Enterprise

Amazon Orders 2 Million More Nvidia GPUs Through 2028

The expanded partnership triples AWS's GPU commitment while Amazon's own chip business hits $25 billion run rate.

Omega Editorial· September 11, 2026· 3 min read

Amazon Web Services has committed to purchasing an additional 2 million Nvidia GPU processors for delivery in 2027 and 2028, according to details first reported by AI Watch. The August 26 announcement expands an earlier agreement that called for more than 1 million Nvidia GPUs starting in 2026, bringing Amazon's total GPU order to more than 3 million units over the next several years.

The new processors will include Nvidia's next-generation Blackwell Ultra, Rubin, and Rubin Ultra architectures. Beyond the hardware purchase, the companies described the arrangement as a broader infrastructure partnership spanning AI factories, CPUs, networking infrastructure, open models, data processing, and robotics.

Why it matters

The scale of this commitment—tripling Amazon's GPU order in five months—illustrates how rapidly cloud providers are locking in chip supply years ahead of delivery. For Nvidia, Amazon represents a critical anchor customer as the chipmaker targets more than $1 trillion in Blackwell and Rubin revenue through 2027. For AWS, securing this compute capacity is essential to maintaining its position in the intensifying competition to provide AI infrastructure to enterprise customers. The question investors will watch: whether this massive capital outlay translates into corresponding AI-driven cloud revenue growth when the chips arrive in 2027 and 2028.

Amazon's parallel chip strategy

The expanded Nvidia order arrives alongside significant momentum in Amazon's own semiconductor business. The company's custom silicon portfolio—including Trainium AI accelerators, Graviton CPUs, and Nitro networking chips—has reached a $25 billion annualized revenue run rate, according to AI Watch.

Amazon has also secured more than $225 billion in revenue commitments for its Trainium chips, with major multi-year deals from Anthropic and OpenAI. This dual approach allows AWS to offer customers both industry-standard Nvidia GPUs and proprietary alternatives optimized for specific workloads.

Institutional backing grows

Hedge fund ownership in both companies increased during the second quarter of 2024. The number of hedge funds holding Amazon positions rose from 353 to 369, with Arrow Street Capital holding the largest stake at $9.6 billion. Nvidia saw similar growth, with hedge fund holders increasing from 275 to 285 and Fisher Asset Management holding an $18.1 billion position.

The institutional interest reflects confidence that both companies are well-positioned in the AI infrastructure buildout, even as questions remain about when and how that spending will generate returns. With delivery timelines extending into 2028, the market will have several years to evaluate whether the current wave of GPU orders matches actual enterprise AI adoption.

These details were first reported by AI Watch.

#amazon web services#nvidia#gpu#ai infrastructure#cloud computing#custom silicon

This is an original analysis by the Omega editorial team. Source reporting: AI Watch.

Want systems like this working for your business?

Book a Call

More in Enterprise

Enterprise· 3 min read

Salesforce unveils Enterprise AI Harness and AI Control Plane

Two new offerings aim to improve reliability, security, and governance for organizations deploying AI agents at scale.

Via AI Watch · Sep 11, 2026
Enterprise· 3 min read

Amazon Quick AI Assistant Adds Mobile Activity Feed

The workplace productivity tool now syncs across desktop and mobile, with cloud-based agents that run even when devices are closed.

Via AI Watch · Sep 10, 2026
Enterprise· 2 min read

GitHub Adds REST APIs for AI Scan Pull Request Configuration

Organization and repository-level endpoints let Advanced Security customers programmatically manage AI-powered code scanning across their repositories.

Via AI Watch · Sep 10, 2026