AWS × OpenAI × NVIDIA: The Next Wave of 'AI Infrastructure Hegemony' Signaled by a $38 Billion Deal
In early November 2025, reports that Amazon's cloud division, AWS, had signed a $38 billion NVIDIA GPU supply deal with OpenAI shook the industry. This is not merely a cloud usage agreement, but a deal that symbolizes a new geopolitical phase in the AI infrastructure race. This article explains the structure of the contract, the intentions of each company, and the impact on the overall AI compute market.
1. Overview of the contract: A 'triangular partnership' via NVIDIA chips
This contract is a seven-year agreement in which Amazon Web Services (AWS) provides computing power to OpenAI using NVIDIA GPUs, with a total value of $38 billion (approximately 5.7 trillion yen). According to Bloomberg, the core of the contract lies in the 'long-term securing of cloud computing resources,' with supply scheduled to begin sequentially by the end of 2026.
AWS plans to deploy NVIDIA's 'Hopper' and 'Blackwell' series GPUs for use in OpenAI's LLM training and inference. While AWS has its own in-house chips, 'Trainium' and 'Inferentia,' it is noteworthy that they have prioritized NVIDIA chips in this instance.
'The fact that AWS is relying on another company's GPUs is itself symbolic,' points out a host from Bloomberg Technology. 'It speaks to NVIDIA's supply capacity and the reality of GPU dominance in the AI industry.'
2. OpenAI's strategy: Diversifying cloud dependency
OpenAI has long maintained an exclusive relationship with Microsoft Azure, but this contract with Amazon is seen as part of a 'de-Azure' strategy.
According to Bloomberg analysts, OpenAI is building a 'multi-cloud system' that combines Google Cloud, AWS, and NVIDIA's direct supply network.
This is driven by the following two points:
Azure supply constraints: Rapidly increasing demand for ChatGPT, Sora, and other services has strained Azure data centers.
Ensuring AI model scalability: For the GPT-5 generation and beyond, training scales are expected to expand to the tens of trillions of parameters, making distributed processing across multiple clouds essential.
'No single cloud can keep up with OpenAI's growth speed anymore.'
— Analyst at U.S.-based Loop Capital Markets
3. AWS's objective: A parallel path with Anthropic
AWS has already invested up to $4 billion in Anthropic and integrated the Claude model into its own AI platform, 'Bedrock.' With this OpenAI contract, AWS has become the provider of the foundation for the two major generative AI models: 'Claude' and 'ChatGPT'.
Expansion of commercialization via Bedrock: Offering OpenAI models to AWS customers to broaden the base of cloud usage.
Demonstrating GPU procurement power: By showing a long-term supply contract with NVIDIA, they gain an advantage in the 'GPU scramble' against Google Cloud and Oracle.
In the third quarter of 2025, AWS recorded a 20% year-on-year increase in AI-related cloud revenue. The partnership with OpenAI could be the catalyst that pushes this to over 30% growth in the medium term.
4. NVIDIA's position: The biggest beneficiary of the AI cycle
NVIDIA has surpassed a market capitalization of $5 trillion and stands at the center of the AI bubble. With this contract, the company has further diversified its supply destinations, solidifying a system where it supplies chips to all three major hyperscalers: Microsoft, Google, and Amazon.
Investment bank Loop Capital described NVIDIA as the 'pick and shovel' in the 'gold rush' of generative AI adoption.
It is estimated that the AI data center market will reach over $3 trillion cumulatively over the next five years, with half of that dependent on NVIDIA products.
5. Investor perspective: Toward the 'second wave' of AI infrastructure
According to Bloomberg's analysis, while Meta and Alphabet are accelerating their AI investments, investors are beginning to prioritize 'AI spending backed by cash flow.' There is a growing trend where companies with cloud revenue, such as Amazon and Microsoft, are being more highly valued by the market.
Furthermore, fundraising from private bond and private credit markets for data center construction is intensifying. Google has issued $15 billion in corporate bonds, and Meta has also initiated new debt financing for AI infrastructure investment.
'AI investment is no longer about FOMO (fear of missing out); it has become national-level infrastructure investment.'
— Partner, Blue Owl Capital (US)
Conclusion: The New 'Cloud Geopolitics' of the AI Era
This Amazon–OpenAI–NVIDIA deal is more than just a commercial transaction.
It symbolizes a struggle for hegemony across the three-layer structure of cloud, semiconductors, and AI models.
OpenAI holds the 'technological core' of the international AI competition,
AWS controls the 'supply chain' of data center capital,
and NVIDIA has positioned itself as the issuer of the 'currency' of computing power.
The competition surrounding AI compute will shift its focus from 'who builds the models' to 'who controls the power, GPUs, and capital.'
The $38 billion deal is nothing less than a signal marking the dawn of this new phase.
