Trillions in hidden obligations are emerging alongside the AI infrastructure boom, as a wave of off-balance-sheet commitments accumulates across the technology sector. Roughly $1.8 trillion in obligations now sit outside traditional corporate accounts, spanning purchase agreements, pending leases, and unpaid liabilities. These exposures could strain liquidity if revenue from AI services fails to scale as anticipated, with potential ripple effects across the broader technology supply chain.
Capital spending forecasts outpace expectations
Capital expenditure by major cloud providers is surging beyond earlier expectations. Goldman Sachs estimates that capex could reach between $1.1 trillion and $1.4 trillion by 2027, driven largely by investments in data centers, chips, and supporting infrastructure. Underpinning this buildout is a vast array of contractual obligations that do not fully appear on balance sheets.
Morgan Stanley data highlights nearly $982 billion in purchase obligations, $822 billion in lease commitments, and about $110 billion in unpaid payables supporting current investment plans. These commitments lock firms into long-term spending trajectories, regardless of how quickly AI revenue materializes.
AI-linked bond issuance has also accelerated sharply. By May 2026, total issuance reached $236 billion, a 357% increase from the prior year. Full-year issuance is expected to exceed $570 billion, with April alone accounting for $74 billion. Data center financing dominates this market, representing 85% of high-yield and 40% of investment-grade issuance during that month.
Rising leverage and tightening cash flow
Balance sheet leverage is rising rapidly among large cloud operators. The gross leverage ratio has doubled from 0.9 to 1.8 within two quarters, surpassing levels often associated with capital-intensive sectors such as energy. This increase reflects heavy borrowing to fund AI-related infrastructure.
Morgan Stanley estimates that free cash flow is deteriorating quickly, with some firms approaching zero or turning negative in 2026. As internal cash generation weakens, new investment is increasingly reliant on external financing. This shift heightens sensitivity to changes in credit conditions and investor appetite for AI-linked debt.
Complex financing shifts risk off balance sheets
A growing portion of new leverage is being routed through special purpose vehicles and other structured entities, allowing companies to keep certain debts off their primary balance sheets. These structures can obscure the true level of risk borne by core operating entities.
One example involves a $35 billion chip-backed private loan for Anthropic, arranged by Apollo and Blackstone. The transaction channels insurance capital into semiconductor purchases backed by Broadcom, using a structure that distances the risk from traditional corporate debt metrics while still expanding overall system leverage.
Research indicates that a small group of technology and hardware firms are repeatedly recycling capital through similar structures. Rather than eliminating risk, this approach redistributes it into private credit markets and insurance portfolios, making it harder for investors and regulators to track aggregate exposures.
Deferred costs mask profitability pressures
Deferred cost recognition is another mechanism masking the full impact of the AI infrastructure surge. Capital expenditures categorized as construction-in-progress have increased sharply, allowing companies to delay recognizing depreciation expenses on unfinished assets.
Oracle, for example, has reported a 200% year-on-year increase in construction-in-progress, reflecting rapid data center expansion. This accounting treatment supports near-term profit margins by postponing depreciation, but it also builds a pipeline of future costs that will eventually flow through income statements.
Analysts estimate that depreciation expenses for major cloud firms will exceed $520 billion over the next three years. In some cases, depreciation as a share of revenue could triple by 2028, compressing margins even if top-line growth remains solid.
Revenue growth lags behind investment surge
Revenue expectations have not kept pace with the escalation in capital spending. Google’s 2026 capex projection has been revised higher by 139%, while Meta and Amazon have increased their forecasts by more than 80%. Oracle’s outlook has risen even more sharply, with a 175% increase in projected spending.
This widening gap between revenue and investment underlines a central vulnerability of the current AI cycle. Despite significant hype and adoption of AI tools, AI-driven revenue remains modest relative to the scale of ongoing infrastructure commitments, raising questions about long-term returns on invested capital.
Concentration risk and funding imbalances grow
More than $2 trillion in remaining performance obligations are now concentrated in a limited set of long-term contracts. This concentration heightens counterparty risk, as the failure of a small number of key customers or partners could disproportionately impact cash flows and project viability.
Analysts identify funding mismatches, delayed cost recognition, and opaque leverage structures as the primary weaknesses in the current AI investment cycle, rather than immediate solvency issues. Shorter-term funding is often being used to finance long-lived assets, creating duration mismatches that could become problematic if credit conditions tighten.
Goldman Sachs notes that if AI infrastructure spending reaches 2% to 3% of global GDP, total investment could climb to around $1.4 trillion. At that scale, AI infrastructure would rival past industrial expansions, amplifying both the potential rewards and the systemic risks.
Credit markets signal emerging stress
Early signs of strain are appearing in credit markets. The rapid increase in borrowing has begun to widen credit spreads for some major issuers, signaling growing investor concern about leverage and return profiles. At the same time, hedging activity in credit default swaps tied to AI-exposed firms is rising.
The gap between spending and realized returns remains substantial. AI-related revenue was approximately $60 billion in 2025, compared with about $400 billion in capital expenditures. Much of the associated debt resides within private credit funds and structured vehicles, rather than on traditional corporate balance sheets.
This arrangement raises the possibility that any slowdown in AI adoption or pricing power could trigger stress in segments of the financial system that are less transparent and less regulated, complicating efforts to manage potential contagion.
Outlook hinges on AI monetization
The durability of the AI infrastructure boom ultimately depends on the industry’s ability to monetize AI services at scale. Sustaining current investment levels will require robust demand, stable or rising pricing, and a steady flow of commercially viable applications that justify the underlying capital commitments.
If revenue growth fails to keep up with the pace of investment, the combination of rising debt, deferred costs, and complex financing structures could expose vulnerabilities that have so far remained largely hidden. In that scenario, off-balance-sheet obligations and concentrated performance commitments may become focal points for investors, regulators, and credit markets assessing the true risk profile of the AI expansion.
Want to hedge against AI’s leveraged boom? Explore diversified crypto strategies with Toobit’s digital assets guide today.
Disclaimer: The content on this page is provided for general informational purposes only and does not represent the views or financial advice of Toobit. We make no guarantees regarding the accuracy or completeness of this information and shall not be held liable for any errors, omissions, or outcomes resulting from its use. Investing in digital assets involves risk; users should independently evaluate their financial situation and the risks involved. For further details, please consult our Terms of Service and Risk Disclosure.
