AWS to Add 2 Million Nvidia GPUs in Massive AI Infrastructure Expansion
Amazon is scaling its cloud computing power with a multi-million chip commitment to maintain its lead in generative AI infrastructure.
Amazon Web Services (AWS) and Nvidia have significantly expanded their long-standing partnership to aggressively scale AI computing capacity. The move ensures the world's largest cloud provider remains equipped to handle the surging demand for generative AI workloads.
As part of the expansion, AWS plans to integrate 2 million additional Nvidia GPUs into its global infrastructure between 2027 and 2028. This massive procurement follows a previous commitment to add more than one million chips starting in 2026. Beyond raw processing power, the deal includes the integration of Vera-based CPUs and critical networking upgrades, specifically NVLink Fusion, to optimize data transfer between chips.
Diversifying AI Capabilities
The partnership extends beyond data center hardware into specialized software and automation. AWS is integrating Nvidia's cuDF for accelerated data science and deploying advanced robotics simulation tools designed to enhance the efficiency of Amazon's warehouse robots. Furthermore, the scale of the rollout includes a strategic carve-out, with 100,000 chips from this expansion specifically reserved for government and defense computing requirements.
The Infrastructure Arms Race
This expansion underscores the critical dependency cloud providers have on Nvidia's hardware to power the current AI boom. By securing millions of GPUs years in advance, AWS is attempting to hedge against supply chain volatility and ensure that its enterprise customers have the necessary compute to train and deploy large language models. In a market where compute capacity is the primary currency of AI development, this scale of investment creates a significant barrier to entry for smaller cloud competitors.
Future Outlook
Industry observers will now watch how AWS integrates these millions of chips to lower the cost of AI inference for end-users. While the hardware roadmap is clear through 2028, it remains to be seen how this massive influx of Nvidia silicon will coexist with Amazon's own in-house chip development efforts, such as the Trainium and Inferentia lines.