AWS has announced plans to deploy an additional two million Nvidia GPUs across its global infrastructure in 2027 and 2028, building upon its earlier commitment for over one million GPUs. The new deployment will include Nvidia Blackwell Ultra, Rubin, and Rubin Ultra GPUs, along with Nvidia CPUs, networking, and interconnect technology that AWS intends to integrate into its infrastructure.
The initial commitment was revealed at Nvidia GTC in March, where AWS stated its intention to add more than one million Nvidia GPUs starting in 2026. Deliveries are set to begin in 2026 and continue through the end of 2027, bringing AWS’s planned Nvidia deployments from both commitments to over three million GPUs. There was no update provided on the progress of the earlier commitment in the recent announcement on August 26.
Furthermore, AWS has outlined plans for 100,000 Nvidia GPUs on secure infrastructure for US federal and national-security workloads. These systems will contribute to the development of AI infrastructure for the US government.
In addition to these developments, Amazon CEO Andy Jassy highlighted that AWS has customer reservations for computing capacity extending into 2028. The company’s contracted backlog reached $496 billion by the end of the second quarter, with cloud revenue reaching $42.2 billion, a 37% increase from the previous year.
Jassy noted that despite the significant investments in computing capacity, there is still an expected shortfall in meeting the demand by 2026. Amazon’s capital expenditure for the second quarter of 2026 surged to $53.1 billion, compared to $31.4 billion a year earlier. The company raised its 2026 capital expenditure forecast to $220 billion in July, citing higher memory costs as a primary factor.
These infrastructure investments by AWS and other tech giants like Microsoft, Alphabet, Meta, and Oracle are on the rise, reflecting the growing demand for cloud services and AI capabilities. The collaboration between AWS and Nvidia extends beyond GPUs to include networking equipment and custom silicon infrastructure for future AI systems.
As Nvidia continues to expand its deployments, it faces component constraints that impact its margins. The company’s data-centre revenue doubled in the second quarter, driven by demand for AI cloud services. Nvidia is also working with cloud providers like AWS, Google Cloud, Microsoft, and Oracle to deploy new platforms like Vera Rubin, which offer enhanced AI capabilities.
The exponential growth in data-centre electricity consumption underscores the need for sustainable infrastructure development to support the expanding cloud services industry. With investments in power, cooling, and networking infrastructure, tech companies are gearing up to meet the rising demand for AI and cloud computing services in the coming years.
