Amazon Web Services and Nvidia expand AI infrastructure deal, adding 2 million GPUs

Amazon Web Services and Nvidia said they are expanding their long-running partnership to meet what they describe as rising demand for AI infrastructure. In an announcement reported by Yahoo Finance, the companies outlined plans to deploy 2 million additional Nvidia GPUs through AWS, adding to the accelerated computing capacity available to customers building and running AI models.

The expanded commitment is framed as a response to growth in usage of AI training and inference systems, which rely on GPUs, or graphics processing units, to accelerate the heavy computations that large machine learning models require. By adding more GPUs into AWS capacity, the companies are effectively increasing the supply of compute resources that developers, enterprises, and other organizations can access via cloud services.

For Nvidia, the agreement reinforces its role as a central supplier of AI hardware, particularly its data center GPUs used in training and running artificial intelligence applications. For AWS, it strengthens the company’s ability to support customers that want to move from experimentation to production workloads without waiting for enough specialized accelerator capacity.

While the announcement points to the scale of the planned deployment, it does not, in the information available here, specify the distribution details such as where in AWS regions the GPUs will be allocated, whether they will be tied to particular Nvidia GPU generations, or how quickly the incremental capacity will come online.

The message also lands in a broader market context where cloud providers are racing to expand AI-ready infrastructure and customers are prioritizing performance and availability as demand for generative AI and other AI workloads increases. Partnerships that secure large GPU volumes can be a differentiator because GPUs are still a constrained resource compared with standard CPU-based computing.

Even with the headline figure, several practical questions remain unanswered in the reported announcement. The disclosed materials in the provided coverage do not indicate total contract value, whether the 2 million GPUs are already committed through specific purchase agreements or represent a capacity plan, what portion of the added hardware is intended for training versus inference, or any expected timeline for deployment milestones.

Nvidia and AWS did not provide additional deal terms in the information captured here, so observers will likely focus next on subsequent disclosures such as more detail about the hardware configuration, deployment schedule, and how the capacity expansion translates into measurable changes for customers. For now, the clearest takeaway is that both companies are aligning around a large, near-term scaling effort for GPU-backed AI compute in the cloud.