Rotating Data
Home/Deal Room/AWS and NVIDIA plan two-million-GPU AI expansion

Cloud capacity · Evidence brief

AWS and NVIDIA plan two-million-GPU AI expansion

Amazon Web Services + NVIDIA

Primary sourcePress release only2026-08-26
Deal structureGlobal AI-compute and AI-factory capacity expansion
GeographyGlobal, including United States government infrastructure
Infrastructure layersGPU cloud capacity + AI factories + networking + federal infrastructure
Disclosed scale2 million additional NVIDIA GPUs planned across AWS global infrastructure in 2027-2028, including 100,000 GPUs for secure U.S. government AI factories

Why it matters

The strategic read

Commits the world's largest cloud platform to an additional two million NVIDIA accelerators on top of the more than one million GPUs AWS announced earlier in 2026. The plan extends beyond accelerator purchases into AI factories, Spectrum networking, Vera CPUs, NVLink Fusion and custom high-bandwidth memory, making it a full-stack capacity expansion with a dedicated 100,000-GPU federal and national-security tranche.

Source-supported terms

What was disclosed

  1. 01

    AWS plans to deploy 2 million additional NVIDIA GPUs across its global infrastructure during 2027 and 2028.

  2. 02

    The planned fleet includes NVIDIA Blackwell Ultra, Rubin and Rubin Ultra accelerators and is incremental to AWS's earlier 2026 announcement covering more than 1 million GPUs.

  3. 03

    AWS and NVIDIA plan a separate 100,000-GPU deployment on secure AWS infrastructure for federal and national-security workloads at Impact Level 6 and above.

  4. 04

    The expanded collaboration includes NVIDIA Spectrum networking, Vera CPU infrastructure, NVLink Fusion and NVIDIA custom high-bandwidth memory for future Trainium systems.

  5. 05

    The release does not disclose purchase price, delivery cadence by region, exact data-center sites, associated power capacity or binding procurement documents.

Disclosure boundary

What is known—and what is not

Strategic infrastructure expansion announced jointly by AWS and NVIDIA in an NVIDIA issuer release. The companies identify the additional two-million-GPU deployment target, 2027-2028 timing, accelerator generations, global AWS scope, 100,000-GPU federal tranche and related CPU, networking, interconnect and memory work. Public materials do not disclose a contract value, purchase orders, binding minimum volumes, regional allocations, site-level power or construction commitments, delivery milestones, cancellation rights or remedies; execution therefore remains a forward deployment plan rather than a completed procurement.