AI infrastructure financing

TOPIC HUB · AI INFRASTRUCTURE FINANCING

AI infrastructure financing

The reference on capital structures funding the AI buildout: DDTL contract-collateralised debt, private-market equity, hyperscaler-contract ABS securitisation, nuclear PPA stacks, sale-leaseback, green bonds, insurance, sovereign wealth, public REITs, and the hyperscaler buy-vs-build allocation frame.

01Direct answer

AI infrastructure financing is the specific set of capital structures being used to fund the $500B+ AI compute + data centre buildout 2024-2030. The largest structural innovations are Delayed Draw Term Loan (DDTL) debt collateralised by hyperscaler contract commitments (estimated ~$95B committed by 2026 per aggregated DDTL disclosures across CoreWeave, Nscale, Crusoe and adjacent operators), and permanent-capital private-equity vehicles (Blackstone BREIT + Brookfield Infrastructure + DigitalBridge deploying ~$200B+ 2020-2026). Adjacent structures include hyperscaler-contract ABS securitisation (nascent 2025-2026, $115B potential 2030), nuclear PPA financing stacks (utility + hyperscaler prepay + DOE + IRA credits), sale-leaseback ($25B expected 2026 volume, 5.5-7.5% cap rates), green bonds ($40B+ annual 2024-2026 issuance), sovereign wealth participation ($80-120B deployed 2020-2026), public REITs (~$150B market cap), and buy-vs-build capital allocation across the hyperscalers themselves.

02Key concepts

DDTL
Delayed Draw Term Loan; primary contract-collateralised debt structure for AI infra buildout.
Permanent capital
PE vehicles without fund-life constraints; enable follow-on capacity for multi-year buildouts.
Contracted revenue
Multi-year hyperscaler off-take commitments; the underlying collateral for most AI infra debt.
ABS securitisation
Asset-backed securities against hyperscaler contract commitments; nascent structure.
Sale-leaseback
Owner sells DC real estate, leases back; capital recycling structure.
BTM / behind-the-meter
Power arrangement where load co-locates with generation to avoid grid interconnection.
Nuclear PPA
Power purchase agreement with nuclear (restart, SMR, or traditional) generation.
DOE LPO
Department of Energy Loan Programs Office; federal loan guarantee for nuclear + clean energy.

04Deep-dive research

06Commercial + investment implications

For PE sponsors + LPs

The financing map is the primary lens for evaluating AI infra allocations. DDTL is the near-term deployment vehicle; ABS is the emerging secondary market. Sovereign wealth is the incremental capital source at scale. See The PE playbook layer by layer for the sponsor-side deployment strategy.

07Frequently asked

What is DDTL and why is it the primary AI infra debt structure?

Delayed Draw Term Loan. Committed facility that draws down against construction milestones, secured against hyperscaler contract commitments. Structure lets the debt fund the build without full drawdown at close. CoreWeave built on it; Nscale scaling with it. Blackstone Credit, Ares, Sixth Street, KKR lead the market.

Why do permanent-capital vehicles win the buildout capital allocation?

Follow-on capacity. Traditional 10-year PE funds struggle with the multi-year AI infra buildout because a single portfolio company may need capital multiple times across the fund life. Permanent capital (BREIT, Brookfield Infra, DigitalBridge multi-vehicle) can support follow-ons without fund-cycle constraints.

How large is the ABS securitisation opportunity?

Nascent 2025-2026 ($5B first transactions). Rating agencies (KBRA, S&P, Fitch, Moody's) have adapted CMBS + solar ABS + fleet lease ABS methodologies. Base case $115B annual issuance by 2030 if adoption tracks solar ABS growth curve.

What does the nuclear PPA financing stack look like?

For restart (Constellation TMI-style): utility balance sheet + hyperscaler prepayment + DOE LPO guarantee + IRA tax credits. For SMR (Kairos, X-energy): more equity, less debt-eligible until first-of-a-kind risk retires. See Financing IV.

Where does sovereign wealth fit?

Materially at scale. PIF (Humain), Mubadala (G42/MGX), GIC, Temasek, ADIA collectively deployed $80-120B into AI infra 2020-2026. Direct co-invest + LP + national-champion routes. CFIUS + FDI review constrain but do not block. See Financing VIII.

Is buy-vs-build converging across hyperscalers?

No. Meta and Microsoft own-heavy; Amazon and Oracle lease-heavy; Google balanced with strong JV. Different capital structures + strategic philosophies drive different equilibria. See Financing X for the specific allocation-by-hyperscaler breakdown.

08Further reading

Definition AI infrastructure financing spans ten distinct capital structures deployed to fund the 2025-2030 buildout, from delayed-draw term loans (DDTL) collateralised against hyperscaler contract commitments through to sovereign wealth deployments, hyperscaler-contract ABS, nuclear PPA structures, sale-leaseback, green bonds and REITs. DDTL structures at CoreWeave, Nscale and Crusoe reached approximately $95B in aggregate committed capital by mid-2026.