AI 버전의 서브프라임 모기지 위기? 1.8조 달러 규모의 우발부채, 이번 광란의 시한폭탄으로 부상
- 핵심 시각: AI 인프라 구축은 전례 없는 규모의 부채 확장을 촉발하고 있으며, 이 중 약 1.8조 달러의 우발부채(구매 확약, 미효성 리스 등)가 대차대조표에 반영되지 않아 막대한 잠재적 금융 리스크를 구성하고 있으며, 그 취약성은 AI 상업화가 기대에 미치지 못할 때 드러날 것입니다.
- 핵심 요소:
- 2026년 5월 말까지 AI 관련 채권 발행액은 2360억 달러에 달해 전년 동기 대비 357% 증가했습니다. 모건스탠리는 연간 발행액이 5700억 달러를 돌파할 것으로 예상합니다.
- 하이퍼스케일 클라우드 기업의 총 레버리지 비율은 단 두 분기 만에 0.9배에서 1.8배로 급등, 이미 에너지 업종을 추월했으며 현금 흐름은 거의 제로에 가깝습니다.
- 약 1.8조 달러 규모의 우발부채는 세 부분으로 구성됩니다: 구매 확약 9820억 달러(부채 미계상), 미효성 리스 확약 8220억 달러, 미지급금 1100억 달러.
- SPV를 통한 순환 파이낸싱 구조(예: Apollo와 Blackstone의 Anthropic 대상 350억 달러 '칩 담보' 거래)는 기업의 레버리지를 공급망과 사모 신용 시스템으로 이전시켰습니다.
- 감가상각 압력은 지연되었습니다: 건설 중인 자산 잔액이 급증했으며(오라클 전년 동기 대비 200% 증가), 향후 4년간 빅4의 누적 감가상각비는 5200억 달러를 초과하여 수익성에 타격을 줄 수 있습니다.
- 매출 예측 상향 조정 폭은 자본 지출 예측 상향 조정 폭(예: 오라클 자본 지출 예측 175% 상향)에 크게 미치지 못해 구조적 불일치가 존재합니다.
- 골드만삭스는 2027년 하이퍼스케일 클라우드 기업의 자본 지출이 1.1조 달러에서 1.4조 달러에 달할 수 있다고 예측하지만, 이는 LLM이 가격 결정력을 유지하고 기업 고객 충성도를 확보한다는 전제 조건이 충족되어야 합니다.
Original author: Bu Shuqing
Original source: Wall Street Journal
Amid the AI infrastructure building boom, a debt expansion of unprecedented scale is quietly taking shape – and the most dangerous part of it has never appeared on any balance sheet.
A recent Goldman Sachs report predicts that capital expenditures by hyper-scale cloud companies will reach $1.1 trillion to $1.4 trillion by 2027, far exceeding market consensus. However, according to in-depth research by Morgan Stanley, this already staggering figure is only the tip of the iceberg.
Nearly $1 trillion in purchase commitments, over $800 billion in non-cancelable lease contracts, and tens of billions in supplier financing arrangements collectively constitute approximately $1.8 trillion in off-balance-sheet exposure – liabilities that exist outside the balance sheet but genuinely lock in future cash outflows.
The market has not yet fully priced in these risks.
Morgan Stanley warns that the leverage ratio of hyper-scale cloud companies has soared from 0.9x to 1.8x in just two quarters, with the growth rate of capital expenditures consistently outpacing that of revenue and free cash flow. Meanwhile, the true impact of depreciation pressure has yet to arrive.
At the same time, private credit institutions represented by Apollo and Blackstone are shifting leverage to the supply chain level through SPVs (Special Purpose Vehicles), creating a highly circular and opaque financing structure. If AI commercialization falls short of expectations, or if enterprise customers massively pivot to cheaper alternatives, the vulnerability of the entire financing chain will be laid bare.
The Debt Issuance Frenzy: AI Has Become the Biggest Variable in Public Markets
According to Morgan Stanley's latest "AI Debt Financing Tracker," global AI-related bond issuance had reached $236 billion by the end of May 2026, a staggering 357% increase compared to the same period in 2025.
Morgan Stanley expects total AI debt issuance for the full year to exceed $570 billion, with the pace accelerating further in the second half of the year as financing needs for capital expenditures are concentratedly released.
In April alone, AI-related bond issuance exceeded $74 billion, setting a new high for the year. Project financing structures (used for data center construction) accounted for 85% of high-yield bond supply and 40% of investment-grade bond supply. Meanwhile, the five hyper-scale cloud companies – Amazon, Meta, Google, Microsoft, and Oracle – now constitute 4% of the entire investment-grade bond index.
On the leverage front, the gross leverage ratio of hyper-scale cloud companies has risen from 0.9x in Q3 2025 to the current 1.8x, increasing by approximately 0.3x per quarter, surpassing the leverage level of the entire energy sector.
Morgan Stanley notes that due to supply pressure, the relevant credit spreads have drifted from the AA range to the A range and could widen further. Meta's credit spread is now wider than the CDX IG benchmark.
In terms of free cash flow, Morgan Stanley predicts that Amazon and Meta's free cash flow in 2026 will approach zero or turn negative, forcing incremental financing to rely almost entirely on new debt.
$1.8 Trillion Off-Balance-Sheet Exposure: Invisible Liabilities, Locked-In Cash Outflows
Todd Castagno of Morgan Stanley's Global Valuation, Accounting & Tax team points out in a report that focusing solely on capital expenditure numbers severely underestimates the true financial commitments of the AI construction cycle. Beyond disclosed capital expenditures, there are three key types of off-balance-sheet exposures:
Purchase commitments amount to approximately $982 billion. The long-term procurement contracts of hyper-scale cloud companies and Nvidia total nearly $1 trillion. According to accounting standards, unless a company expects a contract loss, these obligations are not recorded as liabilities until the goods are delivered. Therefore, nearly a trillion dollars in future cash outflows currently does not appear as any liability on the balance sheet.
Notably, Nvidia's own inventory and purchase obligations have risen to approximately 32% of consensus revenue forecasts for fiscal year 2027, far above the historical range of 15% to 20%, extending supply chain commitment risk to the chip supplier side.
Non-cancelable lease commitments amount to approximately $822 billion. Over $800 billion in lease contracts have been signed but have not yet commenced, and are therefore not included in current lease liabilities. Furthermore, arrangements such as variable lease payments, renewal options, and residual value guarantees also exist outside the balance sheet.
Morgan Stanley estimates that if finance leases were included, Microsoft's capital expenditure as a percentage of sales would jump from 33%/50% (FY2026/FY2027) to 44%/64%, and Oracle's could rise from 76%/115% to 101%/189%.
Unpaid capital expenditures within accounts payable stand at approximately $110 billion. Days Payable Outstanding (DPO) for hyper-scale cloud companies has lengthened significantly – Oracle's DPO increased by 370% year-over-year, Meta's by 73%, and Microsoft's by 69%. This means the entire supply chain is effectively fronting the costs for AI construction, with suppliers bearing the liquidity pressure that should belong to the buyers.
SPVs and Circular Financing: Leverage Shifted into 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 transaction for Anthropic, which perfectly illustrates the logic of this model:
Broadcom provides the backstop for the SPV, Anthropic uses the raised funds to purchase Google chips manufactured by Broadcom, while Google holds a 14% equity stake in Anthropic; Morgan Stanley, which arranged the transaction, also provides loans to the participating investors.
Morgan Stanley's AI Ecosystem Financing Correlation Map reveals a multi-layered circular relationship involving customers, investors, supplier financing, and buybacks between OpenAI, Oracle, Nvidia, Microsoft, CoreWeave, AMD, and Amazon. The same funds circulate repeatedly among a few key entities, with SPVs serving as the core tool to facilitate this cycle.
It is reported that Apollo's insurance subsidiary, Athene, is particularly active in the above structure – raising funds by selling annuities to retirees and then injecting the capital into SPVs to participate in AI infrastructure financing.
This model shifts leverage from the visible balance sheets of hyper-scale cloud companies to the supplier and private credit ecosystem, making the true systemic risk exposure difficult for external observers to identify and aggregate.
The Depreciation Cliff and Monetization Gap: The Deferred Impact
Current financial data contains a systematic optimistic bias. A large amount of capital expenditure is currently recorded as "Construction in Progress" (CIP) and has not yet begun to depreciate. This artificially inflates reported profit margins and underestimates future expense pressures.
Oracle, Meta, and Google's CIP balances have increased by approximately 200%, 90%, and 55% year-over-year, respectively.
Once these assets are gradually transferred to depreciation, the impact will be concentrated and released.
Morgan Stanley predicts that the cumulative depreciation for Microsoft, Oracle, Meta, and Google over the next three years will exceed $520 billion. For Oracle, depreciation as a percentage of revenue could rise from the current 7% to 28% by fiscal year 2028; for Meta, it could increase from 9% to 19%.
In this context, the only path to maintaining profit margins is proportional and significant revenue growth – yet the upward revisions to revenue forecasts are currently lagging far behind the upward revisions to capital expenditure forecasts.
Data shows that consensus estimates for Google's 2026 capital expenditure have been revised up 139% from a year ago, while Meta and Amazon's have been revised up 85% and 81%, respectively. Oracle saw the largest revision, a 175% increase.
At the same time, the magnitude of revenue forecast revisions has clearly lagged. A structural mismatch, where capital expenditure precedes commercial realization, is clearly visible.
Furthermore, over $2 trillion in Remaining Performance Obligations (RPO) are highly concentrated among a small number of large, long-term contracts. This counterparty concentration risk cannot be ignored – a problem for any major participant in the circular system could trigger a chain reaction.
A Timing Mismatch, Not an Immediate Solvency Crisis
Morgan Stanley concludes that the aforementioned risks currently do not constitute an imminent solvency crisis, but rather a confluence of timing mismatches and information disclosure gaps: depreciation pressure is deferred, capital expenditure outpaces monetization progress, leverage is shifted to suppliers and the private credit stratum, and the comparability of capital intensity between companies is significantly undermined by differences in accounting classifications.
Hyper-scale cloud companies are clearly aware of the limited window of current market sentiment and are seizing the opportunity to maximize the scale of their financing.
Goldman Sachs analyst Ryan Hammond points out that if AI infrastructure investment reaches 2% to 3% of GDP, drawing an analogy to the historical construction cycles of the railroad and automotive industries, capital expenditure could reach $1.1 trillion in 2027. In an extreme scenario, considering the cash flows of hyper-scale cloud companies and the capacity of the investment-grade credit market, the upper limit could reach $1.4 trillion.
However, all of this is premised on Large Language Models (LLMs) being able to sustain token pricing and maintain sufficient enterprise customer stickiness. A growing number of enterprises are turning their attention to AI products with comparable performance but significantly lower prices.
If a structural shift occurs on the demand side, the current carefully constructed financing system will face a fundamental stress test.
