OpenAI Inks $38 Billion Deal with Amazon, Marking First AWS Partnership

In a significant move signaling a shift in the AI infrastructure landscape, OpenAI has inked a $38 billion agreement with Amazon Web Services (AWS) for cloud capacity. This marks OpenAI’s first major contract with AWS, the undisputed leader in cloud infrastructure, and suggests a move towards greater independence from its long-time benefactor, Microsoft.

The deal, officially announced on Monday, will see OpenAI immediately begin deploying workloads across AWS infrastructure. This initial phase will leverage hundreds of thousands of Nvidia GPUs within existing AWS data centers in the United States, with plans for substantial capacity expansion in the coming years to meet the burgeoning demands of AI model training and inference.

News of the collaboration sent Amazon’s stock price soaring, with shares climbing approximately 5% on the day.

“It’s completely separate capacity that we’re putting down,” confirmed Dave Brown, Vice President of Compute and Machine Learning Services at AWS. “Some of that capacity is already available, and OpenAI is making use of that,” he added, highlighting the immediate impact of the agreement.

OpenAI’s recent flurry of infrastructure agreements, totaling over $1.4 trillion, including collaborations with Nvidia, Broadcom, Oracle, and Google, has ignited discussions about a potential AI bubble. Critics are voicing concerns about the feasibility of realizing such ambitious projects, particularly regarding the availability of necessary power and resources in the long term.

Until recently, OpenAI maintained an exclusive cloud partnership with Microsoft, who initially invested in the company in 2019 and has since committed a total of $13 billion. However, in January, Microsoft announced a revised arrangement, relinquishing its exclusive status and adopting a right-of-first-refusal model for future requests.

With the expiration of Microsoft’s preferential status, OpenAI is now free to explore partnerships with other hyperscale cloud providers. While OpenAI has already established cloud relationships with Oracle and Google, the AWS agreement represents a strategic alignment with the market leader.

“Scaling frontier AI requires massive, reliable compute,” OpenAI CEO Sam Altman stated in the official release. “Our partnership with AWS strengthens the broad compute ecosystem that will power this next era and bring advanced AI to everyone.”

Despite diversifying its cloud infrastructure, OpenAI remains committed to Microsoft, having recently announced an intention to purchase an additional $250 billion in Azure services.

For Amazon, securing a deal with OpenAI is a significant win, especially considering its existing multi-billion dollar investment in Anthropic, another prominent AI player. Amazon is currently developing an $11 billion data center campus in New Carlisle, Indiana, specifically designed to support Anthropic’s AI workloads, underscoring its commitment to fostering innovation in the AI space.

“The breadth and immediate availability of optimized compute demonstrates why AWS is uniquely positioned to support OpenAI’s vast AI workloads,” remarked AWS CEO Matt Garman in the release.

Amazon’s Q3 earnings report showcased robust 20% year-over-year revenue growth at AWS, surpassing analyst expectations. However, Microsoft and Google reported even faster cloud expansion, with growth rates of 40% and 34% respectively, emphasizing the intensifying competition in the cloud computing market.

The initial phase of the agreement centers on utilizing Nvidia chips, including the highly sought-after Blackwell models. However, AWS executives haven’t ruled out incorporating additional silicon, including Amazon’s custom-built Trainium chip, into OpenAI’s infrastructure in the future. Anthropic is already using Trainium chips in the new Indiana facility, demonstrating the potential for performance and cost optimization.

“We like Trainium because we’re able to give customers something that gives them better price performance and honestly gives them choice,” Brown stated, while refraining from divulging details on specific collaborations with OpenAI on Trainium.

The infrastructure will support both inference, which powers real-time responses in applications like ChatGPT, and the demanding training of next-generation AI models. The agreement allows OpenAI to scale its AWS resources as needed over the next seven years, although detailed plans beyond 2026 are yet to be finalized, leaving room for future technological advancements and strategic realignments.

OpenAI’s foundation models, including open-weight options, are already available on Bedrock, AWS’s managed service for accessing leading AI systems. This allows various companies to leverage these models for diverse applications.

Companies such as Peloton, Thomson Reuters, Comscore, and Triomics are currently utilizing OpenAI models on AWS for an array of tasks, including coding, mathematical problem solving, scientific analysis, and advanced agentic workflows. This highlights the versatility and broad applicability of OpenAI’s technology.

Monday’s announcement signifies a more direct and formalized relationship between OpenAI and AWS.

“As part of this deal, OpenAI is a customer of AWS,” Brown clarified. “They’ve committed to buying compute capacity from us, and we’re charging OpenAI for that capacity. It’s very, very straightforward.”

For OpenAI, currently the highest-valued private AI company, the AWS agreement represents another step toward a potential initial public offering (IPO). By diversifying its cloud partnerships and securing long-term capacity from multiple providers, OpenAI is sending a clear message of independence and operational maturity to the market.

Altman has previously acknowledged that an IPO is “the most likely path” for OpenAI given its significant capital requirements. CFO Sarah Friar has echoed this sentiment, highlighting recent corporate restructuring as a strategic move to prepare for a public offering.