AI Credit Boom Risks Telecom-Era Bust, Sycamore Warns

Sycamore Tree Capital Partners’ CIO Trey Parker flags both opportunity and risk in credit tied to the AI boom, drawing parallels to the late-1990s telecom collapse. He favors shorter maturities and stronger data-center tenant credit ratings, implying a cautious underwriting stance as AI-driven leverage and occupancy risks remain in focus. As this is commentary rather than a discrete company/regulatory catalyst, the likely impact is limited to positioning in credit rather than broad market repricing.

Analysis

The market is likely underpricing the funding-risk asymmetry inside the AI buildout: equity investors can tolerate years of capex, but credit investors get paid to underwrite tenants’ lease renewal and utilization assumptions. The cleanest money is in short-duration, senior paper tied to investment-grade tenants; the bad trade is reaching for yield in longer-dated, lower-rated financing where the repayment story depends on uninterrupted AI capacity expansion and rapid takeout financing.

The late-1990s telecom parallel matters because the first pain point is rarely the incumbent platform names; it shows up in the capital stack where excess supply, falling unit economics, and refinancing dependence collide. Over 1-3 months, the most vulnerable credits are data-center financings with single-tenant concentration, construction draws, or aggressive leverage; over 6-18 months, the risk is that capacity overshoots demand and spreads widen even if headline AI spending stays strong.

Consensus is probably too confident that “AI infrastructure” is a single credit bucket. In reality, hyperscaler-adjacent senior secured debt can remain resilient while mezzanine, bridge, and speculative-grade issuers get repriced fast if rates stay higher for longer or if one or two large tenants slow incremental leasing. Watch for any sign that capex guidance is being funded with shorter tenor debt, because that usually precedes a refinancing wall rather than a benign growth cycle.

The thesis is falsified if utilization remains tight, tenant renewal spreads stay positive, and 2026 refinancing windows are repeatedly extended without concessions. A sustained tightening in credit spreads despite slower equity multiples would also argue the market is not yet signaling telecom-style overbuild risk.

Key Decisions for Investors

• Prefer short-duration investment-grade credit over lower-quality AI infrastructure credit: long LQD / underweight HYG for a 1-3 month quality-spread widening trade if financing conditions tighten.
• Avoid or short the weakest data-center/AI-linked leveraged loans and private credit exposures where repayment depends on rapid lease-up; best risk/reward is in the 12-24 month refinancing bucket, not the near-term construction phase.
• If you want a cleaner expression, buy CDX IG protection and fund it with no offset in CDX HY only if primary issuance in AI-adjacent names reaccelerates; this is a hedge against a telecom-style credit repricing.
• Watch item, not full trade: if any large AI tenant cuts capex or if data-center lease spreads compress, rotate out of BB/B paper quickly; that would be the first signal the credit cycle is peaking.