The past year has been dominated by headlines heralding an AI bubble. Stock valuations have rollercoastered throughout the year alongside the heart rates of investors, and tech journalists have grown accustomed to reporting on billion-dollar deals.
Whispers of an AI bubble have been circulating across financial markets for the past year. Despite these warnings, tech stock prices have continued to climb to unprecedented highs. Microsoft, Alphabet, and Meta alone spent an estimated US$78 billion on infrastructure in Q3, with a substantial proportion dedicated to AI-related capacity expansion.
But behind the headlines lies a deeper story: the way AI infrastructure is financed is changing. Off-balance-sheet entities, debt-laden structures, hybrid equity-debt models, and high-yield investors are now entering what was once exclusively the domain of tech capital expenditures.
The Nvidia/OpenAI deal in September has become the poster child for this new era of complex, circular financing.
In a nutshell, Nvidia pledged $100 billion for the construction of new AI datacentres to OpenAI. Who, in turn, committed to kitting out the data centres with Nvidia’s GPUs. OpenAI receives the capital needed for rapid growth, while Nvidia guarantees itself as the exclusive supplier for that expansion, thus completing the loop of mutual benefit.
Meta has also jumped on the unconventional financing bandwagon. Its $27 billion Hyperion data centre deal with fund manager Blue Owl Capital sees Blue Owl retain an 80% interest in the joint venture and Meta will retain the remaining 20% ownership. Meta said via its statement that both have committed to fund their respective pro rata share of the approximately US$27 billion in total development costs for the buildings and long-lived power, cooling and connectivity infrastructure at the campus. Which according to reports makes it the largest private credit deal in the tech infrastructure sector.
Whilst earlier this week Bloomberg reported that the development of a data centre campus linked to Oracle Corp received financing from as many as 20 banks.
The cost of staying competitive
Rising AI expenditures are placing pressure on valuations. As Meta CEO Mark Zuckerberg commented last month, whilst an AI bubble is “quite possible.” The risk of being left behind is worse “If we end up misspending a couple of hundred billion dollars, I think that that is going to be very unfortunate, obviously.”
“But what I’d say is I actually think the risk is higher on the other side.”
Open AI’s Sam Altman also admitted that we are in a phase where investors are overexcited about AI, but countered with “Is AI the most important thing to happen in a very long time? My opinion is also yes.”
Implications for the market
For banks and fund managers, these deals offer access to a rapidly expanding sector. For tech companies, they provide speed and leverage. But the risk is increasingly shifted away from tech firms onto investors and lenders, who must assume long-term obligations and exposure to market cycles.
For some companies like Meta, this structure frees up capital and allows for more flexibility. For OpenAI, it enables vast infrastructure in anticipation of a presumed surge in demand – backed by long-duration debt, and the assumption of continuous market growth.
The AI arms race is no longer just a technological story – it is a financial one. And the house of cards being built may carry stakes far beyond chips and servers.
As one analyst commented, “When you build a house of cards with so many moving parts, from huge leverage, untested market demand and long payout windows… it might look good until the wind blows it over.”
RELATED STORIES
Capacity Middle East 2026
Capacity Middle East is the region’s leading digital infrastructure event, uniting over 3,500 executives from more than 90 countries for visionary content and unrivalled networking and business opportunities.
