How it works
An SPV is a real company in the boring legal sense: it is incorporated, it has directors, it can be sued, it can go bankrupt. What makes it special is that it is forbidden from doing anything except the one job it was made for. It owns one data centre, or one pool of loans, or one aircraft. It has no employees, no other business and no other creditors. Lawyers call this being bankruptcy-remote, and it is the entire point. Lenders to the vehicle are exposed to that asset and nothing else, and if the collapses, the vehicle's creditors are not dragged into the wreckage.
The accounting question is separate from the legal one, and it is the question that decides whether the debt shows up. A sponsor consolidates an entity onto its own balance sheet when it controls that entity and absorbs most of the upside and downside. Structure the vehicle so that an outside partner holds meaningful equity and meaningful control, and the sponsor does not consolidate. The vehicle's borrowings sit in the vehicle's accounts. The sponsor's involvement appears as whatever it actually promised, typically a long lease, and a lease obligation does not become a balance-sheet liability until the lease commences.
A made-up illustration, round numbers, not data. A tech company wants a campus. It does not build it. An investment manager sets up NewCo, puts in $2bn of equity, and NewCo sells $8bn of bonds to insurers. NewCo buys the land and pays the contractors. The tech company signs a twenty-year lease at, say, $900m a year, beginning when the building energises in 2029. NewCo's bondholders are repaid out of that rent. Today the tech company owns nothing, owes nothing, and has committed roughly $18bn. The commitment is disclosed. It is disclosed in a footnote.
SPVs are not a trick invented for . They are the plumbing of securitisation, project finance and most of structured credit, and in their ordinary use they do something genuinely useful: they let a lender underwrite one asset without also having to underwrite the sprawling company that assembled it. They also have a history. Enron's off-balance-sheet vehicles were SPVs, and the abuse there was not the structure but the fact that the risk had never actually left.
Why it matters to this crash
This is the mechanism underneath the single largest number we have written about. Nikkei totted up the off-balance-sheet obligations of Alphabet, Microsoft, Amazon, Meta and Oracle and got about $1.65tn, more than the roughly $1.35tn of debt these companies carry on their balance sheets, and around eight times the level of four years ago (Nikkei). Meta's share is about $420bn, nearly triple its recorded debt. Oracle's was $273.3bn at the end of May. We wrote this up on 18 August and again on 19 August.
The named example is Meta's Hyperion financing: a $27bn bond raised in a special purpose vehicle, backed by a twenty-year Meta lease promise, and not on Meta's balance sheet. Goldman puts sector-wide leasing commitments at roughly $1.5tn, of which about $1tn is invisible in headline statements. Moody's flagged the ballooning of not-yet-commenced leases in February.
So the period of maximum commitment, with contracts signed and concrete pouring and nothing delivered, is also the period of minimum disclosure. Every ratio computed on these companies during the buildout is measuring a capital-light software business that no longer exists. That is not a forecast risk. It is a delivery queue.
The structure also moves risk somewhere we can see less well. Our 18 August piece on the $14bn Meta–BlackRock campus in El Paso found it carries all-risk cover of up to $427mn in construction and $450mn operating, and is not insured against total loss. If that project is financed the way Hyperion was, the uninsured belongs to bondholders in a vehicle, not to Meta. Nobody outside the deal knows how those documents are written.
What would make this dangerous
The first observable is whether the lease promise is unconditional. If rent abates on casualty, force majeure or failure to deliver power, the sponsor's is worth less than the bond priced it at, and that shows up as a downgrade or a widening spread on SPV paper long before it shows up anywhere else.
The second is consolidation. Auditors and the SEC can decide that a sponsor really does control a vehicle and absorb its variability. A reclassification would land tens of billions of debt onto a balance sheet in a single quarter with no change in the underlying economics.
The third is the equity partner. These structures require an outside investor willing to hold the first-loss piece. If insurers and asset managers stop buying that tranche, or demand equity thick enough that the sponsor must consolidate to get the deal done, the whole financing route closes and the buildout has to be funded with ordinary corporate bonds, visibly.
The fourth is a total loss at a site that is not insured for one, with the debt in a vehicle and the rent contingent. That is the case where a $14bn hole becomes a bondholder's problem, and the answer arrives in a court filing rather than an earnings call.