AdvertiseFee Deal Desk

Macro

ECB Counts $4.5 Trillion in Hidden AI Debt

The European Central Bank (ECB) has put a number on the debt hiding inside the AI buildout: $4.5 trillion.

Be a creator
November 19, 2026, 07:06 PM UTC3 min read
AI SummaryAI
  • The ECB estimate is 287% higher than its 2025 tally of AI industry commitments
  • The ECB tally covers eight firms including Alphabet, Amazon, Meta, Microsoft, Nvidia and Anthropic
  • The Federal Reserve report put off-balance-sheet obligations above $3 trillion across seven firms
  • Oracle's $300 billion OpenAI contract is cited as collateral for hyperscaler borrowing

A $4.5 Trillion Shadow Ledger

The European Central Bank (ECB) has put a number on the debt hiding inside the AI buildout: $4.5 trillion. Its Financial Stability Review, released on Wednesday, estimates off-balance-sheet commitments at eight US technology firms at that sum, 287% above the tally the bank recorded in 2025. The figure covers years of promised payments that AI companies owe for data center rentals, cloud computation and chips, obligations that sit outside their formal balance sheets under US accounting standards. The firms in the count are Alphabet, Amazon, Meta, Microsoft, Nvidia, Oracle, Broadcom and Anthropic, the maker of Claude, whose $1 trillion IPO landed one day after the report. The ECB found that these hidden commitments dwarf everything the companies declare as formal liabilities such as debt. The mechanism is straightforward: hyperscalers turn long-term client compute contracts into collateral and borrow against the committed future payments, and Oracle's $300 billion deal with OpenAI is the largest case the bank cites. "Such cash flow commitments are debt-like in economic terms," the review states, yet under US accounting standards they never appear as liabilities on the balance sheets of the hyperscalers.

Circular Financing and the Fed's Own Count

The accounting gap has a technical root. A US balance sheet records a lease only once the asset becomes available for use, so purchase commitments that have not commenced live in footnotes, where they have ballooned over the past year. The Federal Reserve has been counting along the same lines. In a Monday speech, Fed Governor Michael Barr independently counted more than $100 billion of hyperscaler and GPU-maker stakes in AI labs and cloud startups, and called the pattern "circular financing." The Fed's stability report, published Friday, tallied more than $1 trillion of signed but unstarted leases, then added more than $2 trillion of take-or-pay contracts and guarantees at seven of the same firms. On the Fed's calculation, the nominal value of these off-balance-sheet obligations now exceeds $3 trillion. ECB Vice-President Boris Vujcic described the loop this way: chipmakers and cloud providers may invest in AI labs that, in turn, commit to purchasing their hardware and services. The ECB's review repeatedly flags that ring. Cash rotated inside the industry, without true revenue from consumers or outside sectors, can misleadingly inflate top-line figures and forecasts for AI companies. The bank also warned that the rapid shift toward debt finance invokes historical parallels with the dotcom and subprime bubbles.

On COINOTAG's reading, the two counts differ in method but agree in direction: the AI sector's debt-like exposure runs roughly an order of magnitude beyond what reported liabilities imply, and the wider ECB figure of $4.5 trillion is the one that now sets the size of the question. The risk is refinancing, since the commitments are fixed payments while the revenue backing them depends on AI demand holding. That repricing, in our view, will travel through credit spreads rather than through chart levels of support and resistance on tech equities, which tend to lag the credit signal. The next stress test arrives with the companies' own disclosures on how much of the $4.5 trillion converts into on-balance-sheet debt.

Primary sources

COINOTAG's editorial and research desk.

AI-Assisted

AI-generated, AI-reviewed, under COINOTAG editorial oversight.

Price-Impacting News

Live
All breaking news
Breaking News