Amazon has committed to acquiring over 3 million Nvidia graphics processing units (GPUs) by 2029, a number greater than the total number of its full- and part-time employees as of the end of 2025. This large-scale purchase includes a recent agreement announced on August 26, which involves deploying 2 million additional Blackwell Ultra, Rubin, and Rubin Ultra GPUs across Amazon Web Services (AWS) infrastructure in 2027 and 2028. This follows a prior commitment made at Nvidia GTC 2026 in March, where AWS pledged to acquire over 1 million Nvidia GPUs, expected to be delivered between 2026 and 2027. Nvidia’s CFO mentioned that deliveries of the GPUs will begin this quarter, not in 2027 as previously thought. However, neither company has provided details on how they define a GPU or the total cost of the deal. The way GPUs are counted could significantly impact the numbers. In the past, Nvidia counted Rubin-generation GPUs by the number of compute dies rather than the number of physical packages. This changed before CES 2026, with the flagship model now being sold as Vera Rubin NVL72, which includes 72 two-die packages. The upcoming Rubin Ultra model, expected in 2027, will have four compute dies in each package. The deal also includes a national security component, with around 100,000 GPUs designated for AI workloads related to the U.S. government at Impact Level 6, which refers to the sensitivity of the information being processed. Amazon’s main constraint in this deal is supply, not demand. AWS revenue increased by 37% year over year to $42.2 billion in the second quarter, and its contracted backlog rose to $496 billion, up from $364 billion just three months earlier. Amazon’s CEO, Andy Jassy, noted during the July earnings call that AWS capacity will fall short of demand this year and likely next, with demand for 2028 being particularly strong. In addition to its own custom chips, such as the nearly fully subscribed Tranium3, Amazon is preparing for several years of significant investment in AI infrastructure and rising AI demand. This suggests that the company is positioning itself for continued growth in the AI and cloud computing sectors.