OpenAI has signed a $38 billion, seven-year deal with Amazon Web Services (AWS) to power its next-generation artificial intelligence (AI) workloads. The agreement, announced earlier this week, marks a significant shift in OpenAI’s cloud strategy and reflects broader changes in how hyper-scalers—large-scale cloud service providers that operate massive data centres—are positioning themselves.
“We all need more compute, we need more GPUs for AI just because the demand is growing so much and that's why we are excited to partner with all these cloud vendors, including AWS, Microsoft, Nvidia and others,” said Srinivas Narayanan, vice president, engineering, OpenAI at a press briefing in Bengaluru on Tuesday.
Until recently, Microsoft Azure was OpenAI’s exclusive cloud service provider. This arrangement changed last month when OpenAI restructured its commercial terms with Microsoft, removing the latter’s right of first refusal on cloud contracts. The AWS deal is the first major partnership to follow, and it underscores OpenAI’s pivot towards a multi-cloud strategy, prioritising flexibility, scale, and access to compute resources.
AWS will provide OpenAI with access to hundreds of thousands of Nvidia’s latest GB200 and GB300 GPUs, deployed via Amazon EC2 UltraClusters. These clusters are designed to support both inference and training workloads, including those for ChatGPT and future frontier models. The infrastructure is expected to be fully deployed by the end of 2026, with provisions for expansion into 2027 and beyond. AWS’s ability to deliver this scale of compute—along with its custom silicon (Trainium and Inferentia), security architecture (Nitro System), and global data centre footprint—was a key factor in securing the deal.
“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,” said Matt Garman, CEO of AWS in a press release. “The breadth and immediate availability of optimised compute demonstrates why AWS is uniquely positioned to support OpenAI's vast AI workloads.”
The AWS–OpenAI deal is not an exception. It comes amid a wave of strategic recalibrations across the cloud ecosystem, as hyper-scalers race to secure their place in the AI value chain.
Microsoft, once OpenAI’s exclusive cloud partner, has responded by diversifying its AI alliances. It now licenses models from Anthropic and xAI, while also developing its own foundation models under the MAI initiative. Microsoft’s focus has shifted towards embedding AI across its software suite—most notably through Copilot integrations in Office, Windows, and GitHub—and investing in massive AI-ready infrastructure through a $100-billion joint venture with BlackRock and MGX.
Google Cloud, meanwhile, is leaning into its research roots. It has scaled up its partnership with Anthropic, offering access to its custom TPUs and launching Gemini Enterprise, a full-stack AI platform for businesses. Google’s strategy emphasises tight integration with its data and analytics tools, and its in-house silicon advantage through TPUs.
Oracle, which is often overlooked in the AI cloud race, has emerged as a major infrastructure player. It recently signed a $300-billion deal to build 4.5 GW of AI compute capacity, making it OpenAI’s largest infrastructure partner by volume. Oracle’s approach is to remain model-agnostic, positioning itself as a neutral provider of raw compute power to AI labs and enterprises alike.
Together, these moves reflect a broader shift: Hyper-scalers are no longer just competing on services or pricing; they’re competing on who can deliver the most reliable, scalable, and flexible compute infrastructure.
First Published: Nov 04, 2025, 18:54
Subscribe Now