8.35 Basis Points on Every Bond, a 68 Percent Margin: Why the Rating-Agency Duopoly Is a Business Built on Law — and Where It Wobbles for the First Time in 2026

Any company selling a $1 billion bond in 2026 wires $835,000 to S&P Global before the deal prices — and, almost always, a similar sum to Moody’s. Not because the treasurer values their opinion, but because without it no insurer, pension fund or bond mutual fund is allowed to buy the paper. That is the entire business model, and in the second quarter of 2026 it turned 68 cents of every revenue dollar into operating profit. This article works through what a fee nobody negotiates actually earns, why 92.9 percent of all outstanding ratings in the United States come from three firms, where the duopoly has already been broken without anyone noticing — and why both stocks lost a quarter of their value in four weeks last February, just as the ratings business was setting up the best quarter in its history.

Rating agencies are an anomaly in capitalism. They sell an opinion the rated party pays for, the buyer cannot commission, and the state has written into law. In 2008 that model helped cause the worst financial crisis in eighty years; the congressional inquiry later called the three firms “essential cogs in the wheel of financial destruction.” Eighteen years, two reform laws and two billion-dollar settlements later, the Big Three’s share of U.S. ratings has slipped from 98.8 to 92.9 percent. Moody’s corporate operating margin has risen from 38 percent in 2009 to 55 — and the ratings segment’s to 68.

The fee nobody negotiates: 8.35 basis points

The price is in a regulatory filing almost nobody reads. S&P Global Ratings files its list prices with the SEC every year, and for most U.S. corporate transactions the rate since January 2026 is 8.35 basis points of principal, subject to a $155,000 minimum. In January 2025 it was 8.10 basis points; in 2024, 7.90; in 2023, 7.70. That is an 8.4 percent price increase over three years for a product that has not changed — and in none of those years did a customer of any size walk away.

On a $1 billion issue the list price is $835,000. Because virtually every large deal needs two ratings — most institutional mandates require at least two, and the major bond indices use the median of three — the issuer pays roughly $1.6 million for S&P and Moody’s together. That sounds like a lot until it is set against the interest saved by carrying an investment-grade rating rather than none: at 20 basis points of spread over a ten-year term, the saving is $20 million. The issuer pays $1.6 million to save $20 million. That is exactly why nobody negotiates.

The effective rate does run below the list price, because large issuers have relationship-based pricing and fees are capped. It can be backed out of the quarterly numbers: S&P billed $1,268 billion of issuance in the second quarter of 2026 and booked $746 million of transaction revenue on it — 5.9 basis points on average across all segments. Moody’s, which rated more than $2 trillion of debt for the second consecutive quarter, booked $891 million of transaction revenue, roughly 4.4 basis points. The gap between the two says less about pricing than about counting: Moody’s reports the volume it rated, S&P the volume it billed.

Issue size	S&P list price (8.35 bp)	Two ratings (S&P + Moody’s, assuming equal pricing)	Interest saved at 20 bp over 10 years
$100 million	$155,000 (minimum)	$310,000	$2 million
$500 million	$417,500	$835,000	$10 million
$1 billion	$835,000	$1.67 million	$20 million
$5 billion	$4.18 million (before cap)	$8.35 million	$100 million
$53 billion (largest AI bond of 2026)	$44 million (before cap)	$88 million	$1.06 billion

On top of the issuance fee comes the annual surveillance fee, paid for as long as the bond is outstanding — “non-transaction revenue” at S&P, “recurring revenue” at Moody’s. It is the reason the business makes money even in years when nobody issues anything: Moody’s booked about $1.38 billion of recurring ratings revenue in 2025, a third of the segment, and that figure barely moves with the issuance cycle.

Why there are exactly two: a license dated 1975

The duopoly is not an accident of market forces; it is a regulatory decision. In 1936 the Office of the Comptroller of the Currency barred national banks from holding “speculative” bonds and defined speculative by reference to the ratings of the leading houses of the day. In 1975 the SEC created the status of Nationally Recognized Statistical Rating Organization, NRSRO, and tied its broker-dealer capital rules to it. Over the following decades, insurance regulators, pension law, money-market fund rules and central-bank collateral schedules all pointed at the same ratings. A bond without a recognized rating was simply not purchasable by most institutional capital.

The second step was changing who pays. Until 1970, Moody’s and S&P sold their opinions to investors through subscriptions. The photocopier killed that model — one subscriber could reproduce the manual for all of Wall Street — and Moody’s began charging issuers instead. Ever since, the rated party pays for the grade, and the bond buyer the grade is meant for pays nothing. The conflict of interest has been understood for 56 years, described afresh in every crisis, and abolished in none.

The third layer is the two-ratings convention. Because investment mandates and index rules demand at least two grades, a new competitor does not mean an issuer replaces S&P with Fitch; it means the issuer buys a third rating. Fitch, wholly owned by Hearst since 2018, lives on precisely that: it is usually the third opinion, not the first. A newcomer would not have to persuade issuers but the thousands of investment-policy statements in which S&P and Moody’s are named. That is the moat, and it is legal, not technological.

The quarter’s arithmetic: 68 cents on the dollar

What that structure is worth shows in the second quarter of 2026, the best in the history of both ratings businesses. Moody’s Investors Service booked $1,260 million of revenue — $891 million transactional (up 34 percent) and $369 million recurring (up 6 percent) — at an adjusted operating margin of 68.3 percent. S&P Global Ratings booked $1,339 million: $746 million from transactions (up 25 percent) and $593 million from surveillance and relationship pricing (up 8 percent). Segment expenses fell 2 percent, to $426 million, so operating profit rose 28 percent to $913 million. Margin: 68.2 percent under GAAP, 68.5 percent adjusted.

Q2 2026	Moody’s Investors Service	S&P Global Ratings
Revenue ($ million)	1,260	1,339
of which transaction	891 (+34%)	746 (+25%)
of which recurring / non-transaction	369 (+6%)	593 (+8%)
Adjusted operating margin	68.3% (+410 bp)	68.5% (+310 bp)
Rated / billed volume	> $2,000 billion	$1,268 billion (+25%)
Effective rate (transaction revenue / volume)	~4.4 bp	~5.9 bp
Group revenue	2,185 (+15%)	3,678 pro forma (+11%)
Adjusted group margin	55.3%	54.3%
Credit analysts (2023 SEC filing)	1,644	1,636

The number to remember is not the margin but the expense line: S&P’s ratings costs fell while billed volume rose by a quarter. A rating on a $5 billion bond does not need five times the analysts of a rating on $1 billion; it needs the same ones. That is why the margin does not stop at 60 percent in a boom year but climbs to 68 — and why it falls the other way in a drought.

The cycle the margin hides: 2022

Because the business is more cyclical than its share prices suggest in good years. In 2022 rates rose, issuance collapsed, and Moody’s ratings revenue fell from $3.8 billion to $2.7 billion — down 29 percent in a single year. Adjusted earnings per share fell from $12.29 to $8.57, down 30 percent, and the stock went from $391 at the end of 2021 to $235 at the October 2022 low, down 40 percent. S&P Global lost 39 percent over the same stretch. Anyone who bought the shares of a “fee monopoly” in 2021 owned the shares of an issuance cycle in 2022.

Moody’s Investors Service	2021	2022	2025	H1 2026
Ratings revenue ($ billion)	3.8	2.7 (−29%)	4.1 (+9%)	1.26 in Q2 alone
Transaction share	~72%	~58%	67%	71% (Q2)
Adjusted MIS margin	~63%	~55%	63.6%	68.3% (Q2)
Adjusted EPS, group ($)	12.29	8.57	14.94	2026 guidance: 16.50–17.00
Share price, year-end / low ($)	391	235 (low, Oct 14)	507	469 (Sept 18, 2026)

Recurring fees cushion this, but they do not cushion all of it: a third of revenue is stable, two-thirds depend on whether companies happen to be selling bonds. And that depends on rates, mergers and maturity schedules — none of which the agencies control. The analogy to a gold miner’s leverage is not accidental: the agency owns a fee on a volume it does not steer, with costs that are largely fixed. Volume up a quarter, profit up 28 percent. Volume down a quarter, profit down by more than a third.

The market in numbers: 92.9 percent

Once a year the SEC counts who has how many ratings outstanding. As of December 31, 2025, 92.9 percent of all outstanding U.S. ratings came from S&P, Moody’s and Fitch; at the end of 2023 it was 94.15 percent, and in 2007 — the first year of the count — 98.8 percent. In the European Union, where the regulator ESMA measures by revenue, S&P took 50.42 percent in 2025, Moody’s 29.63 and Fitch 11.82 — 91.87 percent between them in a market with more than two dozen registered agencies. And by revenue across the ten registered U.S. agencies, the Big Three’s share was 91.9 percent in 2023.

Share of outstanding U.S. ratings, end-2023	S&P	Moody’s	Fitch	Big Three	Largest challenger
Corporate issuers	43.5%	26.1%	16.8%	86.4%	Egan-Jones 6.7%
Financial institutions	39.3%	25.4%	24.0%	88.7%	DBRS 5.9%
Insurance companies	31.6%	12.3%	14.7%	58.6%	A.M. Best 33.4%
Asset-backed securities	20.2%	27.5%	19.5%	67.2%	DBRS 22.9%, KBRA 9.8%
Government securities	55.2%	33.5%	10.2%	98.9%	—
All ratings	50.0%	31.7%	12.5%	94.2%	DBRS 3.0%

The table carries two messages that contradict each other. First: in government securities — 76.8 percent of all ratings counted, because every American municipality has dozens of bond series outstanding — there is effectively no competition, and those ratings roll on for decades with annual fees attached. Second: in asset-backed securities, the segment that set off the crisis in 2008, the Big Three now hold only two-thirds.

Where the duopoly has already broken — and why nobody notices

The break has a place and a date. After 2008, S&P and Moody’s lost their most important asset in structured finance, credibility, and issuers discovered that investor mandates in that segment often required “an NRSRO rating” rather than “a rating from S&P or Moody’s.” Kroll Bond Rating Agency, founded in 2010, and Canada’s DBRS, part of Morningstar since 2019, filled the gap. In commercial mortgage-backed securities KBRA has held more than 40 percent of the market by volume every year since 2022 and in 2023 ranked ahead of both Moody’s and S&P; in single-borrower CMBS, DBRS has held more than a quarter since 2022. That is real competition, and it has pushed prices down there.

Why does the group P&L not feel it? Because structured finance is now only $151 million of Moody’s $1,260 million quarterly revenue, twelve percent, and because the competition has not reached the place where the money is made — corporate finance, $651 million at Moody’s alone in the quarter. Egan-Jones, the largest challenger in corporate ratings, holds 6.7 percent of outstanding grades and, per its own SEC filing, employs 20 analysts. Moody’s employs 1,644. The two-ratings convention protects the segment where it is most valuable, and it has lost none of its force there since 2008.

The sins of 2008 and the laws that changed nothing

The crisis ledger is familiar and still astonishing every time. Enron carried an investment-grade rating until four days before it filed for bankruptcy in November 2001. Lehman Brothers was still rated A on the morning of September 15, 2008. And hundreds of billions of dollars of mortgage securities carried a AAA they lost within months. The Financial Crisis Inquiry Commission wrote in 2011 that the three agencies were “key enablers of the financial meltdown”; journalists found internal emails in which analysts mocked the models they were simultaneously applying.

The bill was paid in cash, not in market share. S&P settled with the Justice Department and 19 states in 2015 for $1.375 billion, plus $125 million with the California pension fund CalPERS — $1.5 billion, less than half a year of today’s ratings revenue. Moody’s paid $864 million in January 2017. Both firms denied wrongdoing. No executive was charged.

Congress tried twice. Section 939A of the 2010 Dodd-Frank Act ordered federal agencies to strip references to credit ratings from their rules — a provision that, on paper, removed the model’s foundation. In practice the agencies found no substitute and private-sector mandates stayed as they were. The EU wrote its own regulation in 2009, put the agencies under ESMA in 2011 and added civil liability and mandatory rotation in 2013 — rotation that applies only to re-securitizations, a niche. Issuers keep paying, buyers keep not ordering, and the margin has risen by roughly 25 percentage points since. Rarely has an industry emerged so unpunished from a crisis it helped cause.

What has changed is the relationship with the sovereign itself. S&P stripped the United States of its AAA in August 2011, Fitch in August 2023, and Moody’s, last of the three, on May 16, 2025. The issuer whose rules created the duopoly is now a AA credit at all three agencies.

The 2026 issuance boom: AI pays the fee

The reason for the record quarter is in every hyperscaler’s balance sheet. Between 2020 and 2024, Alphabet, Amazon, Meta, Microsoft and Oracle issued roughly $35 billion of bonds a year on average. In 2025 it was $93 billion, and by July 31, 2026 already $132 billion — including a single multi-tranche deal of roughly $53 billion, among the largest corporate bond sales ever, and a rare century bond. The cause is simple: the five companies’ data-center capital spending is heading for about $800 billion in 2026, double 2025, and for the first time exceeds their operating cash flow.

For the agencies this is perfect clientele: investment-grade issuers with volumes for which a rating fee, even before the cap, is a rounding error. S&P billed $560 billion of investment-grade issuance in the second quarter, up 27 percent, and explicitly named “large transactions to finance AI infrastructure expansion” as the driver. Moody’s raised its 2026 guidance for rated issuance in July from “low-single-digit” to “mid-single-digit” growth, and its adjusted EPS guidance to $16.50–17.00. Add a refinancing wave: the bonds sold at near-zero rates in 2020 and 2021 mature between 2026 and 2028 and must be replaced — and every replacement is a new fee.

The bear argument sits in the same number. If AI capex stops growing in 2027 and merely holds, part of the issuance volume that carried the record quarter goes away. The agencies earn on issuance, not on the stock of debt, and a data center only needs to be financed once.

The new front: private credit and the agency with 20 analysts

The real challenge to the model comes not from a competitor but from a market that bypasses the public rating altogether. U.S. life and annuity insurers hold roughly $1 trillion of their $6 trillion in invested assets in private credit, and about $419 billion of that carries so-called private letter ratings — grades an agency produces only for the issuer and the insurer, never published, which the insurer may use to set its regulatory capital. The NAIC, the insurance regulators’ body, compared those grades with its own assessments and found the smaller agencies rated on average three notches higher than the regulator, in some cases six. According to a Bloomberg report, Egan-Jones rated more than 3,000 such private credits in 2024 — with the 20 analysts mentioned above.

It is the mechanism of 2006 one floor down: the issuer shops for the agency that gives the best grade, and the buyer who needs the grade wants it to be good, because a good grade spares its capital. For S&P and Moody’s the segment is both opportunity and threat. Opportunity, because the NAIC is tightening its requirements for private ratings from 2026 and insurers will prefer the big houses’ grades if the small ones stop being recognized; Moody’s has named private credit as a growth area for two years. Threat, because the duopoly rests on the public bond market — and the share of corporate borrowing that never reaches that market has been rising for a decade.

February 2026: AI as a threat — but to which half?

On February 10, 2026, S&P Global reported its annual results and guided 2026 adjusted EPS to $19.40–19.65, about one percent below consensus. The stock fell 13 percent that day and reached $367 on February 11 — 28 percent below its January 16 high. Moody’s dropped 5 percent the same day and also bottomed on February 11, at $410, 24 percent below its January high. The trigger was a guidance number; the subject was something else: the fear that large language models would hollow out both companies’ data businesses, because an analyst with an AI agent can now pull from annual reports what used to require a terminal from S&P Market Intelligence or Moody’s Analytics.

The fear is legitimate, but it hits the two halves of each company very differently. Moody’s Analytics is 42 percent of group revenue at a 33.6 percent margin; S&P Market Intelligence produced $1,290 million of second-quarter revenue and $293 million of operating profit, a 23 percent margin. Those are data subscriptions, and for data subscriptions a model that reads PDFs is a competitor. The ratings business sells no data; it sells a regulatory permission. No language model can make a bond eligible for an insurer’s portfolio, because no language model is an NRSRO. The 68 percent margin rests on a statute, not a database — and AI lowers the cost of that business while leaving its prices untouched. In February the market punished the whole company for the half that is vulnerable.

S&P Global drew the conclusion and spun off Mobility, the old Carfax business, on July 1, 2026 as Mobility Global; the group now runs on four divisions, of which Indices at a 70 percent margin and Ratings at 68 are the two that carry the model. Post-spin guidance for 2026 is $17.50–17.75 of adjusted EPS and more than $7 billion of share repurchases, about six percent of market capitalization.

Valuation: what 28 times earnings assumes

Moody’s closed at $468.59 on September 18, 2026 — an $81 billion market value, down 7.7 percent year to date in a year in which the S&P 500 has gained almost 12 percent. On the midpoint of guidance, $16.75, that is 28 times earnings; on expected free cash flow of $2.7–2.9 billion, a 3.4 percent cash-flow yield. S&P Global closed at $405.32, $119 billion, 23 times its $17.63 guidance midpoint — though its 17 percent year-to-date decline includes the spin-off, whose shares were credited to holders’ accounts. Berkshire Hathaway owns 24.67 million Moody’s shares, 14.2 percent of the company, at a cost of $248 million; the position is worth $11.6 billion today. Warren Buffett nearly halved it after 2008 and has not touched the remainder since 2013.

Twenty-eight times earnings for a business with a 68 percent margin and a statutory moat is not expensive if volume holds. It is expensive if 2027 turns out like 2022. Three scenarios for Moody’s, each with an assumption on issuance and on the multiple the market would grant such a year:

Moody’s 2027 (our estimates)	Bear	Base	Bull
Issuance assumption	−25% (a 2022 repeat)	flat	+10% (refinancing + private credit)
Ratings revenue ($ billion)	3.6	4.6	5.0
MIS margin	56%	65%	69%
Adjusted EPS ($)	~13	~18	~20
Multiple (P/E)	22×	26×	30×
Implied price ($)	~285	~470	~600
Change from 469	−39%	±0%	+28%

The picture is asymmetric, and not in the buyer’s favor: in the base case the shareholder earns the profit growth and nothing beyond it; in the bear case they lose what the stock lost in 2022; in the bull case they need a record year and a re-rating at the same time. The bear case is no exaggeration — it is a rerun of a year four years old. And the interest rates that choked off issuance in 2022 are not lower in 2026; they are higher.

What it means for investors

The ratings business is one of the few whose moat sits not on the balance sheet but in the statute book — and that moat held through 2008, through Dodd-Frank and through EU regulation. Whoever buys the shares buys three things: a fee that swings with issuance volume; a surveillance fee that barely swings; and a data business whose future AI is currently negotiating. The mistake of 2021 was to take the first for the second. The mistake of February 2026 may have been to take the third for the whole company.

For U.S. investors both names trade on the NYSE and pay qualified dividends — 0.9 percent at Moody’s, about 1.0 percent at S&P Global — taxed at 15 or 20 percent depending on bracket, which makes the yield almost irrelevant to the case; these are buyback stories, with S&P Global retiring six percent of its shares this year and Moody’s up to $3 billion. In a taxable account the sensible framing is a multi-year holding whose return arrives as price appreciation, not income. Whoever balks at the multiple can find the model in diluted form: MSCI and the London Stock Exchange Group sell index licenses on similar logic, and Berkshire Hathaway holds 14 percent of Moody’s at a price no investor will see again.

The arithmetic at the end is the arithmetic at the start. 8.35 basis points on every bond, paid by the rated, required by law, doubled by the convention that there must be two. No one could invent this business — it came into being because a regulator needed a name in 1975, and it has since survived every crisis it helped cause. The question for an investor is not whether the fee survives. It does. The question is how much volume flows through it in 2027, and whether 28 times earnings is a price that can absorb both answers.