Amazon and Nvidia expanded their corporate partnership on Wednesday, locking in a massive agreement to ship an additional two million Nvidia graphics processing units directly to Amazon Web Services data centers.
These advanced processors are built to handle high-level training and real-time execution for modern software models. The massive order includes Nvidia Blackwell Ultra, Rubin, and Rubin Ultra architectures. Amazon plans to install these processors across its cloud network throughout 2027 and 2028.
The announcement came during Nvidia’s quarterly earnings call, arriving five months after Amazon originally agreed to deploy over one million Nvidia chips across its cloud operations. Company leaders stated that demand for processing power exceeded initial expectations, driving the decision to triple total chip volume.
Neither party disclosed exact financial terms for the expanded order, but industry experts estimate the multi-million chip deal reaches tens of billions of dollars.
This massive hardware order comes even as Amazon continues designing its own in-house silicon chips. Amazon built custom Trainium chips as alternatives to Nvidia hardware, while also deploying Graviton central processors across its server racks to challenge traditional chipmakers like Intel and AMD. Amazon recently reported that its custom chip business hit a $25 billion annualized revenue run rate.
Despite Amazon’s internal hardware push, Nvidia remains the primary supplier for high-end processing hardware. Along with supplying graphics processors, Nvidia will ship standalone Vera central processing units to Amazon Web Services. Nvidia chief executive Jensen Huang expects Vera chips to open up a fresh multi-billion dollar market for the chipmaker.
Beyond cloud server racks, the partnership extends into Amazon warehouse operations. Amazon plans to adopt Nvidia Jetson computers, Isaac development platforms, and Omniverse simulation tools to build and control its physical warehouse robotics fleet.
On the software side, Amazon Web Services will host Nvidia Nemotron open models across its Amazon Bedrock platform and SageMaker cloud management systems, giving corporate clients direct access to pre-built training frameworks.
Nvidia reported quarterly sales reaching $96.2 billion, topping Wall Street expectations. Data center hardware sales generated $89 billion of that total, marking an 87 percent increase compared to the same period last year. Company leadership projects third-quarter revenues to reach $108 billion as production shifts to next-generation Rubin architecture.
To secure future manufacturing capacity and keep pace with global demand, Nvidia committed $279 billion to supply partners, up from $199 billion in the previous quarter. The company allocated $42 billion for current fiscal year production and another $87 billion for next year.
Massive hardware investments across major tech companies highlight a clear market reality. As cloud providers and research labs build out advanced software systems, securing fast hardware and reliable server infrastructure remains the primary focus for industry leaders worldwide.