Big tech keeps $300 billion of AI exposure off its balance sheets, FT finds
Alphabet, Meta, Broadcom and Nvidia use residual value guarantees to back data-center and chip financing without booking most of the debt themselves.
Published
Big tech companies are using financial guarantees to keep roughly $300 billion of artificial intelligence exposure off their balance sheets, according to the Financial Times.
The mechanism is a residual value guarantee. A tech company agrees to cover part of the shortfall if a financed data center or its chips are later sold for less than an agreed value. Because the underlying debt sits with a special-purpose vehicle rather than the guarantor, most of the exposure does not initially show up as ordinary debt on the guarantor's own books, the Financial Times found.
Alphabet's data-center lease guarantees rose from $16.9 billion to $43.8 billion over six months, with less than 2% of that total booked to its balance sheet, according to the Financial Times. Meta's roughly $50 billion Hyperion data-center project in Louisiana runs through a Delaware vehicle called Beignet Investor, in which Blue Owl holds 80% and Meta 20%, backed by about $28 billion of residual value guarantees supporting roughly $27 billion of debt from Pimco, BlackRock and Apollo. Broadcom has taken on about $29 billion of exposure tied to chip financing for Anthropic, while Nvidia has provided $105 billion in guarantees linked to a SoftBank data-center development for OpenAI, the Financial Times reported. In a separate structure, Fluidstack leases and builds data centers stocked with Google's chips and rents the capacity to Anthropic, the Financial Times said.
A Wall Street Journal analysis cited alongside the Financial Times' findings put the scale higher still, estimating that nine major tech and chip firms, including Alphabet, Amazon, Meta, Microsoft, Oracle, Nvidia, Broadcom, SpaceX and AMD, carry close to $3 trillion in AI-related off-balance-sheet commitments, nearly five times their combined capital spending over the past year. Credit-rating agencies still examine these guarantees even though accounting rules keep much of the exposure out of headline balance-sheet liabilities, the Financial Times noted.
The scale of the structures is notable given the size of AI capital spending commitments, but it fits a pattern of big tech managing investor expectations through accounting treatment rather than changing its underlying AI spending.