Amazon Web Services announced a massive expansion of its partnership with Nvidia on Wednesday, committing to deploy an additional 2 million GPU chips across its global data center network over the next two years. The deal, revealed during Nvidia’s quarterly earnings call, effectively triples Amazon’s previous GPU order and signals that enterprise demand for AI compute continues to outstrip even the most aggressive supply projections.
The expanded commitment covers Nvidia’s next-generation Blackwell Ultra, Rubin, and Rubin Ultra GPUs, all scheduled for deployment in AWS data centers during 2027 and 2028. While neither company disclosed exact financial terms, industry analysts estimate the order is worth tens of billions of dollars based on current GPU unit costs. The deal comes just five months after Amazon agreed to deploy more than 1 million Nvidia GPUs across AWS infrastructure starting this year.
Surging Demand Forces Faster Expansion
Nvidia said in a statement that since the original 1 million GPU agreement was announced at GTC 2026 in March, “demand has exceeded those expectations.” The rapid doubling-down reflects a broader industry trend: generative AI workloads, large language model training, and enterprise AI adoption have created a sustained crunch for high-end compute capacity that shows no sign of easing.
For AWS, the world’s largest cloud provider, securing such a massive allocation of Nvidia’s most sought-after chips serves a dual purpose. It guarantees capacity for its own customers who need scalable GPU clusters for AI development, and it tightens the global supply of Nvidia hardware, potentially making it harder for competitors like Microsoft Azure and Google Cloud to secure comparable orders.
More Than Just Chips
The expanded partnership extends well beyond a simple buyer-supplier transaction. Nvidia’s full technology stack will be integrated across AWS, including networking hardware that connects thousands of GPUs into unified systems, open models, CPUs, data processing software, and the company’s robotics platform encompassing Omniverse, Cosmos, Isaac, and Jetson.
This deeper integration means AWS customers will have access to a more tightly coupled Nvidia ecosystem within the cloud, potentially improving performance and ease of deployment for AI workloads that depend on coordinated hardware and software components.
A Hybrid Strategy Under Pressure
The Nvidia deal stands in tension with Amazon’s own aggressive chip-building program. Amazon’s custom silicon business, which includes Trainium AI chips and Graviton Arm-based CPUs, crossed a $25 billion annualized revenue run rate on its last earnings call, driven by $225 billion in total commitments from AI labs including Anthropic and OpenAI.
Amazon’s AI chief Peter DeSantis has said AWS is in talks to sell its Trainium chips directly to external data centers, and the company has positioned Trainium as a direct alternative to Nvidia’s H100 and Blackwell chips for deep learning workloads. Yet the tripling of the Nvidia order suggests Amazon sees custom silicon and Nvidia GPUs as complementary rather than competing paths.
The Economics of AI Compute
The sheer scale of Amazon’s Nvidia commitment underscores how central GPU infrastructure has become to the economics of AI. As Jensen Huang, Nvidia’s CEO, noted during the earnings call, “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.”
That logic is driving an industry-wide infrastructure arms race. Hyperscalers are pouring hundreds of billions into next-generation data centers, and the companies that control the most compute capacity are positioned to set the pace of AI innovation. Amazon’s 2-million-GPU expansion raises the stakes for every major cloud provider.
Implications for the AI Arms Race
The deal also reflects how the AI infrastructure market has evolved beyond simple GPU procurement. Nvidia’s platform now includes networking, software, and systems integration, making it not just a chip vendor but a full-stack infrastructure partner. For Amazon, betting heavily on Nvidia while simultaneously building its own chips is a pragmatic acknowledgment that the AI compute shortage is real and likely to persist through the end of the decade.
As AI models grow larger and inference costs from millions of daily users continue to climb, access to GPU power has become the defining competitive advantage in cloud computing. Amazon’s latest Nvidia deal ensures AWS will have the inventory to meet demand while keeping pressure on rivals to match the scale of its investment.
discussion