Amazon to Deploy 2 Million Nvidia GPUs in AWS Data Centers by 2028

Amazon Web Services announced a major hardware expansion that will see 2 million Nvidia graphics processing units installed across its data centers between 2027 and 2028, a move aimed at bolstering the cloud provider's capacity for artificial‑intelligence workloads.

The new purchase builds on a prior commitment in which AWS secured more than one million Nvidia chips, underscoring the growing reliance of cloud operators on specialised accelerators to meet the computational demands of generative AI, machine‑learning model training and inference services.

While GPUs will form the backbone of the upgrade, Amazon is also preparing to roll out its own Vera‑based central processing units. The Vera line, a successor to the company’s earlier Graviton processors, is designed to complement the GPU fleet by handling the increasingly complex data‑pre‑processing and orchestration tasks that AI applications require.

Industry analysts note that the scale of the acquisition positions AWS as one of the largest single‑customer buyers of Nvidia hardware to date. Nvidia’s GPUs have become the de‑facto standard for high‑performance AI, and the partnership reflects a broader trend of cloud providers locking in supply chains to guarantee performance and pricing stability as demand spikes.

Amazon’s investment comes at a time when competition among the big cloud players—Microsoft Azure, Google Cloud, and Alibaba Cloud—has intensified around AI services. By securing a massive inventory of GPUs and pairing them with custom silicon, AWS hopes to offer lower‑cost, higher‑throughput AI compute options that could attract enterprises seeking to run large language models or advanced analytics in the cloud.

From a strategic perspective, the expansion also serves to diversify Amazon’s hardware portfolio. Relying solely on third‑party accelerators can expose a provider to supply constraints, especially as Nvidia grapples with semiconductor shortages and the need to meet the needs of multiple industry sectors.

Looking ahead, the deployment timeline suggests that the bulk of the new GPUs will become operational by the end of 2028. Customers can expect incremental availability of upgraded instances as data centers are retrofitted, with Amazon likely announcing pricing and performance benchmarks closer to rollout.

Stakeholders will be watching how the Vera CPUs integrate with the Nvidia GPUs, as the synergy between custom CPUs and external accelerators could set a new standard for AI‑focused cloud infrastructure. If successful, the move may accelerate the adoption of large‑scale AI services across a broader range of industries, reinforcing AWS’s position as a leading cloud platform.