AI's Data Center Debt Binge: What Credit Markets Should Price In Now
AI has lifted equities, but the build-out behind the story is getting financed with debt at a pace worth your attention. The center of gravity has shifted from cash-rich tech balance sheets to bond markets, private credit, and securitization.
UBS pegs AI data center and project financing at about $125 billion this year, up from roughly $15 billion during the same stretch last year. T. Rowe Price's Anton Dombrovskiy points to the acceleration in both public and private credit, and the need to watch it closely. The Bank of England has flagged that heavy debt reliance in this cycle could become a financial stability issue if valuations cool or cash flows under-deliver. Their Financial Stability Report lays out the macro risks.
Who's Issuing-and How It's Reordering the IG Stack
Christopher Kramer at Neuberger Berman notes a structural change: mega-cap tech is tapping debt to fund AI infrastructure instead of relying mainly on internal cash. Recent months show a wave of large tech issuance, including names like Oracle and Meta.
JP Morgan estimates AI-linked issuers now make up 14% of its investment grade index, surpassing U.S. banks as the top sector. That concentration raises crowding risk and index-tracking pressures. Oracle's 13% stock drop after a weak read-across reminded the market how quickly sentiment can swing.
Not Everyone's Buying the Paper
Al Cattermole at Mirabaud Asset Management hasn't touched the new AI-linked IG or HY deals. His team questions whether data centers will be delivered on time, within budget, and at the promised compute capacity.
His stance is clear: the risk looks closer to equity than debt in parts of the build cycle. If investors are taking construction and technology risks, coupons should reflect that.
Beyond Bonds: Private Credit and ABS Step In
Private credit is underwriting a growing slice of data center projects, often with customization that banks won't touch. At the same time, securitized products are bundling loans and even data center rents into asset-backed securities (ABS).
Bank of America notes digital infrastructure is still a small corner of U.S. ABS, but it's scaling. Post-2008, ABS draws extra scrutiny because complexity and liquidity can turn quickly. If you need a refresher on ABS mechanics, the SEC's overview is useful: Asset-Backed Securities.
Key Risks You Need to Underwrite
- Completion and cost risk: EPC terms, liquidated damages, contingency buffers, and interconnect timelines for power.
- Power and permitting: grid access, long-term supply contracts, cost pass-throughs, and regulatory constraints.
- Tech obsolescence: refresh cycles, architecture shifts, stranded-capex risk if workloads move or designs pivot.
- Tenant concentration and offtake: take-or-pay strength, termination options, escalators, and counterparty health.
- Valuation sensitivity: higher cap rates, slower lease-up, or weaker pricing can hit coverage and refinancing options.
- Duration and rate risk: floating-rate exposure, hedge effectiveness, and interest coverage under stress.
- Covenant quality: maintenance tests vs. incurrence-only, asset-level security, step-in rights, and reserves.
- ABS complexity and liquidity: structural subordination, triggers, collateral performance data, and dealer support.
Questions to Ask Before You Add Exposure
- What's the pre-lease percentage, weighted average lease term (WALT), and DSCR under base and downside cases?
- How are power costs hedged and passed through? Any exposure to curtailment or price spikes?
- What's the sponsor's equity check, support package, and incentive to finish? Any keepwell or completion guarantees?
- What credit enhancements exist: capex reserves, interest reserves, make-whole, or springing collateral?
- Can the structure withstand: 6-12 month delays, 15% capex overrun, 200-300 bps higher spreads, and slower lease-up?
- Where does the refinancing wall sit (2026-2028), and what's the plan if risk-free stays higher for longer?
Trading and Portfolio Implications
Spread compression in AI-tagged IG can mask hidden construction or tenant risks. Compare to non-AI comps with similar leverage and duration-don't overpay for a hot label.
Watch index mechanics as AI-linked weight climbs: crowding can amplify moves both ways. Equity drawdowns in key issuers can spill into credit; Oracle's swing is a case study.
What to Watch Next
- Issuance calendar from mega-cap tech and large colocation platforms; any pivot to secured vs. unsecured.
- Private credit club deal sizes, pricing, and covenant drift.
- ABS share of funding, collateral transparency, and trigger design in new shelves.
- Regulator commentary from central banks on debt concentration and systemic links.
- Power constraints and permitting delays that can ripple into covenants and DSCR.
Bottom Line
The AI build is real, but the funding stack has shifted risk onto credit investors. Price construction and technology uncertainty like you mean it, structure for delays, and demand sponsor alignment.
If you need to upskill teams evaluating AI-related issuers and vendors, these resources can help: AI tools for finance.
Your membership also unlocks:
