Meta, Elon Musk’s xAI, Oracle and data centre operator CoreWeave have led the charge on complex financing deals designed to shield them from the chunky borrowing needed to build AI data centres.

Financial institutions, including Pimco, BlackRock, Apollo, Blue Owl Capital and US banks such as JPMorgan, have supplied at least $120 billion in debt and equity for these tech groups’ computing infrastructure, according to a Financial Times analysis.

That money is funnelled through special-purpose holding companies known as SPVs, and because the financings do not appear on the tech companies’ balance sheets, the risks can look like somebody else’s problem.

SPV structures also increase the risk that financial stress for AI operators could cascade across Wall Street in unpredictable ways.

One large financing institution, a senior executive said: “18 months ago, this would have been unfathomable, and fast forward to today, it’s very much the norm, about the tens of billions of dollars flowing into SPVs to fund data centres. The tech industry can access more capital than any other meaningfully because of the credit profile."

Silicon Valley giants have traditionally generated a lot of cash and had little debt, giving them excellent credit ratings and a fan club of investors who assume the money machine never breaks.

The race to secure computing power for advanced AI has pushed tech groups to borrow more than ever, and off-balance sheet structures protect credit ratings and flatter financial metrics for anyone still pretending spreadsheets have feelings.

In October, Meta completed the largest private credit data centre deal: a $30 billion agreement for its proposed Hyperion facility in Louisiana, which created an SPV called Beignet Investor with the New York financing firm Blue Owl Capital.

The SPV raised $30 billion, including about $27 billion of loans from Pimco, BlackRock, Apollo and others, as well as $3 billion in equity from Blue Owl.

The deal meant Meta could borrow $30 billion without any of the debt appearing on its balance sheet, which made it easier to raise a further $30 billion in the corporate bond market a few weeks later.

Oracle has led the way in structuring large debt deals through third parties to pay for its enormous commitments to lease data centre power to OpenAI, which is an expensive way of saying it has signed up for a lot.

Larry Ellison’s tech group has partnered with builders and financiers such as Crusoe, Blue Owl Capital, Vantage and Related Digital to build numerous data centres that will ultimately each be owned by SPVs.

Its off-balance sheet financing deals include about $13 billion invested by Blue Owl and JPMorgan, including $10 billion of debt, into an SPV that owns its OpenAI facility in Abilene, Texas, plus a $38 billion debt package for two data centres in Texas and Wisconsin and an $18 billion loan for a site in New Mexico.

In each case, Oracle has agreed to lease the facilities from the SPVs, and in the event of a default, lenders would have recourse over the assets rather than the companies managing the sites.

That means the data centre, the land and the chips take the hit first, which is comforting right up until the day it is not.

Raising off-balance-sheet debt through an SPV has become more popular as the capital needed for AI infrastructure has skyrocketed, stretching cash reserves and leaving even rich tech groups looking for someone else’s balance sheet.

Morgan Stanley estimated that $1.5 trillion in external financing would be needed to fund tech companies’ AI plans.

In many cases, investors have convinced themselves that the financial risk ultimately still lies with the tech company leasing the site if AI demand disappoints and the value of these computing warehouses takes a hit.

In the case of Beignet Investor, Meta owns 20 per cent of the SPV and has given a “residual value guarantee” to the other investors. This means the social media group would have to repay the SPV investors if the value of the data centre drops below a certain level by the end of the lease and Meta decides not to renew.

Musk’s AI start-up xAI is pursuing a similar structure as part of a $20 billion fundraise, including as much as $12.5 billion in debt, with the SPV using the money to buy Nvidia graphics processing units and lease them to xAI.

CoreWeave said in March that it had created an SPV to fulfil a $11.9 billion contract to supply computing power to OpenAI, which would “incur indebtedness to finance its obligations”.

In July, it borrowed $2.6 billion to fund its OpenAI contracts, a move that helps keep the party going when the bar tab arrives.

Private capital investors are keen to get in on the AI boom, boosting demand for these novel structures. Tech companies have borrowed about $450 billion from private funds as of early 2025, $100 billion more than over the previous 12 months, according to UBS.

This year, about $125 billion flowed into “project finance” deals such as the Meta and Blue Owl transaction, UBS said, because nothing says stable like reinventing plumbing while the water is running.

Data centre construction has mainly become reliant on deep-pocketed private credit markets, a rapidly inflating $1.7 trillion industry that has prompted concerns over steep rises in asset valuations, illiquidity and the concentration of borrowers.

One banker close to data centre financing deals said: “There is risky lending and underlying credit risk built up in private credit already. This creates an interesting set-up for the next several years, as you have two material risks to the outlook becoming more intertwined.”

The risk depends on how widespread these structures become, since stress spreads faster when everyone uses the same clever template and then acts surprised when it burns.

If multiple AI companies are using SPVs, pain can hit the private credit funds behind them simultaneously with little transparency, which is a lovely feature if you are selling the paper and a dreadful one if you are holding it.

The AI data centre boom relies on a small group of clients, and OpenAI alone has made more than $1.4 trillion in long-term computing commitments across most of the sector’s big players.

Lenders across multiple data centres could therefore be exposed to the same risks if one tenant falters, while also facing uncertainty about access to power, AI regulation, or technological shifts that make current hardware obsolete.

Not all of the big hyperscaler data centre companies have joined the trend, and Google, Microsoft and Amazon, which already had large data centre businesses before the AI boom, have continued to finance construction using cash.

While Google and Amazon have recently tapped bond investors to raise more debt directly, the three companies have not yet disclosed any significant SPV financing, which is either restraint or just better hiding.

The cocaine nose jobs of Wall Street are also pushing into more obscure structures in data centre transactions, with tech bankers saying they have seen securitisation deals on AI debt in which lenders pool loans and sell slices as asset-backed securities to investors.

Two bankers estimated these deals currently numbered in the single-digit billions of dollars.

UBS head of public and private credit strategy Matthew Mish said most investors “feel that actually it’s a good thing you ultimately end up with hyperscaler risk” given these companies’ strong balance sheets and credit profiles.

Mish added SPV financings still “add outstanding liabilities” for tech companies, meaning the “overall credit quality for the hyperscaler would be worse than what’s currently being modelled.”