Investment-grade credit is skating on thin ice. Even after a burst of cross-asset turbulence, spreads in US high-grade bonds tightened to fresh three-decade lows this week, leaving investors with the slimmest cushion against default and downgrade risk since the 1990s. That complacency meets a new test: a record, AI-fueled funding cycle led by the biggest names in tech, with borrowers flooding primary markets to bankroll data center buildouts at a pace Wall Street has rarely seen.
The reset in US investment-grade option-adjusted spreads to new cycle tights underscores how quickly fears faded after a bout of global volatility. The market is now priced for benign defaults, resilient cash flows, and a Goldilocks macro path that keeps Treasury yields contained even as supply swells. That works until it doesn’t. At three-decade lows, the spread buffer is so thin that even routine earnings misses, guidance cuts, or rating downgrades can force a rethink. In 1997-98, spreads also looked bulletproof—until they weren’t. Today’s setup is different, but the math is the same: when the premium for bearing credit risk shrinks, the margin for error evaporates.
The scale is eye-catching. Large-cap technology issuers sold a record $108.7 billion of bonds in the final quarter of 2025, then another $15.5 billion in the first two weeks of 2026, as hyperscalers raced to secure long-dated funding before the next rate pivot. The rationale is simple: AI infrastructure spending is massive, ongoing, and front-loaded. Corporate guidance and third-party estimates point to more than $1.5 trillion of capital investments over five years, with as much as $300 billion to $400 billion annually likely to be financed in the debt markets depending on cash generation, buyback commitments, and tax dynamics. For investors, that means a structurally heavy primary calendar, frequent mega-deals, and tighter new-issue concessions as underwriters test how far appetite stretches with spreads already hugging historic lows.
AI-linked issuers now represent about 14% of the US investment-grade index, larger than the banking sector’s roughly 11.5% share. That marks a regime shift in a $9 trillion market historically anchored by financials, industrial stalwarts, and utilities. The passive money that tracks benchmarks will be forced to absorb a rising share of tech credit, amplifying concentration risk across the curve. Portfolio construction gets trickier: if one or two platform companies hit a speed bump—say, a slower ramp in AI monetization or regulatory friction—index-linked buyers will be pulled along for the ride. In the 1990s, telecom and media volumes swelled before the sector underperformed; today, the concentration sits atop highly profitable franchises, but it is concentration nonetheless.
The market’s early warning lights are flickering in places like Oracle, where credit default swaps widened to the highest in about three years amid concern that aggressive investment and acquisition legacies leave less room for error. That is not a verdict on any single balance sheet so much as a marker of growing sensitivity to leverage in the cohort should cash flows lag capex. The paradox of ultra-tight spreads is that they attract more issuance, which can weaken credit metrics at the margin and, in turn, make spreads more volatile. If a handful of marquee names guide to heavier borrowing needs or push maturities out on aggressive terms, secondary spreads could finally stop grinding tighter and begin to gap wider on disappointment. The tape will punish pro forma debt turns that drift up without a visible, near-term revenue line to match.
There is a credible bull case: balance sheets at the largest platforms are stronger, free cash flow is deeper, and product cycles are more predictable than during the dot-com era. Many megacaps can carry gross debt without stressing net leverage thanks to cash and marketable securities. And if AI workloads translate into durable software and cloud revenue, today’s capex spike could look like prudent investment when viewed over a decade. Still, the late-1990s lesson is that markets can mis-time payoffs. In that cycle, high-grade spreads sat at or near tights even as the telecom buildout overreached; when growth faltered and financing shut, repricing was abrupt. The 2020s risk is not a carbon copy, but it rhymes: extrapolated demand assumptions, a financing boom that outpaces realized economics, and a credit market that leaves little room for disappointment.
The near-term risk is supply meeting a macro wobble. With spreads near historic tights, the all-in yield investors earn is now driven as much by Treasury duration as by credit premium. If growth cools or policy stays higher for longer, that duration can sting just as issuance accelerates. Issuers are already leaning into 10-, 20-, and 30-year tranches to lock funding; that makes the investor base more sensitive to rate volatility and curve steepening. The soft-landing consensus helps—so does a still-sturdy consumer—but any shift in the rate path could force underwriters to widen concessions. A thin spread buffer turns a rate market hiccup into a credit repricing, especially for names crowding the long end of the curve where life insurers and pensions are price-setters.
Benchmark concentration also raises liquidity questions. In calm markets, mega-deals clear and ETFs hum. In stress, passive outflows can force selling into pockets where dealers are balance-sheet constrained and single-name liquidity evaporates. With LQD and similar vehicles increasingly dominated by tech and communications paper, a shock to AI sentiment won’t stay contained to a few tickers. That feedback loop is not hypothetical; it is how modern credit sells off. If a marquee earnings print disappoints, a guidance cut hits capex timing, or a regulatory headline dents perceived AI monetization, high-grade spreads can gap wider not because default risk jumps, but because the market needs a higher premium to warehouse concentrated exposure.
Watch three things: capex guidance, capital structure discipline, and new-issue pricing. On guidance, investors need credible bridges from today’s data-center capex to tomorrow’s software and services revenue. On structure, boards will be forced to prioritize between buybacks and AI spend; if buybacks stay elevated alongside rising gross leverage, rating agencies will react, and spreads will reprice. On pricing, primary concessions at the front of the calendar will signal whether buyers are pushing back now that spreads have little room to compress. Expect marquee issuers like Microsoft, Alphabet, Amazon, Meta, and Oracle to tap across the curve; how those books build and where final pricing lands relative to secondaries will set the tone for the quarter.
The investment-grade market wanted a clean macro narrative and got one; spreads did the rest. Now the AI buildout is arriving at full speed, with record issuance, rising index weight, and investors forced to choose between embracing the growth story or demanding more compensation for risk. The soft landing may hold. AI workloads may scale. But when credit spreads mark three-decade lows while a new sector re-levers in plain sight, the burden of proof shifts to issuers. If the next wave of deals clears on tighter terms with modest concessions, euphoria survives. If not, we will find out how much true demand exists for AI debt when credit is no longer priced for perfection.
