Amazon.com, Inc. (AMZN.US) has placed another large order for 2 million NVIDIA Corporation GPUs, making a strong statement for the demand for AI computing power.

At the same time NVIDIA Corporation (NVDA.US) reported a record quarterly earnings of $96.2 billion, one of its largest customers, Amazon.com, Inc. (AMZN.US), made a powerful remark on this achievement with what it called the "largest GPU order in history." On August 26, local time, Amazon Web Services (AWS) and NVIDIA Corporation announced the expansion of their strategic partnership, planning to deploy an additional 2 million high-end NVIDIA GPUs in AWS's global infrastructure between 2027 and 2028. This order is double the 1 million GPUs procurement plan that Amazon.com, Inc. announced in March this year, incorporating NVIDIA's Blackwell Ultra, Rubin, and Rubin Ultra chips into AWS's AI landscape. Order Overview: 2 million GPUs, from Blackwell Ultra to Rubin Ultra The scale of this deal is unprecedented in the AI infrastructure sector. Amazon.com, Inc. will complete the deployment in phases in 2027 and 2028, covering NVIDIA Corporation's current three strongest product generations: Blackwell Ultra: An enhanced version of NVIDIA Corporation's current flagship AI chip; Rubin: The core GPU of NVIDIA Corporation's next-generation AI platform; Rubin Ultra: The flagship product in the Rubin series. The unit price of these GPUs is expected to be at least tens of thousands of dollars. Assuming a conservative estimate of tens of thousands of dollars per chip, the contract value for 2 million GPUs could reach several hundred billion dollars. With the procurement plan of 1 million GPUs announced in March, AWS's total commitment for NVIDIA GPUs in 2026 has already reached 3 million units. From "buying chips" to "full-stack collaboration": Vera CPU included in AWS for the first time This is not just a GPU procurement, but also a full-stack technology binding. AWS will deploy NVIDIA Corporation's Vera CPU for the first time in its infrastructurea high-performance processor designed specifically for agent-based AI workloads. NVIDIA Corporation stated that some of AWS's Vera CPUs will be integrated with Rubin, while others will be used as standalone products. The two parties will also deepen their collaboration in NVLink Fusion high-speed chip interconnect technology, pairing it with NVIDIA Corporation's new custom high-bandwidth memory (NVHBM), allowing Amazon.com's in-house Trainium chips to access faster and more energy-efficient memory. Additionally, AWS and NVIDIA Corporation plan to deliver 100,000 GPUs for AI factories for the U.S. government on AWS infrastructure compliant with Impact Level 6 security rating. AWS: NVIDIA Corporation's major buyer This deal confirms a key trend: even as Amazon.com, Inc. invests heavily in its in-house AI chip Trainium, the demand for NVIDIA Corporation GPUs continues to accelerate. NVIDIA Corporation CEO Jensen Huang stated in a statement: "Demand is exceeding every forecast. For 16 years, we have been expanding the computational scale of NVIDIA Corporation in the cloud together. Now, we are extending our partnership to full-end technologycovering GPUs, CPUs, networking, open models, and software." NVIDIA Corporation CFO Colette Kress revealed in the earnings call that the capital expenditures of the "top five hyperscale customers" are expected to increase from $800 billion in 2026 to $1.3 trillion in 2027. AWS, as NVIDIA Corporation's largest buyer of accelerators, is further consolidating its position in this deal. This expansion builds on AWS's announcement at the GTC 2026 conference of a deployment plan for over 1 million GPUswhich this new commitment directly doubles. Competitive Game: Self-developed chips and NVIDIA Corporation's "dual-track parallelism" The most dramatic backdrop of this order is that Amazon.com, Inc. is simultaneously advancing its "alternative strategy" against NVIDIA Corporation. In recent years, Amazon.com, Inc. has been increasing its focus on self-developed AI chips. Its Trainium AI accelerator has attracted clients like Anthropic, paying hundreds of billions for use. In June of this year, Amazon.com, Inc. confirmed it is negotiating with potential clients, planning to sell Trainium chips directly to enterprise data centers, moving from AWS exclusivity to external sales. However, the order for 2 million GPUs indicates that self-developed chips and NVIDIA Corporation GPUs are not an "either-or" relationship, but rather a strategy of parallel expansion. AWS CEO Matt Garman stated that this expansion provides "more ways for frontier labs, enterprises, and governments to build and deploy AI on AWS." Institutions predict that NVIDIA Corporation will still retain at least 70% of AI training workloadswhile the demand scale locked in by this AWS transaction is difficult for competitors' custom chips to match. Conclusion With the additional order of 2 million GPUs, Amazon.com, Inc.'s AWS has made a clear statement in the global AI infrastructure race: even under the multiple pressures of self-developed chips, regulatory scrutiny, and GEO Group Inc. politics, the "arms race" in AI computing power is far from slowing down. From Blackwell Ultra to Rubin Ultra, from GPUs to Vera CPUs, from commercial cloud to federal government AI factoriesthis collaboration between AWS and NVIDIA Corporation is pushing the competition in AI infrastructure from "hundreds of billions" into the "trillion-dollar era."