Moody's Q3 and the Borrower That Actually Pays the Fee

Moody's reported on October 21 and missed. Revenue rose 1% to $2.02B against a consensus near $2.08B, and the ratings business, Moody's Investors Service, shrank 1%. In the same quarter investment-grade issuance set a third-quarter record, supported, in the company's words, by AI infrastructure and hyperscaler financing. The usual story puts the rating agencies on the winning side of the AI build, next to the chip and power suppliers: whoever lends to the hyperscalers, Moody's and S&P Global rate it.

The chart above is the problem with that story. This year Moody's ratings revenue has moved with high-yield bonds and leveraged loans, and hardly at all with investment grade. Q1 had the fastest investment-grade growth of the year and the slowest ratings growth of the first half, because leveraged loans fell. Q2 was the big quarter because leveraged loans rose by half. Q3 had record investment-grade borrowing and a decline, because high yield and leveraged loans both fell by a fifth or more. The AI borrower supplies volume. The leveraged borrower supplies the growth.

Read line by line, Q3 is not a story about investment-grade fees getting cheaper. Investment-grade revenue grew a little faster than investment-grade volume. Corporate Finance as a whole lost a seventh of its transaction revenue because the leveraged borrowers stepped back from a record Q3 2025, and that one line, worth $59M less than a year earlier, more than accounts for the ratings decline. The AI paper did its part; it just could not replace the borrowers who were missing.

Management added the second piece on the call. Year to date, chief executive Rob Fauber said, AI-related issuance has been a little over half of Moody's issuance growth but closer to 40% of its revenue growth, and the big hyperscaler deals now sit on a frequent-issuer commercial arrangement. That is a modest discount, not a collapse. But it confirms the shape of the chart: the largest new source of rated debt in years is also the one that earns Moody's the least for each dollar it adds.

Moody's does not publish fees by category, so what follows is my reading of the numbers above rather than a disclosed fee schedule.

• Repeat issuers negotiate. A hyperscaler that comes back to the bond market in the tens of billions several times a year can agree a programme price. Moody's said that is how the hyperscaler deals are now structured, and it is consistent with AI being more of the volume growth than the revenue growth.
• Big deals spread the fee thin. The analytical work on a $25B investment-grade deal is not twenty-five times the work on a $1B high-yield bond, and fees appear to follow the work more than the size. That would explain high-yield revenue falling twice as fast as high-yield volume.
• The speculative borrower needs the public rating most. For a high-yield issuer selling to public investors, the rating sets much of the coupon and opens the buyer base. It has the least room to negotiate the fee in public markets, and its main alternative is to leave them for a private loan.
• Monitoring follows every rating. MIS recurring revenue grew 8% to $373M in Q3 2026 while transaction revenue fell. Each new AI bond adds a fee for monitoring over its life, so the AI build does pay over time, just slowly.

If the agencies' growth depends on the leveraged borrower, the names that matter next to them are not the hyperscalers, whose borrowing is covered in CoreWeave, AI Debt and the Rate the Moat Sets. They are the private-credit managers who lend to that borrower directly. KKR owns the insurer Global Atlantic and Blackstone manages insurance money for others, so both sit on both sides of the private-rating question: they make the loan, and an insurer they own or advise holds it and needs a rating it can use for capital.

The grid shows where the agencies' advantage lives: in capital rules that reference ratings, in a century of default data, and in investors who will not buy an unrated public bond. Nothing in Q3 touched any of that. What the grid cannot show is that the rules part of the moat is only as strong as the regulator's willingness to treat a Moody's or S&P rating differently from a cheaper one. For an insurer's private loan, that question is being decided this quarter.

The strongest objection is in that table. In every quarter of 2026, Moody's investment-grade revenue grew at least as fast as investment-grade volume, and in Q2, a heavy hyperscaler quarter, much faster. If AI debt were seriously discounted, the line it dominates should show it, and it does not. The only direct evidence for a discount is one spoken ratio on the call, and 40% of revenue growth from one theme is a lot of money. Moody's own analysts expect AI capex to rise from about $800B in 2026 to more than $1T in 2027. If hyperscaler issuance keeps compounding, it could carry the ratings line through a weak leveraged-finance year by sheer size.

The second objection is timing. Moody's said the Q3 decline came against a record prior-year quarter, and leveraged loans lost less revenue than volume, which fits the repricing wave that ran through Q3 2025 rather than a lasting loss. Spreads were stable for most of the quarter and widened only in late September. Leveraged borrowers wait; they do not disappear. And private credit may be a holding pen rather than an exit: PitchBook counted about $12B of direct-lending deals refinanced into syndicated loans in Q3, and the number of private credit deals Moody's rated rose more than 40%.

The third objection is that AI paper gets more lucrative as it gets riskier. Not all of the build is funded by Alphabet and Meta: neoclouds and data-centre project vehicles borrow too, and their paper is high yield, project or structured debt. Moody's said issuance in sectors tied to AI and power investment rose about fourfold in Q3, and Public, Project and Infrastructure Finance transaction revenue rose 14%. If that continues, the AI build becomes a leveraged-finance customer itself.

That leaves private ratings, which is where the bears are partly right and the evidence is contested. A Columbia study posted in June put US life insurers' privately rated assets at about $481B, roughly 12% of their rated credit, up from under 2% in 2018, with most of those ratings from Egan-Jones, KBRA and Morningstar DBRS, and found private grades running two to three notches above public grades for the same loss risk. KBRA disputed the paper a week later: it measured accounting impairments rather than defaults, inferred inflation rather than measuring it, and by its own model implied about $4B of extra capital rather than the headline $4.5B. Moody's has its own numbers on the size of the pool: life insurers' private and illiquid bonds reached about $807B at the end of 2025, about a fifth of their fixed income. If insurers can keep using cheaper private ratings for capital, the leveraged borrower that pays the duopoly's richest fee has a route around it. If the NAIC tightens, that route narrows.

Nothing in Q3 touches the reasons the framework rates the agencies' moats as it does, and the matrix reads the same as before the print. The refresh of the Moody's analysis that ships with this article moves only the ratings driver, from accelerating to decelerating, and leaves the moat and the price ladder alone. MCO is an IM25 holding in the book effective October 26, as is SPGI.

The adjustment is in what to read. A record month of hyperscaler bonds is not evidence that the agencies are winning the AI build, and a quiet quarter for the hyperscalers is not a threat to them. The numbers that move Moody's and S&P Global are high-yield and leveraged-loan issuance, the share of leveraged lending that stays private, and whether insurers can keep using private ratings on current terms after the NAIC decides. Moody's Analytics, which grew faster than the ratings business this quarter, depends on none of it; S&P Global and the Half With the Moat argued the opposite split would hold, and one quarter of inversion is not the full year its own test asks for.

The lesson reaches past rating agencies. A mandate tells you a toll will be collected; it does not tell you which customer pays the most of it. When a protected franchise's biggest new customer is also its strongest negotiator, the volume headline will flatter the earnings, and the customer who matters is the one with the fewest alternatives.