S&P Global and the Half With the Moat

S&P Global reported the second quarter of 2026 before the open on July 28 and the stock fell around 5%. The wires had their reason ready: adjusted earnings came in below consensus, and full-year guidance appeared to have been cut by close to 10%. Both statements are, in the narrow sense, things that happened. Neither describes the company.

The guidance was not cut — the company got smaller, on a scheduled date, and the guide followed it. Mobility Global separated from S&P Global on July 1, 2026. The new full-year range covers continuing operations only, measured against a pro forma 2025 base the company published on July 6, and S&P Global states plainly that the two ranges are not comparable. The reported reduction in the guide is the same order of magnitude as the earnings that walked out the door with the spin. Consensus, meanwhile, was not struck on one company at all: the Q2 estimate the stock was marked against appeared in the press as $4.75, $4.95 and $5.02 depending on the vendor.

A separation is one of the few corporate events that breaks year-over-year comparison at the level of arithmetic rather than judgement. For one or two quarters the guide, the consensus and the prior-year base can each describe a different entity, and the difference between them is reported as performance. That is worth knowing, but it is not an investment thesis — it is a scheduling artifact that resolves itself by the next print.

Ratings and Indices supplied roughly three-quarters of the four segments' revenue growth in the quarter, at margins near 70%, while the two larger-headcount data businesses grew 6% and 3%. This is not a story about a good quarter and a bad quarter. It is the same split S&P Global has been running for years, made legible because the spin-off removed a segment and forced everyone to look at the parts.

If the split were specific to S&P Global it would be a stock note. It is not. Three of the five names in this cohort reported the same quarter within a week of each other, and each one shows a benchmark-owning segment growing at multiples of the data segment sitting next to it.

Now apply the framework to the same five names and watch it fail to sort them.

The scores cluster. They should — these are five of the most durable franchises in financial services, and a framework that separated them sharply would be telling you something false. But the clustering is also the point: a company-level score is an average across segments whose economics have almost nothing in common, and averaging is exactly the operation that destroys the information you need here. The moat pillars land the same way.

Read the grid honestly and it says the cohort is uniformly well defended. That assessment is accurate about each company as a whole and nearly useless for deciding which one to own, because the pillars are earned by one part of each business and credited to all of it. S&P Global's regulatory lock-in is a property of Ratings. Its network effects are a property of Indices. Neither pillar has much to say about a Capital IQ seat or a commodity data subscription — and those two lines are the majority of the company's revenue.

The distinction is not data quality, headcount, or how much AI a segment has shipped. It is who has to change their own documents in order to stop paying you.

• The mark is written into third-party paper. An S&P rating is named in bond covenants, fund mandates and capital rules; an MSCI index is named in prospectuses. Switching is not a procurement decision, it is an amendment — and the counterparty who must agree to it is not the vendor.
• The fee is levied on someone else's asset base. Index licensing and asset-based fees scale with AUM the licensee raised, not with headcount at the licensee. Revenue grows without a renewal conversation, which is why $6.35T of ETF-linked AUM converts into a 71.5% segment margin.
• There is no comparison shop. A buyer evaluating a data terminal can price three of them against each other. A buyer who needs the S&P 500 cannot buy an equivalent from anyone, because equivalence is the one property the product does not have.
• The data half fails all three. Market Intelligence, Moody's Analytics and MSCI Analytics sell genuinely good products into markets with substitutes, annual renewals, and buyers whose budgets are negotiated line by line. That is a decent business. It is not the same business, and it grows at a third of the rate.

The cohort contains a name where this argument inverts, and it inverts hard. FICO's Scores business was the purest benchmark in the group — a single number written into mortgage underwriting across the American housing finance system. Since July 2025 the FHFA has approved a competing score for conforming loans, Fannie Mae has eliminated minimum score requirements outright, and the VA, FHA and USDA have accepted verified rent and utility histories as primary evidence of creditworthiness. The benchmark half is the one under pressure. The growth is coming from the platform half, where Platform ARR grew 49% year over year in Q2 FY2026 — precisely the inversion of the pattern everywhere else in this table.

That does not break the thesis; it bounds it, and the bound is the most useful thing in this article. A benchmark embedded in private contracts — bond covenants, ISDA masters, fund prospectuses, investment mandates — can only be dislodged by persuading thousands of counterparties to amend their own documents one at a time. A benchmark embedded in a federal programme's rulebook can be dislodged by one agency changing the rulebook. Both look like regulatory lock-in on a moat matrix. Only one of them is durable against a change of administration, and FICO is the demonstration that the difference is not academic. ICE sits partway along that spectrum: its energy benchmarks are private-contract embedded, while MERS is closer to public infrastructure.

For S&P Global the separation is a genuine improvement in mix: what remains is more weighted toward the two segments that own something, and the buyback guided above $7B for 2026 compounds that on a shrinking share count. The complication is that the market worked this out three weeks ago. The stub separated on July 1 from a cum-distribution close of $424.42 on June 26, with Mobility then estimated at $25–33 per share, and it has traded up since — even after today's drop. The re-rating this article argues for has substantially happened.

So the framework's caution on SPGI now sits in the valuation pillar rather than the moat or growth pillars, and its scenario targets have been rebased to continuing operations — the old ones were struck against earnings that no longer exist. Within the cohort, the names where the scorecard still shows room are the ones whose benchmark half has not yet been re-rated for the same reason. Click through to any of the five to see where each sits today.

The generalisable lesson is smaller than the thesis and more portable. Any company that reports segments with a 15-point growth spread and a 30-point margin spread is two companies wearing one ticker, and both the market's multiple and our own composite score are averages over that difference. Spin-offs are useful mainly because they force the average apart for a quarter and let you see what was inside it. The mistake on July 28 was not mispricing S&P Global. It was reading a consolidated line on a company that had just told everyone the consolidated line changed.