Neither company disclosed the financial terms of the expanded agreement.
The deal comes as Amazon continues developing its own AI hardware. AWS offers Trainium accelerators as an alternative computing option, and Amazon said the additional Nvidia systems are intended to sit alongside its custom silicon rather than replace it. Amazon’s custom-chip business has reached a $25 billion annualized run rate, supported by $225 billion in commitments from AI labs including Anthropic and OpenAI. Nvidia said demand since the companies’ March agreement has been stronger than anticipated.
AWS is also adding Nvidia’s Vera CPUs. Nvidia CFO Colette Kress said Vera processors will be offered both alongside Rubin GPUs and as standalone products. The move extends the partnership beyond accelerators and gives AWS another processor option as demand grows for infrastructure supporting agent-based computing.
Amazon will also work with Nvidia on NVHBM memory technology for Trainium. Nvidia said the technology delivers 30% more bandwidth and 15% greater power efficiency than HBM4E, allowing Amazon’s own AI accelerators to use Nvidia-developed memory technology even as the two companies compete in parts of the compute market.
Another portion of the agreement will support government workloads. Amazon said 100,000 of the GPUs covered by the three-million-chip commitment will be assigned to AI factories designed for the U.S. government and capable of serving Impact Level 6 security classifications.
The companies are extending their work into robotics as well. Amazon plans to use Nvidia technologies including Omniverse, Cosmos, Isaac and Jetson across its warehouse robotics operations. AWS will also make Nvidia’s Nemotron open models available through Amazon Bedrock and Amazon SageMaker.
The expansion was announced alongside Nvidia’s fiscal second-quarter results. Nvidia reported $96.2 billion in quarterly revenue, including $89 billion from its data center business, which increased 117% from a year earlier. The company projected $108 billion in third-quarter revenue and said Rubin GPUs began production shipments during the quarter.
The scale of Amazon’s new commitment shows that its investment in proprietary chips is continuing alongside, rather than displacing, large purchases of Nvidia hardware. AWS is building Trainium as its own AI accelerator platform while simultaneously reserving substantially more Nvidia capacity for customers requiring access to the company’s latest GPU systems.
“NVIDIA and AWS have built one of the great growth engines of the AI era, and demand is running ahead of every forecast,” Nvidia CEO Jensen Huang said. “For 16 years, we have scaled NVIDIA computing in the cloud together. Now we are expanding our partnership across the full stack—GPUs, CPUs, networking, open models and software—to make agentic and physical AI real at an unprecedented pace and scale that only AWS and NVIDIA can deliver. This expansion reflects customers’ demand for NVIDIA’s platform on AWS.”
This analysis is based on reporting from wccftech.
Image courtesy of The Motley Fool.
This article was generated with AI assistance and reviewed for accuracy and quality.
