In a move that redefines the cloud computing landscape, Amazon Web Services (AWS) and OpenAI have entered a multi-year, $38 billion agreement that will enable OpenAI to run its core AI workloads on AWS infrastructure. The deal positions Amazon as a key pillar in OpenAI’s global AI operations while marking one of the largest commitments in the cloud computing sector.
The agreement grants OpenAI immediate and expanding access to AWS’s high-performance computing environment. This includes hundreds of thousands of NVIDIA GPUs deployed in AWS data centers, with capacity to scale to tens of millions of CPUs over the next several years. The infrastructure will support everything from real-time inference for ChatGPT to the training of OpenAI’s next-generation foundation models.
“Scaling frontier AI requires massive, reliable compute,” said OpenAI CEO Sam Altman. “Our partnership with AWS strengthens the broad compute ecosystem that will power this next era and bring advanced AI to everyone.”
Big Plus for AWS
The deal represents a shifting trend in the AI ecosystem. For years, Microsoft’s Azure cloud was OpenAI’s exclusive home, a relationship reinforced by Microsoft’s $13 billion investment. But with Microsoft’s exclusivity ending earlier this year, OpenAI has been free to diversify its cloud relationships.
For AWS, the deal underscores its determination to reclaim leadership in the cloud market’s most high-stakes arena: generative AI. After years of dominance in cloud infrastructure, Amazon has faced fierce competition as Microsoft and Google signed multibillion-dollar partnerships with AI startups. The OpenAI pact suggests a turn of fortunes for AWS.
Matt Garman, CEO of AWS, touted the deal as a sign of the company’s cloud prowess. “As OpenAI continues to push the boundaries of what’s possible, AWS’s best-in-class infrastructure will serve as a backbone for their AI ambitions,” he said.
Prepping for an IPO
Behind the numbers is an infrastructure project of extraordinary complexity. AWS engineers are deploying enormous clusters of NVIDIA GPUs, connected through Amazon’s EC2 UltraServers and linked on a high-bandwidth, low-latency network fabric. This architecture will enable OpenAI to process the computationally dense tasks required to train and refine large language models. All of the new capacity is expected to be online by 2026, with expansion into 2027 and beyond.
For OpenAI, the partnership offers not just additional compute, but strategic independence. As Altman’s company prepares for a potential public offering, diversifying its infrastructure partners signals a maturing operational model, one less dependent on a single backer. By distributing workloads across multiple hyperscalers, OpenAI moves toward resiliency, lower cost per compute cycle, and greater flexibility as it develops its next AI models.
The collaboration also expands the companies’ existing integration. OpenAI’s foundation models, now available through Amazon Bedrock, are already used by customers such as Thomson Reuters, Peloton, and Comscore for tasks ranging from agentic workflows to scientific modeling. This new announcement formalizes that relationship and greatly increases its scale.
New Era for Cloud Competition
The sheer magnitude of spending underscores the escalating economics of AI. OpenAI’s broader infrastructure commitments, now totaling more than $1.4 trillion across multiple vendors, have raised concerns about sustainability. Yet Altman has dismissed skepticism, insisting that the market demand for AI capabilities justifies the investment.
For Amazon, the OpenAI partnership arrives as it accelerates its own AI build-out. The company recently opened an $11 billion data center complex in Indiana, dubbed Project Rainier, to host AI workloads, including for Anthropic. Combined with the OpenAI agreement, these moves highlight the depth of AWS’s capacity: it can offer world-class infrastructure for multiple top AI model developers.
As the largest AI players work to secure computing power, the AWS–OpenAI alliance signals that the cloud wars are entering a new phase, one defined less by storage and networking and more by who can deliver the most advanced, most efficient infrastructure for artificial intelligence. In that contest, this deal places Amazon squarely back in the center of the field.
