AI’s Off-Balance-Sheet Bill Tops $4.1 Trillion: Should You Be Concerned?

A sharp sell-off in Intel on 30 October briefly pushed its intraday loss to 30%, taking the shares as low as $64.24 and erasing nearly $150 billion in market value. It was the company’s worst single-day performance since the 1970s. Behind the decline was Q4 revenue guidance of $15 billion to $16 billion, more than $1 billion short of analysts’ average estimate of approximately $17 billion. CFO David Zinsner also made clear that Intel does not expect to generate positive cash flow this year. An incident at its Israeli factory is expected to lower Q4 revenue by approximately $300 million. Meanwhile, PC shipments are projected to fall by approximately 12% this year, followed by another 10% in 2027. Between the start of the year and 29 October, Intel’s shares had at one stage climbed nearly 150% as investors valued it as an AI beneficiary. The earnings release exposed a contrasting picture: fabrication plants were absorbing cash flow, and higher memory prices were squeezing the core PC business. Operational missteps at one company explain Intel’s difficulties. But the subsequent compilation of the $4.11 trillion figure indicated a problem extending well beyond a single company’s operational failures.

What Makes Up the $4.11 Trillion?

Rather than conventional debt, the $4.11 trillion consists principally of three types of obligation: non-cancellable purchase commitments, leases already signed but not yet commenced, and guarantees and credit support extended to data centre and chip projects.

Meta’s third-quarter figures show non-cancellable purchase commitments climbing to $455 billion from $349 billion, while leases yet to commence grew to $356 billion from $279 billion. Residual value guarantees covering the Hyperion and El Paso joint ventures amounted to $41.5 billion. At Amazon, unconditional purchase obligations almost doubled to $254 billion from $130 billion, and leases yet to commence expanded to $165 billion from $137 billion. The company added $131 billion in commitments over a single quarter, or 41%, the largest percentage rise among hyperscalers. Alphabet, meanwhile, recorded a $95 billion increase in purchase commitments. Its leases yet to commence rose to $131 billion from $85 billion, energy equipment guarantees doubled to $16.2 billion, and new TPU hardware repurchase guarantees amounted to $6.6 billion. With total off-balance-sheet commitments of $1.059 trillion, Alphabet became the first individual company to cross $1 trillion. Microsoft’s expansion was the most restrained, at $35 billion, or 6%. Its purchase commitment figures, however, still come from the disclosures in its end-June Form 10-K. The actual figure will most likely be revised upwards in the next annual report.

The most aggressive expansion came from SpaceX. Off-balance-sheet commitments stood at $29 billion at the close of the second quarter, reached $40 billion by the end of the third quarter and climbed again to $51 billion on 26 October, representing a rise of 174%.

These categories together clarify what off-balance-sheet commitments represent: not funds already borrowed, but expenditure to which companies have already committed. Signing a lease creates a rental obligation whether or not servers have been fully installed or computing capacity has been sold. Making a purchase commitment similarly requires payment, irrespective of whether chips arrive on time or demand materialises. Rather than appearing as interest in the income statement or liabilities on the balance sheet, these obligations sit in the financial statement footnotes as “commitments”. Their cash outflows are nevertheless just as inflexible as those associated with on-balance-sheet debt.

Against these commitments, the companies have approximately $1 trillion in actual debt and lease liabilities recognised on their balance sheets. Off-balance-sheet commitments therefore exceed on-balance-sheet liabilities by a ratio of more than 4 to 1.

How Was Nearly $1 Trillion Added in 90 Days?

In just over four months, off-balance-sheet commitments more than doubled: they stood at approximately $1.8 trillion at the end of June, reached $3.12 trillion by the end of August and rose to $4.11 trillion in October. Even considering only the five companies that have already submitted third-quarter reports, the combined figure increased to $3.05 trillion from $2.48 trillion. That represents a quarterly addition of $577 billion, or 23%.

The speed of that increase is a signal in its own right. Credit ratings, interest costs and the bond market’s ability to absorb new issuance limit the growth of on-balance-sheet debt. Off-balance-sheet commitments, by contrast, can expand in a much “quieter” manner from an accounting perspective. If additional balance-sheet leverage would lead to a rating downgrade and higher funding costs, a company can structure expenditure through long-term leases or purchase commitments at a substantially lower “cost” in its financial statements.

The clearest illustration is Meta. Despite operating cash flow of $97 billion and credit ratings of Aa3/AA-, its CDS climbed to an all-time high after the earnings announcement. It subsequently announced plans to raise up to $95 billion through equity. Of that amount, $37.5 billion had already been raised, while approximately $60 billion remained to be executed through an at-the-market offering programme. Thus, an AA- rated company is simultaneously building off-balance-sheet commitments, facing market-driven increases in its CDS and raising equity to replenish capital. Together, these developments indicate that investors have started factoring off-balance-sheet commitments into credit pricing, with the company opting to respond through equity rather than borrowing.

A different view emerges from Amazon’s figures. Capital expenditure for the quarter hit $77.26 billion, while trailing twelve-month free cash flow moved from positive $14.8 billion a year earlier to negative $42.4 billion. In the same quarter, its off-balance-sheet commitments rose by $131 billion, recording the largest percentage increase among hyperscalers. It is no coincidence that capital expenditure and off-balance-sheet commitments accelerated together: both reflect, in accounting terms, the same race to build AI infrastructure.

Why Do Rating Methodologies Differ?

Differences in leverage calculations were highlighted by Goldman Sachs credit strategist Amanda Lynam: S&P makes adjustments for off-balance-sheet leases, whereas Moody’s and Fitch do not. As a result, the leverage shown for an individual company varies with the framework applied. Such methodological differences have limited consequences when off-balance-sheet commitments total a few hundred billion dollars. At $4.11 trillion, however, or more than four times on-balance-sheet liabilities, they are large enough to change credit assessments.

Meta’s outlook was cut on 30 October by S&P from “stable” to “negative”, although its AA- rating was retained. The stated reasoning directly linked AI capital expenditure to free cash flow turning negative and specifically highlighted the increase in Meta’s future lease obligations to $355.7 billion. Although the company’s recognised debt remains modest, adding those $355.7 billion in lease obligations through adjustments under S&P’s methodology changes the path of its leverage.

Another concern was raised by Mark Habib, head of digital infrastructure at S&P Global. He noted that implicit guarantees for data centre construction or smaller AI companies may leave hyperscalers with substantial undisclosed AI-related debt, describing these as “interconnected financing structures”. Over the past year, guarantees supplied by US technology giants for AI data centres and chips have reached up to $300 billion. They may sit across several line items as residual value guarantees, joint venture commitments or credit support. Examples include Meta’s $41.5 billion in residual value guarantees for Hyperion and El Paso; Alphabet’s energy equipment guarantees and TPU hardware repurchase guarantees fall into another category.

Warnings have already come from the Bank for International Settlements that opaque AI financing compounds financial vulnerabilities, with circular financing the most prominent risk. A further warning from S&P identifies revenue inflation arising from circular financing as one of the main sources of increasing uncertainty in the AI industry.

How Is the Market Assessing the Risk?

It is the inflexibility of cash outflows, set against uncertain corresponding revenue, that makes off-balance-sheet liabilities risky, rather than the fact that they are “hidden”. The sequence of transmission runs as follows: cash receipts undershoot expectations → project cash is insufficient → collateral values fall and funding conditions tighten → parent companies or guarantors inject additional funds → construction and procurement are scaled back.

Cash receipts set this process in motion. A project’s revenue hinges on its ability to rent out computing capacity at sufficiently high rates. Lower rental rates leave cash inflows unable to meet lease payments and purchase commitments, with project companies coming under pressure first. Declining collateral values and tighter financing follow, requiring parent companies or guarantors to supply more funds. Off-balance-sheet commitments then become cash outflows on the balance sheet. Where the parent company’s own cash flow is also contracting, meeting that funding requirement entails further borrowing or equity financing.

A threshold is central to this transmission process. With normal cash receipts, off-balance-sheet commitments remain no more than an entry in the footnotes. Once a project faces a cash shortfall, however, their inflexibility can carry the pressure back to the parent company much sooner than anticipated. The 30% loss in Intel’s market value in a single day on 30 October illustrates this at the micro level. At the macro level, simultaneous shortfalls at multiple project companies could rapidly convert part of the $4.11 trillion into actual funding requirements.

On-balance-sheet measures continue to provide the principal inputs for mainstream valuation models. Yet that approach is starting to change. After Meta’s earnings announcement, its CDS hit an all-time high, its outlook was lowered by S&P, and Goldman Sachs publicly drew attention to differences in rating methodologies. Across credit markets, rating agencies and sell-side research, off-balance-sheet commitments are now beginning to be assessed as a separate variable.

A systemic repricing may be triggered less by the absolute scale of commitments than by the divergence between their growth and that of operating cash flow. As long as operating cash flow covers on-balance-sheet obligations and a portion of off-balance-sheet commitments, these liabilities remain manageable future expenditure. But when commitment growth outpaces operating cash flow growth, their inflexibility turns a “long-term arrangement” into “near-term financing pressure”.

The bill for Intel has already come due, reflected in its 30% fall on 30 October. The $4.11 trillion represents the bill still to come. Investors needed one day to price the first. The time required to price the second will depend on when the first transmission chain progresses from “cash receipts falling short of expectations” to “parent companies providing additional funding”.