Seattle: Amazon.com Inc. is expanding its partnership with Nvidia by adding 2 million high-end Nvidia graphics processing units to its data centre infrastructure, underscoring the technology giant’s continued push to expand artificial intelligence computing capacity.
The additional GPUs are expected to be deployed across Amazon Web Services data centres during 2027 and 2028. The latest commitment comes only months after Amazon announced plans to install more than 1 million Nvidia chips in AWS infrastructure beginning in 2026.
The expansion highlights the extraordinary demand for computing power as companies accelerate spending on generative AI, AI agents and other advanced workloads. It also demonstrates that Amazon continues to rely heavily on Nvidia’s technology even as it develops its own artificial intelligence chips.
Amazon adds 2 million Nvidia GPUs
Amazon and Nvidia announced that AWS will deploy an additional 2 million Nvidia GPUs across its global infrastructure in 2027 and 2028.
The chips include Nvidia’s Blackwell Ultra, Rubin and Rubin Ultra GPU platforms. These processors are designed to handle demanding workloads involved in training and running artificial intelligence models.
The latest order comes on top of the more than 1 million Nvidia GPUs that Amazon had already planned to install in AWS data centres starting this year.
That means Amazon’s planned Nvidia GPU deployment will expand substantially as it builds additional computing capacity for customers.
Neither Amazon nor Nvidia disclosed the financial terms of the latest agreement. However, the chips are high-end products that can carry list prices running into tens of thousands of dollars per unit, although large customers typically receive discounts.
Demand for AI computing continues to surge
The deal comes as demand for AI infrastructure accelerates worldwide.
Businesses are increasingly moving AI applications from experimentation into production, requiring significantly greater computing power to train and operate models.
AWS says customers are scaling workloads involving agentic AI, scientific research, enterprise automation and robotics. The additional Nvidia capacity is intended to support these workloads across Amazon’s global infrastructure.
The move also reflects the changing nature of AI workloads.
Large language models require enormous amounts of computing power during training, while AI agents and other applications can require substantial resources when they are running continuously.
For cloud providers such as Amazon, access to sufficient GPU capacity has therefore become an important competitive advantage.
Nvidia remains central to Amazon’s AI strategy
Amazon’s decision to expand its Nvidia GPU deployment demonstrates that Nvidia remains a key supplier despite the company’s efforts to develop its own chips.
Nvidia’s GPUs have become a central component of the AI boom, powering many of the systems used to train and operate large AI models.
Amazon, Microsoft and Alphabet-owned Google are among Nvidia’s largest customers, but all three companies are also developing alternative hardware designed specifically for their own cloud infrastructure.
Amazon has invested heavily in Annapurna Labs, its in-house chipmaking operation.
The company has developed its own Trainium AI accelerators as well as processors designed for cloud computing. These chips are intended to give Amazon greater control over costs and performance while reducing its reliance on external suppliers.
However, the new Nvidia agreement shows that Amazon is pursuing both strategies at the same time.
Amazon continues to develop its own AI chips
Amazon’s custom silicon strategy remains an important part of its long-term AI infrastructure plans.
Annapurna Labs has developed Trainium, an AI accelerator designed to compete with Nvidia’s GPUs for certain workloads.
Amazon has also said that its upcoming Trainium generation will use Nvidia networking technology, demonstrating that the relationship between the two companies extends beyond GPU purchases.
The strategy allows Amazon to offer customers different computing options.
Nvidia’s GPUs can handle a broad range of AI workloads, while Amazon’s own chips can be optimised for specific applications and potentially deliver efficiency advantages.
This combination could become increasingly important as the cost of operating AI systems rises.
AWS expands its AI infrastructure
The additional GPUs will be deployed across AWS’s global infrastructure, including dedicated AI factories.
Amazon and Nvidia are also expanding their collaboration beyond GPUs to include CPUs, networking, open AI models, data processing and robotics.
One element of the agreement involves bringing Nvidia Vera CPU-based infrastructure to AWS.
Vera is designed to support the CPU-heavy workloads associated with agentic AI, including code execution, tool use, data processing and orchestration.
The addition gives AWS another option for customers that require both high-performance CPUs and accelerated AI computing.
Nvidia’s Blackwell Ultra and Rubin chips
The latest deal covers several generations of Nvidia’s AI hardware.
Blackwell Ultra is part of Nvidia’s current high-end AI computing portfolio, while Rubin and Rubin Ultra represent the company’s next-generation platforms.
Deploying these systems across AWS infrastructure will allow Amazon to offer customers access to newer generations of Nvidia technology as AI workloads become increasingly demanding.
The scale of the deployment also gives Nvidia a major long-term customer for its upcoming products.
AI data centres require huge investment
The GPU agreement is part of a much broader expansion of AI infrastructure.
AI data centres require not only GPUs but also high-speed networking, storage, power infrastructure and cooling systems.
Cloud companies are therefore investing billions of dollars to build and expand facilities capable of supporting increasingly large AI workloads.
Amazon has separately announced major investments in AI and cloud infrastructure in several markets.
In India, for example, the company announced an additional $13 billion investment to expand AWS cloud and AI infrastructure through 2030, taking its total planned investment in the country between 2026 and 2030 to $48 billion.
US government AI infrastructure also expands
The Amazon-Nvidia partnership also includes plans to support AI infrastructure for the US government.
AWS and Nvidia said they are working to build AI factories for the US government, including 100,000 GPUs on secure AWS infrastructure for federal and national-security workloads.
This reflects the growing importance of AI computing capacity not only for commercial businesses but also for government agencies.
Amazon has previously announced plans to invest up to $50 billion to expand AI and high-performance computing infrastructure for US government customers. The investment is expected to add nearly 1.3 gigawatts of computing capacity across AWS government cloud regions.
Nvidia faces intense demand
The Amazon deal comes at a time when Nvidia is benefiting from exceptionally strong demand for AI chips.
The company’s data centre business has become the main driver of its growth as technology companies race to build infrastructure for generative AI.
However, Nvidia is also facing supply constraints and rising component costs as the industry attempts to expand production.
The scale of Amazon’s new deployment illustrates how large cloud providers are planning years ahead to secure sufficient computing capacity.
Amazon’s AI ambitions continue to grow
For Amazon, the Nvidia agreement represents more than simply purchasing additional hardware.
AWS is one of the world’s largest cloud platforms, and its ability to offer customers access to powerful AI infrastructure is becoming increasingly important to its competitive position.
The company is competing against Microsoft Azure and Google Cloud, both of which are investing heavily in AI infrastructure and developing their own silicon.
By combining Nvidia GPUs with its own Trainium processors and other custom technologies, Amazon is attempting to build a flexible AI computing ecosystem.
A major boost for Nvidia
The deal is also significant for Nvidia because it reinforces the company’s position as a leading supplier to the world’s largest cloud platforms.
Amazon’s decision to deploy another 2 million Nvidia GPUs demonstrates that demand for Nvidia’s technology remains strong despite the availability of competing custom chips.
The agreement also gives Nvidia an important route into the rapidly expanding market for AI agents, scientific computing and physical AI applications.
As AI models become more capable, the computing requirements associated with them are expected to increase, creating additional opportunities for chipmakers and cloud providers.
Conclusion
Amazon is set to deploy an additional 2 million Nvidia GPUs across AWS’s global infrastructure in 2027 and 2028, adding to the more than 1 million Nvidia chips it had already planned to install from 2026.
The latest chips include Nvidia Blackwell Ultra, Rubin and Rubin Ultra GPUs, with the financial terms of the agreement remaining undisclosed.
The move highlights the extraordinary scale of investment taking place in AI infrastructure as companies increase spending on model training, AI agents, scientific research, automation and robotics.
Amazon is simultaneously developing its own AI accelerators through Annapurna Labs and its Trainium programme. However, the decision to substantially increase its Nvidia GPU deployment shows that custom chips and Nvidia hardware are likely to coexist within AWS for the foreseeable future.
For Nvidia, the agreement reinforces its position at the heart of the AI infrastructure boom. For Amazon, it provides the computing capacity needed to support the next phase of AWS’s AI expansion.