AI version of the "Subprime Crisis"? Under the frenzy, 1.8 trillion hidden debts are accumulating in the shadows

Title: "The Taste of the AI Subprime Crisis? $1.8 Trillion Off-Balance-Sheet Exposure Becoming a Time Bomb in This Round of Frenzy"
Author: Bu Shuqing, Wall Street Insights

Author: Lydong BlockBeats

Source:

Reprinted from Mars Financial

Amid the boom in AI infrastructure development, an unprecedented scale of debt expansion is quietly taking shape—and the most dangerous part has never appeared on any balance sheet.

Goldman Sachs' latest report forecasts that by 2027, the capital expenditure of mega-scale cloud computing companies will reach $1.1 trillion to $1.4 trillion, far exceeding market consensus. However, according to in-depth research by Morgan Stanley, this already jaw-dropping figure is just the tip of the iceberg.

Nearly $1 trillion in procurement commitments, over $800 billion in unexecuted lease contracts, and hundreds of billions in vendor financing arrangements collectively constitute approximately $1.8 trillion in off-balance-sheet exposure—these liabilities are outside the balance sheet but truly lock in future cash outflows.

The market has yet to fully price in these risks.

Morgan Stanley warns that the leverage ratio of mega-scale cloud companies has soared from 0.9 times to 1.8 times within just two quarters, with capital expenditure growth outpacing revenue and free cash flow growth, and the true impact of depreciation pressure has yet to arrive.

Meanwhile, private credit institutions such as Apollo and Blackstone are shifting leverage to the supply chain level through SPVs (Special Purpose Vehicles), forming highly cyclical and opaque financing structures. If AI commercialization progresses slower than expected, or if corporate clients shift en masse to cheaper alternatives, the fragility of the entire financing chain will be exposed.

Debt Issuance Frenzy: AI Has Become the Largest Variable in the Public Market

According to Morgan Stanley's latest "AI Debt Financing Tracking Report," as of the end of May 2026, global AI-related bond issuance has reached $236 billion, a 357% increase compared to the same period in 2025.

Morgan Stanley expects total AI debt issuance for the year to surpass $570 billion, with the pace further accelerating in the second half as capital expenditure financing needs are concentrated.

In April alone, AI-related bond issuance exceeded $74 billion, hitting a new high for the year, with project financing structures (used for data center construction) accounting for 85% of high-yield bond supply and 40% of investment-grade bonds. Meanwhile, Amazon, Meta, Google, Microsoft, and Oracle, the five mega-scale cloud companies, now account for 4% of the entire investment-grade bond index.

On the leverage front, the overall gross leverage ratio of mega-scale cloud companies has risen from 0.9 times in Q3 2025 to the current 1.8 times, increasing by about 0.3 times each quarter, surpassing the leverage levels of the entire energy sector.

Morgan Stanley points out that, under supply pressure, credit spreads have drifted from the AA range to the A range and may widen further. Meta's credit spread is currently wider than the CDX IG benchmark.

In terms of free cash flow, Morgan Stanley forecasts that Amazon and Meta's free cash flow in 2026 will approach zero or turn negative, meaning incremental financing will almost entirely rely on new debt.

$1.8 Trillion Off-Balance-Sheet Exposure: Invisible Liabilities Locking in Cash Outflows

Todd Castagno from Morgan Stanley's global valuation, accounting, and tax team notes in the report that focusing solely on capital expenditure figures significantly underestimates the true financial commitments of the AI build cycle. Beyond disclosed capital expenditures, there are three key types of off-balance-sheet exposure:

Procurement commitments of about $982 billion. The long-term procurement contracts of mega-scale cloud companies and Nvidia total nearly $1 trillion. According to accounting standards, unless the company expects contract losses, these obligations are not recorded as liabilities before goods are delivered. Therefore, nearly $1 trillion in future cash outflows are not reflected on any balance sheet.

Notably, Nvidia's inventory and procurement obligations have risen to about 32% of the consensus revenue forecast for FY2027, well above the historical range of 15% to 20%, extending supply chain commitment risks to chip suppliers.

Unexecuted lease commitments of approximately $822 billion. Over $800 billion in lease contracts have been signed but not yet executed, not included in current lease liabilities. Additionally, arrangements such as variable lease payments, renewal options, and residual value guarantees also remain off-balance-sheet.

Morgan Stanley estimates that if finance leases are included, Microsoft's capital expenditure as a percentage of sales would jump from 33%/50% (FY2026/2027) to 44%/64%, and Oracle could rise from 76%/115% to 101%/189%.

Unpaid capital expenditures in accounts payable amount to about $110 billion. The days payable outstanding (DPO) for mega-scale cloud companies has significantly lengthened—Oracle increased by 370% year-over-year, Meta by 73%, and Microsoft by 69%—indicating the entire supply chain is effectively financing AI development, with suppliers bearing liquidity pressures that should be borne by buyers.

SPV and Circular Financing: Leverage Moving to the Shadows

Another core dimension of off-balance-sheet risk is the circular financing structure built through SPVs.

This week, Apollo and Blackstone jointly completed a $35 billion "chip-backed" private credit deal for Anthropic, exemplifying this model's operational logic:

Broadcom provides backing for the SPV, Anthropic uses the raised funds to purchase Google chips manufactured by Broadcom, while Google holds a 14% stake in Anthropic; Morgan Stanley arranged the deal and also provided loans to participating investors.

Morgan Stanley's AI ecosystem financing map shows multiple cyclical relationships among OpenAI, Oracle, Nvidia, Microsoft, CoreWeave, AMD, and Amazon—client, investor, supplier financing, and buyback cycles—where the same funds circulate repeatedly among a few entities, with SPVs serving as the core tool enabling this cycle.

Notably, Apollo's insurance subsidiary Athene is particularly active in this structure—raising funds by selling annuities to retirees and injecting the capital into SPVs for AI infrastructure financing.

This model shifts leverage from the visible balance sheets of mega-scale cloud companies to the supply chain and private credit ecosystems, making systemic risk exposure difficult for outsiders to observe and aggregate.

Depreciation Cliff and Monetization Gap: Delayed Impacts

Current financial data contain systemic optimism bias. Much capital expenditure is recorded as "Construction in Progress" (CIP), not yet depreciated, artificially inflating reported profit margins and underestimating future expense pressures.

Oracle, Meta, and Google's CIP balances have increased by approximately 200%, 90%, and 55%, respectively, year-over-year.

Once these assets begin to depreciate, the impact will be released in a concentrated manner.

Morgan Stanley forecasts that Microsoft's, Oracle's, Meta's, and Google's cumulative depreciation over the next three years will exceed $520 billion. For example, Oracle's depreciation as a percentage of revenue could rise from the current 7% to 28% by FY2028; Meta's could increase from 9% to 19%.

In this context, the only way to maintain profit margins is through substantial revenue growth—yet the upward revisions of revenue forecasts lag far behind those of capital expenditure forecasts.

Data shows that Google's capital expenditure consensus forecast for 2026 has been raised by 139% over the past year, while Meta and Amazon have increased by 85% and 81%, respectively. Oracle's revision is the largest, at 175%.

Meanwhile, the revisions to revenue forecasts are significantly lagging, revealing a structural mismatch where capital expenditure outpaces commercialization progress.

Additionally, over $2 trillion in remaining performance obligations (RPO) are highly concentrated among a few large long-term contracts, with counterparty risk that cannot be ignored—any issues among key participants in the cycle could trigger a chain reaction.

Timing Mismatch Rather Than Immediate Liquidity Crisis

Morgan Stanley concludes that these risks do not currently constitute an imminent solvency crisis but are a combination of timing mismatches and information disclosure gaps: deferred depreciation pressures, capital expenditure outpacing monetization, leverage shifting to suppliers and private credit, and comparability issues among companies due to accounting classification differences.

Mega-scale cloud companies are clearly aware of the limited window of current market sentiment and are rushing to maximize financing scale.

Goldman Sachs analyst Ryan Hammond points out that if AI infrastructure investment reaches 2% to 3% of GDP—drawing parallels with the historical construction cycles of railroads and the automotive industry—capital expenditure could reach $1.1 trillion by 2027; in an extreme scenario, considering cash flows of mega-scale cloud companies and the capacity of the investment-grade credit market, the upper limit could be $1.4 trillion.

However, all of this hinges on large language models (LLMs) continuing to improve token pricing and maintaining enough enterprise customer stickiness. More companies are turning their attention to AI products that are nearly as performant but significantly cheaper.

If there is a structural shift in demand, the carefully constructed financing system will face a fundamental stress test.