The rapid expansion of AI infrastructure financing is accelerating, with core risks concentrated in approximately $1.8 trillion of off-balance-sheet exposures. These include nearly $1 trillion in purchase commitments and over $800 billion in leases and supply chain financing. Concurrently, AI-related debt issuance for the full year is projected to surpass $570 billion and is accelerating. The combination of rising leverage and lagging depreciation is causing risks to spread from off-balance-sheet entities into the broader credit system.

Amid the fervor for building AI infrastructure, a debt expansion of unprecedented scale is quietly taking shape. The most perilous part of this expansion, however, never appears on any company's balance sheet.

A recent report from Goldman Sachs forecasts that capital expenditure by hyperscale cloud companies could reach $1.1 to $1.4 trillion by 2027, far exceeding current market consensus. Yet, according to an in-depth study by Morgan Stanley, this already staggering figure is merely the tip of the iceberg.

Key Off-Balance-Sheet Exposures

Nearly $1 trillion in purchase commitments, over $800 billion in non-operational lease contracts, and tens of billions in supplier financing arrangements collectively form roughly $1.8 trillion in off-balance-sheet exposures. These liabilities exist outside of balance sheets but effectively lock in future cash outflows.

The market has not yet fully priced in these risks. Morgan Stanley warns that the leverage ratio of hyperscale cloud companies has surged from 0.9x to 1.8x in just two quarters. Their capital expenditure growth continues to outpace revenue and free cash flow growth, and the full impact of depreciation pressure has yet to be felt.

Simultaneously, private credit firms like Apollo and Blackstone are transferring leverage to the supply chain level through Special Purpose Vehicles (SPVs), creating a highly circular and opaque financing structure. Should AI commercialization fall short of expectations, or if enterprise clients shift en masse to cheaper alternatives, the fragility of this entire financing chain could be exposed.

The Surge in Debt Issuance

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

Morgan Stanley projects full-year AI debt issuance will exceed $570 billion, with the pace accelerating in the second half of the year as capital expenditure financing demands are concentrated.

In April alone, AI-related bond issuance surpassed $74 billion, a yearly high. Project financing structures (for data center construction) accounted for 85% of high-yield bond supply and 40% of investment-grade bond supply. Meanwhile, five hyperscale cloud firms—Amazon.com Inc (NASDAQ: AMZN), Meta Platforms Inc (NASDAQ: META), Alphabet Inc (NASDAQ: GOOGL), Microsoft Corporation (NASDAQ: MSFT), and Oracle Corporation (NYSE: ORCL)—now constitute 4% of the entire investment-grade bond index.

On the leverage front, the gross leverage ratio for these hyperscale cloud companies has risen from 0.9x in Q3 2025 to 1.8x currently, increasing by approximately 0.3x per quarter, surpassing the leverage level of the entire energy sector.

Morgan Stanley notes that due to supply pressure, related credit spreads have drifted from the AA range to the A range and may widen further. Meta's credit spreads are currently wider than the CDX IG benchmark.

Regarding free cash flow, Morgan Stanley predicts that Amazon and Meta will see their free cash flow approach zero or turn negative in 2026, meaning incremental financing will rely almost entirely on new debt.

Unseen Liabilities, Locked-In Cash Outflows

Todd Castagno of Morgan Stanley's Global Valuation, Accounting & Tax team points out that focusing solely on capital expenditure numbers severely underestimates the true financial commitments of the AI build-out cycle. Beyond disclosed capital expenditure, there are three key types of off-balance-sheet exposures:

1. Purchase Commitments (~$982 billion): Long-term procurement contracts by hyperscale cloud firms and NVIDIA Corp (NASDAQ: NVDA) total nearly $1 trillion. Under accounting rules, unless a company expects a contract loss, these obligations are not recorded as liabilities until goods are delivered. Thus, nearly $1 trillion in future cash outflows are not currently reflected as liabilities on any balance sheet. Notably, NVIDIA's own inventory and purchase obligations have risen to about 32% of consensus revenue forecasts for fiscal 2027, well above the historical range of 15-20%, indicating supply chain commitment risks are extending to chip suppliers.

2. Non-Operating Lease Commitments (~$822 billion): Over $800 billion in lease contracts have been signed but are not yet operational and are not included in current lease liabilities. Additionally, variable lease payments, renewal options, and residual value guarantees also remain off the balance sheet. Morgan Stanley estimates that if finance leases were included, Microsoft's capex-to-sales ratio would jump from 33%/50% (FY2026/2027) to 44%/64%, while Oracle's could rise from 76%/115% to 101%/189%.

3. Unpaid Capital Expenditure in Accounts Payable (~$110 billion): Days Payable Outstanding (DPO) for hyperscale cloud firms has significantly lengthened—Oracle up 370% year-over-year, Meta up 73%, Microsoft up 69%—meaning the entire supply chain is effectively financing the AI build, with suppliers bearing liquidity pressure that should be the buyer's responsibility.

Leverage Shifted into the Shadows

Another core dimension of off-balance-sheet risk is the circular financing structure built via SPVs. This week, a $35 billion "chip-collateralized" private credit deal for Anthropic by Apollo and Blackstone exemplifies this model: Broadcom Inc provides backing for the SPV, Anthropic uses the funds to purchase Google chips manufactured by Broadcom, while Alphabet holds a 14% stake in Anthropic. Morgan Stanley (NYSE: MS), which arranged the deal, also provides loans to participating investors.

Morgan Stanley's AI ecosystem financing map reveals multiple, circular client, investor, supplier financing, and repurchase relationships among entities like OpenAI, Oracle, NVIDIA, Microsoft, CoreWeave, Advanced Micro Devices Inc (NASDAQ: AMD), and Amazon. The same capital circulates repeatedly among a few key players, with SPVs as the core tool enabling this cycle.

Reportedly, Apollo's insurance subsidiary, Athene, is particularly active in such structures—raising funds by selling annuities to retirees and then channeling that capital into SPVs for AI infrastructure financing. This model shifts leverage from the visible balance sheets of hyperscale cloud companies to suppliers and the private credit ecosystem, making the true systemic risk exposure difficult for external observers to identify and aggregate.

The Depreciation Cliff and Monetization Gap

Current financial data exhibits systematic optimism bias. A significant portion of capital expenditure is currently recorded as "Construction in Progress" (CIP), not yet subject to depreciation. This artificially inflates reported profit margins and underestimates future expense pressure. CIP balances at Oracle, Meta, and Google have grown approximately 200%, 90%, and 55% year-over-year, respectively.

The impact will be concentrated once these assets begin to depreciate. Morgan Stanley predicts that 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 7% currently to 28% in fiscal 2028; for Meta, it could increase from 9% to 19%.

In this context, the only path to maintaining profit margins is a simultaneous, substantial increase in revenue. However, the upward revisions to revenue forecasts are lagging far behind the upward revisions to capital expenditure forecasts. Data shows consensus 2026 capex forecasts have been raised by 139% for Google, 85% for Meta, 81% for Amazon, and a massive 175% for Oracle compared to a year ago.

Meanwhile, revisions to revenue forecasts are noticeably lagging, revealing a clear structural mismatch where capital expenditure is outpacing commercialization. Furthermore, over $2 trillion in remaining performance obligations (RPO) is highly concentrated in a few large, long-term contracts, posing significant counterparty concentration risk. A problem with any major participant in this circular system could trigger a chain reaction.

A Crisis of Timing, Not Solvency

Morgan Stanley's conclusion is that these risks do not currently constitute an imminent solvency crisis, but rather a series of timing mismatches and disclosure gaps: depreciation pressure is deferred, capital expenditure outpaces monetization, leverage shifts to suppliers and private credit, and the comparability of capital intensity across companies is compromised by accounting classification differences.

Hyperscale cloud companies are clearly aware of the limited window of favorable market sentiment and are seizing the moment to maximize financing. Goldman Sachs analyst Ryan Hammond notes that if AI infrastructure investment reaches 2-3% of GDP, analogous to historical construction cycles for railroads and the auto industry, 2027 capital expenditure could hit $1.1 trillion. In an extreme scenario, considering hyperscale cloud cash flows and investment-grade credit market capacity, the upper limit could be $1.4 trillion.

However, all this hinges on Large Language Models (LLMs) being able to sustain or increase token pricing and maintain sufficient enterprise customer loyalty. An increasing number of enterprises are looking towards AI products with comparable performance but significantly lower prices. Should a structural shift occur on the demand side, the current, intricately constructed financing system will face a fundamental stress test.