The note Goldman Sachs sent its credit clients this month made an odd pitch for a bull case. Buy investment-grade bonds, it said, and watch the calendar.
The argument rested less on the borrowers than on their absence. AI-related bond issuance has come to an almost complete stop over the past two months, and that silence has held investment-grade spreads flat through a stretch in which the markets around them moved hard, according to the note from Jeffrey Papai, a credit sales trader at the bank.
The divergence shows up in a single figure. From Aug. 15 to Oct. 15, the extra yield investors demand to hold investment-grade credit widened by one basis point. In the opening months of the last two tightening cycles it widened by 27 basis points and 15 basis points. Over the same six weeks this time, the MOVE index, which tracks expected Treasury volatility, rose above 110, the Federal Reserve resumed raising rates, and French sovereign risk widened.
The calm, in other words, is not a verdict on the market’s largest borrowers. It is a measurement taken while the heaviest of them were out of the room.
What sits behind the flat line is a market that has already repriced. The share of total yield accounted for by credit spreads fell below 13% last month for the first time on record, Mr. Papai wrote, against more than 80% during the worst weeks of 2008 and 2020. Investment-grade index yields are above 6%, a level breached only twice since the financial crisis. In France, roughly a quarter of corporate bonds now yield less than government debt, a market that prices LVMH’s credit above that of the French Republic.
The queue waiting to test that pricing is long. Roughly $150 billion of chip-related financing has entered preparation or issuance, tied to programs that include Broadcom’s APCHIP 2.0, SpaceX’s chip borrowing and a Broadcom-OpenAI effort known as Jalapeno, according to people familiar with the pipeline. Goldman Sachs Research estimates that AI-related debt issuance has reached nearly $500 billion in 2026, and analysts there expect gross issuance from hyperscalers to hit a record $420 billion next year, up about 60% from this year’s estimate.
The composition of that supply matters more than its size. By 2027, Goldman estimates, AI-related investment-grade bonds maturing in seven years or less will reach $275 billion, or 20% to 25% of all short-term investment-grade issuance, against roughly 10% today. Regional banks, utilities and loans backed by OpenAI’s graphics processors will be bidding for the same capital, from buyers who until now have had little reason to choose between them.
The rate backdrop has made the exercise more delicate. A central bank that has resumed raising rates lifts the return on risk-free paper while raising the cost of rolling existing debt, and a six-week jump in expected Treasury volatility would ordinarily argue for wider credit spreads rather than narrower ones. That spreads held steady instead says less about the borrowers than about a market in which the marginal seller of risk has stepped aside.
Those buyers are, increasingly, insurers. Some of their confidence in the paper rests in part on the market value of the companies that issued it, which makes a stock price a credit input as well as an equity one. The banks that arrange the deals have declined to keep the debt on their own books, analysts said, a division of risk that leaves the exposure with investors who cannot easily trade out of it.
Demand has been strong enough to absorb what little has been offered. Investment-grade funds took in $12 billion in the six weeks after the Jackson Hole meeting, a 65th percentile reading since 2020, with $5 billion arriving last week against a weekly average this year of about $3 billion. Money looking for a home has met a market with nothing large to sell it. That mismatch, not sentiment, is what the flat spread measures.
Corporate bond funds have also had few alternatives. With government yields moving sharply and the supply of new paper thin, managers holding fresh inflows have little choice but to buy what is offered, a dynamic that flatters prices in a market where the biggest issuers are simply absent.
There is nothing fraudulent in any of this. A market where supply pauses borrows stability from the calendar, and the calendar hands it back. The question is what the January print looks like when the $150 billion pipeline converts into deals and investors who have been paid a record-thin premium discover they are being asked to extend duration into a heavier supply calendar.
Goldman’s own framing is conditional in a way that undercuts the buy case it accompanies. The rally holds, Mr. Papai wrote, until just before the AI supply crunch of 2027.
For credit investors, the practical implication is that the buffer is thin at the moment it is most needed. A spread that has already compressed to a record low share of yield has little room to absorb a repricing gently, and the assets most likely to trigger one sit in the sector that has temporarily stopped selling.
The next two months may look like the best conditions of the year. They are better understood as the last quiet ones, borrowed from borrowers who have not yet come to market and have every reason to arrive together.
