OpenAI has entered into a multi-year agreement with Amazon Web Services valued at approximately US$38 billion, designed to power its next-generation AI infrastructure. The deal is set to last at least seven years and gives OpenAI access to hundreds of thousands of high-performance GPUs (graphics processing units) and vast data-centre capacity via Amazon’s cloud infrastructure. Under the agreement, OpenAI will begin using AWS compute immediately, with all planned capacity expected to be in place by end of 2026, and expansion into 2027 and beyond possible.
This marks a significant pivot for OpenAI: from being heavily reliant on Microsoft’s Azure cloud (its prior exclusive provider) to diversifying its cloud partnerships and signaling an end to Microsoft’s first-refusal rights under their prior contract.
What the Deal Includes: Scale, Technology & Timing
At its heart, the agreement is about compute scale. OpenAI will receive hundreds of thousands possibly millions over time of Nvidia’s latest AI accelerator chips (including the GB300 and GB200 lines) in AWS data clusters. These compute resources will train and run large-scale models, power services like ChatGPT, and support OpenAI’s ambition to build agentic AI systems.
AWS will provide UltraServer EC2 instances, optimized architectures, massive-scale CPU and GPU clusters and data-centre support. The deployment timeline: immediate usage begins now; full-scale availability by late 2026; expansion further into 2027 and beyond.
OpenAI CEO Sam Altman highlighted that scaling “frontier AI” demands massive, reliable compute and this partnership cements OpenAI’s infrastructure backbone for the next era.
Why This Matters: Strategic Shifts in AI & Cloud
The deal provides OpenAI with infrastructure certainty as it scales its ambitions. With revenue run-rates reportedly nearing US$20 billion and commitments to spend at least US$1.4 trillion on compute capacity (to support roughly 30 gigawatts of computing power) this partnership gives muscle behind those assertions.
By moving away from exclusive reliance on Microsoft, OpenAI opens its operations to competitive cloud supply, potentially driving better economics and resilience.
For Amazon / AWS
This is a major endorsement of AWS’s capability to support the heavy compute demands of leading AI model developers, a public boost of confidence at a time when AWS faces competitive pressure from Microsoft and Google. In markets, Amazon’s stock responded positively (up about 5 % on the news).
This deal signals that large-scale AI players view cloud infrastructure not just as a utility but as a strategic asset. The arms race for compute is intensifying: infrastructure, chips, data-centres and cloud services are becoming key differentiators.
It also underscores the importance of partnerships and supply-chain diversification in an era of geopolitical and export-control risk.
Despite the scale and ambition, there are multiple risks to monitor:
- Execution risk: Deploying hundreds of thousands of GPUs, synchronising them across data-centres, ensuring reliability, latency and efficiency is non-trivial. If OpenAI fails to integrate or scale as planned, costs and delays may mount.
- Financial sustainability: While OpenAI is generating large revenue run-rates, its infrastructure commitments (including this deal and other compute agreements) total over US$1 trillion. That raises concerns about how and when the returns will materialise. Analysts are increasingly asking whether this is a classic “AI bubble” risk.
- Cloud provider dynamics: With OpenAI diversifying cloud partners, competition among providers (AWS vs Azure vs Google Cloud) for large-model workloads may intensify. For AWS, the challenge will be delivering performance and cost competitiveness at scale and keeping OpenAI satisfied.
- Competitor reactions: Microsoft, which previously held exclusivity with OpenAI, might face strategic and operational implications. How it adjusts will shape cloud-AI dynamics.
- Model & application risk: Infrastructure is only one part of the equation. The value ultimately depends on how well OpenAI can monetise its models, manage operational costs and scale applications beyond research and prototypes. The infrastructure alone doesn’t guarantee business success.
The US $38 billion deal between OpenAI and AWS is a landmark in the AI-cloud era: a massive investment in compute infrastructure, a strategic shift in partnerships, and a force-multiplier for OpenAI’s model deployment ambitions. For AWS, it’s a major validation of its AI-hardware and cloud-scale credentials.
Yet, the size of the deal also magnifies the stakes. Execution complexity, cost pressures, competitive reactions and the broader question of monetising these investments all loom large. If everything falls into place, this partnership could underpin the next wave of AI applications and business models. But if bottlenecks or mis-execution arise, the risk may be equally large.
This isn’t just a contract, it’s a statement of intent by OpenAI and AWS about the future of AI infrastructure, and the race to build it.
