AI infrastructure financing is shaping a new architecture in credit markets
AI summary card
AI infrastructure financing is shaping a new architecture in credit markets
Morgan Stanley believes hyperscale cloud providers have substantially raised AI capex guidance, and credit markets are accommodating data center, GPU, and energy infrastructure demand through lending, asset-based financing, ABS, high-yield debt, and project finance structures.
- Expected capex for the five largest hyperscale cloud providers has been revised significantly upward: about $800 billion in 2026 and about $1.16 trillion in 2027, above earlier forecasts of about $450 billion in each year made a year ago.
- OpenRouter shows global weekly token usage has grown about 350% since early January, from roughly 6 trillion to 28 trillion, reflecting the rapid expansion of underlying compute demand.
- GPU financing has started to shift from almost entirely equity-capital-led support to broadly syndicated lending and asset-based financing, and may further evolve into ABS structures.
- Financing structures are becoming integrated across project finance, mezzanine-style arrangements, residual value support, and high-yield issuance backed by hyperscale cloud lease agreements, blurring the boundaries between public and private financing and between corporate and project finance.
- Grid interconnection, power equipment, skilled labor, permitting delays, and political and regulatory frictions at local, national, and international levels are becoming key constraints on the pace of AI infrastructure buildout and financing.
Report interpretation
Overview
The report discusses how the capex wave from AI and data center buildout is reshaping the credit market financing ecosystem. Morgan Stanley notes that the required funding scale and drawdown speed for AI infrastructure are meaningfully above prior expectations, and the financing scope has expanded beyond data center buildout itself to include GPUs and supporting energy infrastructure. The role of credit markets is expanding, with structural innovation across public markets, private markets, project finance, asset financing, and high-yield debt.
Core views
The core view is that AI infrastructure capex is expanding exponentially, and credit markets will be a durable source of funding throughout this build cycle. Financing innovation can broaden the investor base, reduce financing friction, and dissolve traditional boundaries between public and private credit as well as corporate versus project financing. But this cycle is not unconstrained: power shortages, grid access, equipment supply, permitting, labor, insurance capital supply, interest-rate changes, and geopolitics may all influence financing availability and buildout pace at the margin.
Analysis framework
The report uses a thematic macro and credit-market framework, combining hyperscale cloud-provider capex forecasts, proxy indicators such as global weekly token usage for compute demand, financing tool evolution, energy constraints, and sources of capital supply to assess the trajectory of the AI infrastructure financing cycle. Its analytical focus is on the interaction between capex demand, financing channel expansion, structural innovation, and potential bottlenecks, rather than valuation of individual companies.
Methodology notes
Decompose AI buildout demand into data centers, GPUs, energy supply, and capital supply, and observe the corresponding credit financing instruments.
The report explains how AI capex is reshaping credit market boundaries by comparing channels such as public credit, private credit, project finance, asset-based financing, ABS, and high-yield issuance.
Use token usage as a proxy indicator for compute demand.
OpenRouter data shows global weekly token usage rising from about 6 trillion to 28 trillion, a rise of roughly 350%, used to support the view that AI compute demand is rapidly expanding.
Assess the constraints that may emerge as AI infrastructure development moves from funding need to execution bottlenecks.
The report highlights that grid interconnection, power equipment, skilled labor, permitting delays, regulatory friction, insurance capital supply, and Middle East capital flows may all affect financing and development pathways.
Asset mapping & comparison
Structured mapping from thesis to named assets (strengths, weaknesses, peers, risks).
- AI infrastructure credit assetsDirectly benefit from expanding AI capex and data center buildout financing demand.
- Strengths
- Large demand scale, long cycle, high strategic importance, and continually evolving financing structures.
- Weaknesses
- Underlying project execution is complex and may be constrained by power, permitting, equipment, and labor bottlenecks.
- Comparison
- Compared with traditional corporate credit, they resemble a hybrid of corporate finance, project finance, and asset financing.
- Risks
- If capex forecasts are revised down, projects are delayed, or salvage-value assumptions prove wrong, credit risk could rise.
- GPU financingShifting gradually from equity-capital-dominated financing toward credit markets.
- Strengths
- Rapidly growing compute demand lays the foundation for asset-based financing and eventual ABS structures.
- Weaknesses
- Fast technology turnover creates uncertainty around residual values and reusability.
- Comparison
- Compared with data center construction financing, GPUs have more pronounced depreciation, residual-value, and technology risks.
- Risks
- Changes in model demand, chip supply cycles, collateral valuation volatility, and refinancing risk.
- Data center energy infrastructurePower availability is becoming a gating factor for AI buildout, pulling energy financing into the AI capex universe.
- Strengths
- Grid-constrained conditions increase the strategic value of off-grid power, fuel cells, gas turbines, energy storage, and brownfield mine-repurpose solutions.
- Weaknesses
- Heavily exposed to regulation, interconnection, equipment delivery, and local political factors.
- Comparison
- Compared with pure data center financing, energy assets depend more on permitting and physical delivery capabilities.
- Risks
- Persistent power shortfalls, project approval delays, energy-price volatility, and policy shifts.
- Insurance capital and fixed-income capitalLife insurers are a meaningful source of funding for AI and data center infrastructure.
- Strengths
- Higher interest-rate environments support stronger annuity inflows, giving insurance companies yield-seeking end-investor incentives.
- Weaknesses
- Capital supply is highly sensitive to interest-rate conditions and annuity sales.
- Comparison
- Compared with Middle East strategic capital, insurance capital is more exposed to the macro interest-rate cycle.
- Risks
- If rates decline materially, annuity sales may slow and insurance capital availability could weaken.
Key data
- 2026 capex outlook for the five largest hyperscale cloud providersabout $800 billionMorgan Stanley internet stock analyst team revised the forecast upward after Q1 results and call discussions.
- 2027 capex outlook for the five largest hyperscale cloud providersabout $1.16 trillionMaterially above last year’s forecast of roughly $4.5 trillion for each of 2026 and 2027.
- Change in global weekly token usageabout 6 trillion to 28 trillion, up about 350%OpenRouter data, from early January.
- Estimated data center power shortfall55 gigawattsStephen Byrd estimates data center developers face roughly a 55 GW electricity shortfall.
- China social financing outlookRMB 1.25 trillionForecast expected to be released between May 10-15 in the weekly report; previous reading was RMB 1.16 trillion in April 2025.
- US April core CPI forecast0.36% m/m and 2.7% y/yThe report expects tariffs to keep core goods inflation positive, with services inflation accelerating via housing and oil pass-through.
Impact & implications
For investors, AI infrastructure financing is likely to become one of the core themes in credit markets over the coming years, creating opportunities in loans, private credit, asset financing, ABS, high-yield debt, and project finance. At the same time, these opportunities will increasingly depend on underlying asset quality, lease protections, residual-value structures, power availability, and stable capital supply. As AI construction and energy-system financing converge, energy infrastructure assets and solutions that shorten power-connection timelines may increasingly be acquired or financed directly by AI participants.
Risks
- Grid interconnection, power equipment, skilled labor shortages, and permitting delays may limit the speed of AI infrastructure buildout.
- Political and regulatory frictions at local, national, and international levels could alter project implementation and financing pace.
- A sharp decline in rates could weaken fixed annuity sales, reducing life insurers' support for AI and data center infrastructure.
- Middle East capital flows are geopolitically sensitive and could slow under risk-off conditions or strengthen under strategic long-duration investment demand; direction remains uncertain.
- Innovation in GPU and AI infrastructure financing structures may introduce residual, collateral valuation, maturity mismatch, and complexity risks.
- The report includes conflict-of-interest and regulatory disclosures; Morgan Stanley and affiliates may have business relationships or positions in covered companies or related instruments.
What to watch
- Subsequent capex guidance and actual spending execution from hyperscale cloud providers.
- Persistence of global token usage, AI model query volumes, and compute demand.
- Whether GPU financing develops further into a mature ABS or other securitized structure.
- Progress in data center power solutions such as grid connection, off-grid power, gas turbines, fuel cells, energy storage, and repurposing of former Bitcoin mining sites.
- Allocation shifts by life insurers, private credit, public bond markets, and Middle East capital toward AI infrastructure financing.
- The weekly impact of U.S. CPI, PPI, retail sales, U.K. GDP, China social financing, and inflation on rates and risk appetite.