$1.2 Trillion in Data Centre Commitments Isn't on Anyone's Balance Sheet

The mechanics, in plain English

Here is the structure the BIS described in its March 2026 Quarterly Review, and it is worth understanding because it is now the dominant financing form for AI infrastructure.

A hyperscaler wants a data centre campus. Rather than building it and recording the debt, it partners with a private credit firm to create a special purpose vehicle — a separate legal entity, typically bankruptcy-remote. The SPV is capitalized with equity from a consortium and raises debt through private placements. The hyperscaler holds a minority stake, commits to a long-term operating lease or capacity offtake agreement, and may provide guarantees.

Economically, the hyperscaler has taken on a multi-decade financial commitment. On paper, it has agreed to rent capacity. The lease appears on the income statement as an operating expense rather than on the balance sheet as debt.

The scale

The growth rate is the alarming part. Total commitments rose from $969 billion to $1.2 trillion in five months. Alphabet's disclosed not-yet-commenced data centre leases grew from $23.9 billion in Q2 2025 to $42.6 billion by Q3, with terms extending to 2031 and noncancelable periods running as long as 25 years.

Why this is a market story, not an accounting story

Moody's language was direct: a "material shift in the structure of their balance sheets is becoming evident." That is a ratings agency describing the credit profile of the largest, most widely held companies in every index fund on earth.

Three mechanisms turn this into a price problem:

Free cash flow compression. Committed lease payments are contractual. They consume cash regardless of whether AI revenue materializes on schedule. The companies that built their valuations on software-like margins and fortress balance sheets are becoming capital-intensive infrastructure operators.

The depreciation wall. Assets placed on balance sheets in 2025 and 2026 begin generating substantial depreciation charges by 2027 and 2028. Reported earnings absorb that even if cash flow holds.

Analytical repricing. Rating agencies are increasingly looking past reported debt to total cash commitments. When agencies change how they measure leverage, spreads move before any operational deterioration occurs.

The BIS also noted that hyperscaler credit default swap spreads have risen, particularly for lower-rated issuers — reflecting both supply volume and uncertainty about project payoffs.

The counter-case, stated fairly

The bull argument is not weak. These are among the most cash-generative businesses ever built, with genuine investment-grade credit and real revenue growth attached to the buildout. Operating leases for infrastructure are a legitimate, long-standing financing tool, not an accounting trick — the disclosure exists in the filings for anyone who reads them. And if AI demand is as durable as the capex implies, these commitments are ordinary capacity investments that will look conservative in retrospect.

The honest version of the bear case is not "this is fraud." It is that the aggregation is invisible: no single company's disclosure conveys the systemic total, and the systemic total is what determines whether the private credit funds on the other side of these structures can absorb a downturn.

The exposure chain most investors don't see

Private credit funds — principally the largest alternative asset managers — originate most of this debt. A February 2026 Federal Reserve Bank of Chicago study found banks' direct exposure to AI-adjacent industries averaged just 0.8% of total assets, but cautioned that banks likely hold additional exposure through lending to nonbank financial institutions.

That is the chain: hyperscaler → SPV → private credit fund → bank credit line → depositor. Each link looks small. The chain is not.

If you hold a broad U.S. index fund in a TFSA or RRSP, you hold the top of this chain at index weight. Check your actual concentration with the Quorum AI scanner.

Bottom line

The largest capital commitment in corporate history is being made through a structure specifically designed not to be recorded as debt. That is legal, disclosed, and entirely conventional. It is also why the leverage in this cycle will be measured accurately only after it matters.

Frequently asked questions