OpenAI Signs $38 Billion Deal With Amazon

OpenAI Signs $38 Billion Deal With Amazon

Summary

OpenAI has agreed a multi-year purchase of approximately $38 billion in AWS cloud infrastructure to train models and deliver services. The arrangement deepens OpenAI’s ties to major cloud players even as it maintains multi-cloud deployments, and it sees Amazon building custom hardware stacks featuring Nvidia GB200 and GB300 chips for both training and inference.

Key Points

• OpenAI will buy about $38 billion worth of AWS compute over multiple years to train and serve AI models.
• Amazon will provide custom infrastructure using Nvidia GB200 and GB300 accelerators, claiming access to hundreds of thousands of NVIDIA GPUs and the ability to scale to tens of millions of CPUs.
• The deal illustrates OpenAI’s strategy of diversifying cloud providers rather than relying solely on Microsoft.
• It tightens industry interdependence: OpenAI already has major relationships with Microsoft, Google, Oracle, Nvidia and AMD.
• Observers warn this arms-race in compute could be a symptom of an AI infrastructure bubble, with US firms projected to spend massive sums on AI infrastructure in coming years.
• OpenAI recently moved to a new for-profit structure (a public-benefit corporation) to raise capital more easily — a context for large-scale infrastructure commitments.

Content summary

The Wired report explains that OpenAI’s $38 billion AWS deal is intended to supply the compute capacity needed for frontier-model training and agentic workloads. Amazon says it will deploy specialised stacks combining GB200 and GB300 Nvidia chips for both training and inference, and that the agreement can scale extensively across GPUs and CPUs.

The move is notable because OpenAI historically leaned on Microsoft; choosing Amazon signals a deliberate multi-cloud posture and highlights how startups and hyperscalers are now financially entangled. Analysts say the agreement shows Amazon is asserting itself in AI infrastructure, while critics worry about the sustainability of huge, upfront compute commitments amid signs of an infrastructure spending rush.

Context and relevance

This story matters if you follow cloud competition, AI economics or the geopolitics of compute. It highlights three industry trends: (1) the compute arms race — huge capital committed to GPUs and specialised hardware; (2) strategic diversification by AI companies to avoid vendor lock-in; and (3) the rise of complex, multi-billion-dollar commercial deals that tie startups and cloud providers together and could reshape market power and margins.

Author style

Punchy: this is a big, strategic move. If you care about who controls the clouds that power future AI — and who profits when models scale — read the detail.

Why should I read this?

Quick and useful — this tells you who’s buying what, and why it matters. If you want the headlines without wading through financial filings: Amazon just got a huge vote of confidence from OpenAI, and that has ripple effects for cloud competition, model costs and where AI gets built.

Source

Source: https://www.wired.com/story/openai-amazon-multi-billion-dollar-deal/