In the battle for AI dominance, computing power has become the new oil, and OpenAI’s $38bn (£29bn) partnership with Amazon is the latest sign of just how expensive that race has become.
The ChatGPT maker’s seven-year agreement with Amazon Web Services (AWS) marks a turning point not just for OpenAI, but for the wider cloud industry.
The deal, giving OpenAI access to hundreds of thousands of Nvidia GPUs across Amazon’s data centres, underscores a wider shift away from exclusive alliances and toward multi-cloud independence as AI companies compete for scarce computing power.
“Scaling frontier AI requires massive, reliable compute,” said OpenAI CEO Sam Altman. “Our partnership with AWS strengthens the broad compute ecosystem that will power this next era and bring advanced AI to everyone.”
It’s a big moment for Amazon, which has been playing catch-up in the AI infrastructure race.
AWS’s share of the global cloud market slipped from 34 per cent before ChatGPT’s launch in 2022 to around 29 per cent this year, as Microsoft Azure and Google Cloud gained ground.
However, Monday’s announcement pushed Amazon shares to a record high, adding more than $140bn to its valuation.
Amazon tries to regain AI ground
The agreement is a clear win for AWS chief Matt Garman, who said the company is “uniquely positioned to support OpenAI’s vast AI workloads”, after years of criticism that Amazon was slow to capitalise on the AI boom compared with rivals.
The tech giant has ramped up spending to change that narrative, opening an $11bn AI data centre in Indiana, US, known as Project Rainier, as well as investing heavily in custom chips.
AWS already hosts Anthropic, another leading AI firm partly owned by Amazon.
“While it’s small relative to other OpenAI contracts, this is a symbolic win,” said Mamta Valechha, analyst at Quilter Cheviot. “It shows Amazon is back in the conversation with the company spending more than a trillion dollars on computing power.”
The scale of the deal also gives AWS a marquee client as it faces tougher competition from Azure, which recently secured a $250bn commitment from OpenAI, and Oracle, which has its own $300bn arrangement with the AI firm.
A reshaped OpenAI
For OpenAI, the agreement cements a new phase of independence following a corporate restructure that loosened its ties with Microsoft.
The move gives the company flexibility to partner with multiple cloud providers, which is necessary as it races to train next-generation AI models that demand unprecedented computing capacity.
“The deal with AWS shows OpenAI’s path to leadership is paved with access to as much computing power as it can get its hands on,” said Kim Forrest, chief investment officer at Bokeh Capital Partners.
The company has signed over $1tn in infrastructure deals this year alone, spanning Oracle, Broadcom, AMD and Nvidia.
However, it remains deeply unprofitable, with Microsoft’s latest quarterly update revealing that OpenAI burned through around $12bn last quarter.
Market reaction
Investors have largely cheered the news. “Once again that AI magic has been casting a spell on investors,” said Danni Hewson, head of financial analysis at AJ Bell.
“This deal gives Amazon a big leg up in the AI stakes, while giving OpenAI the infrastructure it needs to keep its products running.”
Still, there’s a growing unease about how sustainable this investment spree really is, with the Bank of England, the IMF, and JP Morgan’s Jamie Dimon having all recently warned of “unprecedented” levels of tech spending and market exuberance.
To some, though, this is no bubble. “AI isn’t speculative, it’s compounding – you can’t un-invent intelligence,” said Roman Stanek, chief executive of GoodData.
“This isn’t about adding AI to business processes; it’s about rebuilding those processes around AI.”
The OpenAI–Amazon tie-up captures a defining tension of the AI age, where the rush to build limitless capability in a world with finite resources.
The power, chips and capital required to train advanced systems are stretching even the biggest tech firms.
For now, the partnership appears mutually beneficial, offering Amazon prestige and providing vital breathing room for OpenAI as it races toward AGI.
Yet it also signals that the AI power race is no longer about who has the most sophisticated algorithms but who actually controls the infrastructure beneath them.