Global Edition
The Doom Ledger
Est. 2026
AI is not a new church, and people don’t need a new pope.

Special Purpose Vehicles and the Creative Accounting of AI Deals

When a deal is hard to price it gets structured. The structures have become elaborate enough to obscure what is being bought and sold.

Financial and legal documents — photo by RaphaelQS, licensed under CC0 via Wikimedia Commons.

The transactions underpinning the artificial intelligence build-out have become structurally complex in ways that make them difficult to compare and, in some cases, difficult to interpret at all.

Leases, prepayments, preferred liquidation terms

  • Joint ventures and special purpose vehicles that hold data centre assets while the operator takes a lease.
  • Compute commitments that function as prepayment for services but are reported as revenue by the provider.
  • Strategic investments with favourable liquidation terms that make headline valuations misleading.
  • Sale-leaseback arrangements that move capital expenditure off a balance sheet.

The problem is adding them together

Each structure has a legitimate commercial purpose. The problem is aggregation. When several large companies use different structures, their reported figures are not directly comparable, and an investor trying to assess the sector’s total capital intensity is comparing accounting treatments rather than businesses.

Advertisementin-article · responsiveAfter the opening section of a long article. Never between a heading and its own body.

The second-order effect is on risk. If the same asset appears in multiple balance sheets, or if a commitment is recognised as revenue by one party and as an obligation by nobody, then the system’s true leverage is understated.

Detailed disclosure against a headline number

Disclosure quality is the best available signal. Companies that describe off-balance-sheet arrangements in detail, state their commitments, and explain how related parties transact are more trustworthy than those that report a single headline number. Where the disclosure is thin, the sceptical assumption — that the structure exists to achieve a reporting outcome — is not unreasonable.

Compute as something between a product and an asset

In an environment where accelerators are both scarce and strategically important, compute functions as something between a product and an asset. It can be prepaid, pledged, resold and used as consideration in a transaction.

Advertisementin-article-2 · responsiveRoughly two thirds down a long article.

That flexibility produces arrangements that are difficult to classify: an investment denominated in processing time, a cloud credit used to acquire an equity stake, or a capacity reservation treated as a revenue commitment. Each has precedent somewhere in finance and none has a settled treatment here.

Sale, financing, or related-party deal

As these structures accumulate, the audit function becomes more important and more difficult. An auditor must determine whether a commitment is a sale, a financing or a related-party transaction, and the answer determines how it appears in the accounts.

The absence of settled practice means comparability between companies is poor and will remain so until either a standard emerges or a correction forces one.

The cleanest available answer is to compare a company’s cash flow from customers against its cash commitments to suppliers. Where the second is growing faster than the first and the gap is bridged by financing, the structure is doing work that the accounts do not fully display. That check requires only public filings, and surprisingly few analysts run it.

Image credit and licence details for every photograph on this site are listed on the credits page. This article is editorial content; it carries no sponsored material.

Related

Advertisementfooter-banner · 970x90End of page, above the site footer. Never inside the footer itself.