Data Center Debt Is Making Investors Choosier: Cisar

A surge in AI infrastructure financing is bringing billions of dollars of data-center debt to the high-yield market, according to CreditSights' global head of strategy Winnie Cisar. The increased supply is making investors more selective in capital allocation, potentially raising financing pressure for lower-quality issuers and competing high-yield borrowers.

Analysis

The relevant transmission is not simply wider high-yield spreads; it is a rationing of capital toward assets with contracted hyperscaler cash flows and away from speculative capacity builds. Levered data-center developers and AI-cloud operators face a reflexive risk: a 100bp increase in refinancing cost can erase much of the modeled equity IRR on long-dated facilities, forcing slower build schedules and reducing near-term orders for equipment. Public REITs Digital Realty (DLR) and Equinix (EQIX) should be relatively insulated where pre-leasing and investment-grade balance sheets preserve access, while smaller private operators become more dependent on expensive private credit.

The second-order beneficiary is the equipment layer. Vertiv (VRT), Eaton (ETN), and Schneider Electric (SU.PA) sell into projects that have reached committed financing and can gain share if weaker developers cancel or defer expansion; however, their order backlogs are only as durable as customer funding. Oracle (ORCL) and CoreWeave (CRWV) merit closer credit scrutiny than the hyperscalers because capacity growth requires continuous external capital, making their equity multiples sensitive to funding spreads rather than only GPU demand. Over the next 1-3 months, new-issue concessions, covenant quality, and project-level pre-leasing—not headline AI demand—will determine whether this becomes a broader capex slowdown.

Consensus may be too indiscriminate in treating all AI infrastructure financing as equivalent. A selective credit market can consolidate capacity into investment-grade platforms, ultimately improving pricing and utilization for DLR/EQIX while impairing marginal competitors. The bearish thesis is falsified if high-yield AI-linked deals clear with shrinking concessions and strong order books; it strengthens if spreads widen while reported backlog conversion or construction starts decelerate over the next two quarters.

Key Decisions for Investors

• Prefer a 6-12 month long DLR / short CRWV relative-value position: DLR has lower funding-risk sensitivity and can benefit from constrained competitor supply, while CRWV remains more exposed to capital-market access. Reassess if CRWV demonstrates sustained contracted backlog growth funded without materially higher interest expense or equity issuance.
• Accumulate VRT and ETN only on financing-driven pullbacks, with a 6-18 month horizon; use 10-15% downside limits from entry because a broad project-deferral cycle would compress backlog multiples before revenue slows. Monitor quarterly book-to-bill, cancellation commentary, and receivables growth as the thesis checkpoint.
• Avoid broad HYG shorts solely on this signal: AI infrastructure issuance is not yet sufficient evidence of systemic high-yield deterioration. Instead, establish an alert for a persistent 75-100bp widening in CCC versus BB spreads alongside weak AI-related deal execution; that combination would support a 1-3 month defensive short in HYG or long CDX HY protection.
• For credit books, favor senior secured or investment-grade exposure backed by contracted hyperscaler leases over unsecured debt from AI-cloud and smaller data-center sponsors. Require disclosure of customer concentration, lease tenor, debt maturity schedule, and construction funding before underwriting new exposure.