Pimco: Data Center Construction Slowdown to Compress AI Bond Spreads
nashnova research
Pimco argues that construction delays will curb new AI-infrastructure bond supply, compressing spreads from today's 146 bp premium — meaning the feared debt flood may be stretched over time rather than crashing into credit markets all at once.
Spreads are already narrowing — what's driving it?
Investment-grade bonds backed by hyperscaler tenants (Amazon, Microsoft and peers) trade at roughly 146 basis points above comparable tenant-backed debt.
This means → the market still charges an "AI-label" risk premium, but it is shrinking — in late September the gap was closer to 180 bp.
The catalyst is not a demand surge but supply contraction: permitting bottlenecks and rising costs are delaying construction, which slows new-bond issuance.
Is the spending boom over?
Consensus expects Amazon, Alphabet, Meta, Microsoft and Oracle to spend a combined $1.2 trillion-plus per year over the next two years. Pimco sees some of that being cut or deferred.
In plain terms = AI infrastructure is not being cancelled — it is shifting from a sprint to a marathon. Financing needs spread across more years, so fewer bonds hit the market in any single period.
Pimco strategist Lotfi Karoui's framing: the build-out will be "longer-duration and more non-linear," not dramatically smaller.
Slower building actually helps existing data centers?
Construction delays → new compute capacity comes online later → utilization and rental rates at existing facilities stay supported.
This means → operating data centers face less dilution from fresh supply, and their credit quality firms up rather than erodes.
Supply tightening delivers a double tailwind: less debt on the financing side, sturdier rents on the operating side.
How does this differ from the market's prior worry?
The prevailing fear was straightforward: relentless AI borrowing → overwhelms investor demand → pushes up funding costs across the board.
Pimco's counter-narrative: construction delays flatten the issuance curve. The flood becomes a steady stream, and credit markets are not swamped in one wave.
After Democrats won the U.S. midterm elections, doubts about large-scale AI investment returns resurfaced — but Pimco expects that once volatility settles, issuer-level fundamentals will reassert control over pricing.
Which bonds look attractive, and which carry risk?
Pimco favors hyperscaler-backed data-center bonds: scarcity premium + still-elevated spreads → potential to outperform senior unsecured tech-company debt.
High-leverage infrastructure providers (smaller players that borrow to build) warrant more caution.
In plain terms = if construction stalls and revenue is delayed, these firms still owe interest, lease payments and chip-purchase commitments on schedule — cash-flow pressure does not ease with a slowdown; it can intensify.
市场有风险,内容仅供研究参考,不构成投资建议。
