We are the Frontier
World

Amazon Committed to 3 Million Nvidia GPUs by 2029

The massive hardware acquisition aims to address persistent capacity shortfalls in global AI infrastructure through 2028.

Updated on Sept. 21, 2026 in Data Centers

Isometric editorial illustration of a dense, orderly stack of silicon processor chips in muted teals and oxblood tones.
Amazon has committed to acquiring 3 million Nvidia GPUs by 2029 to bolster its cloud infrastructure against persistent global AI capacity shortfalls. AI Illustration. Upload story photo >

Live Poll

Is now a good time for major tech companies to invest heavily in AI infrastructure?

Amazon announced a commitment on August 26, 2026, to purchase 3 million Nvidia GPUs by 2029. This includes a previously announced order of 1 million units and a newly finalized deal for 2 million additional Blackwell Ultra, Rubin, and Rubin Ultra processors.

Why it matters

The investment represents an attempt to bridge the gap between rapidly rising customer demand for AI compute and currently constrained availability. Despite 37 percent year-over-year revenue growth in AWS during the second quarter of 2026, the company expects capacity to remain below demand through 2027.

AWS will deploy 2 million Blackwell Ultra, Rubin, and Rubin Ultra units across its global footprint in 2027 and 2028. Shipments for this phase are scheduled to begin in the third quarter of 2026, with 100,000 units specifically reserved for U.S. government workloads at Impact Level 6.

The players

Amazon

A global cloud computing and e-commerce giant that operates AWS, the world's largest provider of on-demand cloud infrastructure.

Nvidia

The dominant designer of graphics processing units and data center accelerators that currently define the hardware standard for AI compute.

The details

The rollout utilizes Nvidia's high-performance architecture, including the Blackwell Ultra and future Rubin-series chips designed for large-scale generative AI training and inference. These GPUs function as parallel processing engines that accelerate matrix multiplication, the fundamental mathematical operation required for deep learning workloads. AWS integrates this hardware into its global cloud infrastructure to scale capacity for customers, a necessary move following the purchase of 196,000 Hopper-generation GPUs in 2024.

Timeline

  1. 2024: Amazon purchased 196,000 Hopper GPUs.

  2. March 2026: AWS committed to over 1 million Nvidia GPUs.

  3. August 26, 2026: Amazon announced the 2 million GPU deal.

  4. Third quarter 2026: Shipments of the new GPUs commence.

  5. 2027-2028: AWS will deploy the 2 million additional units.

The Tech Race

The transition from 2024 Hopper-based infrastructure to the Blackwell and Rubin architectures defines the current arms race for AI dominance in the cloud. This massive capacity ramp reflects Amazon's strategy to maintain its cloud leadership as the industry shifts toward significantly more compute-intensive generative AI models.

Enterprise and government customers requiring high-scale AI infrastructure will see increased access to compute capacity beginning in 2027. Smaller-scale users may see eventual cost efficiencies as AWS broadens its availability of specialized hardware for machine learning workflows.

The takeaway

Amazon is prioritizing large-scale hardware procurement to avoid bottlenecks that could stall the growth of its cloud AI services. Observers should track mid-2028 deployment benchmarks to confirm whether this massive capacity influx successfully meets projected demand for generative AI.

Further reading

For broader trends in infrastructure, visit Data Centers.

Source note: This article includes information reported by TechRadar.

Live Poll

Is now a good time for major tech companies to invest heavily in AI infrastructure?