers Nvidia’s Blackwell Ultra, Rubin, and Rubin Ultra GPUs, which will “head to Amazon Web Services’ data centers in 2027 and 2028,” according to TechCrunch. The announcement came during Nvidia’s quarterly earnings call, and it builds directly on an earlier pledge: at Nvidia’s GTC 2026 conference, AWS said it would add more than 1 million Nvidia GPUs starting in 2026, and “since then, demand has exceeded those expectations,” according to the Nvidia Newsroom release.
AWS CEO Matt Garman said in the joint release that “customers want the freedom to choose the best tools for their AI workloads, and they want confidence that everything works seamlessly together,” adding that AWS has “invested deeply with NVIDIA to make AWS the best place to run NVIDIA AI technologies, optimizing performance across our infrastructure from networking and security to deployment.” Nvidia founder and CEO Jensen Huang said the companies “have built one of the great growth engines of the AI era, and demand is running ahead of every forecast,” and that after 16 years of scaling Nvidia computing together in the cloud, the partnership is now expanding “across the full stack—GPUs, CPUs, networking, open models and software.”
Beyond raw GPU volume, the companies detailed several other integrations. AWS and Nvidia are working to bring Nvidia’s Vera CPU-based infrastructure to AWS, and Nvidia’s Annapurna Labs collaboration is extending NVLink Fusion — first announced for Trainium chips at re:Invent 2025 — to work with Nvidia’s new custom high-bandwidth memory technology. The two companies also plan to build AI factories for the U.S. government that include 100,000 GPUs on secure AWS infrastructure for workloads classified at Impact Level 6 and above, per the Amazon release. On the instance side, AWS said it will expand Blackwell capacity with Nvidia RTX PRO 4500 Blackwell Server Edition GPUs for its EC2 G7 instances, which the company says deliver 4.6x the AI inference performance and 2.1x the graphics performance of the prior-generation G6 instances — making AWS, per its own release, the first major cloud provider to offer compute instances accelerated by the RTX PRO 4500.
The partnership extends to data infrastructure and robotics as well. AWS said GPU-accelerated data processing on Amazon EMR using Nvidia’s cuDF library delivers up to 3.7x faster processing and 30% better price-performance than CPU-based configurations, while GPU-accelerated vector indexing on Amazon OpenSearch Service delivers up to 9x faster indexing at a quarter of the cost. Nvidia’s Nemotron family of open models will continue to be available on Amazon Bedrock and SageMaker, and Amazon Robotics is adopting Nvidia’s physical AI platform — including the Jetson, Omniverse, and Isaac tools — for simulation, synthetic data generation, and robot training.
Nvidia CFO Colette Kress said on the earnings call that the company will also send an unspecified number of Vera CPUs to AWS, “some integrated with Rubin, others standalone,” and 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,” which she named as Oracle and SpaceXAI, according to TechCrunch. Nvidia separately reported second-quarter sales of $96.2 billion, with data center revenue of $89 billion, up 117% from a year earlier, and guided to $108 billion in third-quarter revenue.
What We Don’t Know
Neither company disclosed the financial terms of the expanded deal. TechCrunch noted that, based on typical GPU unit costs, the agreement is likely worth tens of billions of dollars, though that figure is the outlet’s own estimate rather than a number confirmed by AWS or Nvidia. It is also not yet clear how the new Nvidia commitment will interact with AWS’s own expanding custom-silicon roadmap — Amazon has said its custom chip business, which includes the Trainium and Graviton lines, crossed a $25 billion annualized revenue run rate on the strength of $225 billion in commitments from AI labs including Anthropic and OpenAI, per TechCrunch’s report of Amazon’s prior earnings call.
Analysis
The deal underscores how quickly hyperscalers’ AI infrastructure commitments are compounding: a pledge of “more than 1 million” GPUs made at GTC 2026 was followed, within months, by a further 2-million-unit commitment for 2027 and 2028 — a pattern Nvidia and AWS both attributed directly to demand running ahead of their own forecasts. The expansion is also notable for extending well beyond raw chip volume into CPUs, memory architecture, government infrastructure, and robotics, suggesting both companies see the relationship as a full-stack platform play rather than a series of one-off GPU orders, even as AWS continues to build out competing in-house silicon.