AWS and Nvidia Expand Partnership with 2 Million Additional GPUs

Amazon Web Services and Nvidia plan to deploy 2 million additional Nvidia GPUs across the global AWS infrastructure during 2027 and 2028. [3] This deployment will include Nvidia's Blackwell Ultra, Rubin, and Rubin Ultra graphics processing units. [1] These additions will build upon an earlier commitment of 1 million GPUs that the companies announced in March. [1] The expanded capacity aims to serve frontier labs, global enterprises, startups, and governments as they scale AI workloads. [3] AWS and Nvidia have a history of joint innovation spanning 16 years, and the companies are expanding this strategic collaboration to meet rising global demand for AI infrastructure. [3]

The expanded partnership brings co-engineered AI solutions to AWS customers, including the introduction of Nvidia Vera CPUs and advanced networking capabilities. [3] The infrastructure is designed to support emerging use cases such as agentic AI, enterprise automation, scientific discovery, and physical AI. [3] Customers are moving workloads from the pilot phase into production, which requires broader model choice and faster data pipelines for mission-critical applications. [3] To support this shift, the companies will deepen their work across AI factories, open models like Nemotron, data processing, and robotics. [3] AI workloads are scaling quickly in terms of how models are trained and how data is indexed. [3]

The capacity expansion coincides with massive financial growth for Nvidia, which reported second-quarter revenue of $96.22 billion. [1] This figure represents a 106 percent increase from a year ago and surpassed estimates of $92.17 billion. [1] Nvidia's Data Center revenue surged 117 percent year-over-year to hit $89.0 billion. [1] Net income for the quarter climbed 126 percent to $59.7 billion, ranking as one of the most profitable quarters any publicly listed company has ever recorded. [1] The company expects revenue growth of 70 percent heading into fiscal year 2028, with customer forecasts pointing to growth doubling next year. [1]

The sheer scale of the AWS procurement underscores how aggressively major cloud providers are securing compute for the next phase of AI development. With an influx of Blackwell Ultra and Rubin GPUs, AWS is positioning itself to support intensive physical and agentic AI workloads that standard infrastructure cannot handle efficiently. Nvidia's staggering $89 billion data center revenue reflects that cloud giants are absorbing nearly every unit of hardware capacity available in the market. What the sources don't address: How this massive enterprise hardware buildout will affect cloud pricing and availability for smaller developers who lack enterprise-level budgets.