Nine of the biggest names in technology have quietly accumulated about $3.1 trillion in financial commitments that do not appear on their balance sheets, with nearly all of the exposure tied to artificial intelligence infrastructure. The total rose by $1.3 trillion in just three months, underscoring how rapidly AI-related buildouts are translating into long-dated contractual obligations rather than only reported debt.

The companies named are Alphabet, Amazon, Meta, Microsoft, Oracle, Nvidia, Broadcom, AMD, and SpaceX. Their combined off-balance-sheet obligations now exceed their reported on-balance-sheet leases and long-term debt by roughly three times, even though their total annual capital expenditure is about $600 billion.

What is included in the off-balance-sheet total

The $3.1 trillion figure is mainly made up of two categories. Roughly $1.2 trillion consists of uncommenced leases, or rental agreements for data centers, land, and facilities that have not yet taken effect. The remaining $1.9 trillion is in purchase obligations, which are contractual commitments to buy chips, equipment, and other infrastructure components.

Morgan Stanley said the total exceeds $3.1 trillion when guarantees and leases from hyperscalers and chip companies are included. According to the report, hyperscalers’ guarantees alone are more than $2.7 trillion. These commitments are disclosed in the footnotes of quarterly filings, where investors can see them but they do not sit alongside traditional debt on the face of the balance sheet.

A Nikkei study in July estimated that hidden obligations among five major technology companies totaled $1.65 trillion, an eightfold increase over four years.

Alphabet’s commitments surged in one quarter

Alphabet provided one of the clearest examples of how quickly these obligations are growing. The Google parent’s purchase commitments increased from $332 billion at the end of March to $811 billion by the end of June, a 152% jump in a single quarter.

Cash flow and disclosure concerns

Some hyperscalers have already moved into negative free cash flow as capital spending and commitment-related expenses outpace revenue growth.

Because off-balance-sheet obligations do not trigger the same covenants or rating agency scrutiny as traditional debt, they can accumulate with relatively little external pressure. If AI demand does not materialize at scale, uncommenced leases would still need to be honored or renegotiated, while purchase obligations would still need to be fulfilled or written off.

For context, $3.1 trillion is larger than the United Kingdom’s GDP and roughly five times the annual capital expenditure currently spent by the nine companies. The total also increased by more than 70% in a single quarter, adding a new layer of attention to how much of the AI infrastructure push is being financed through future commitments rather than current spending alone.

Credit analysts have begun questioning whether current disclosures are sufficient. Morgan Stanley’s decision to publish an explicit tally suggests growing concern on the sell side about the gap between reported liabilities and actual contractual exposure.