S&P Global

S&P Global does four things. Two of them are among the finest businesses that exist, one is good, and one is the reason the shares have fallen twenty per cent.

The first is credit ratings. A company that wants to sell a bond needs a rating, because the institutions that buy bonds are required — by their own mandates and by regulation — to know what it is. There are effectively two firms whose opinion counts: S&P and Moody's. This is not a market in the ordinary sense. It is a licensing regime with two licensors, and the fee is charged on the size of the issue rather than on the work involved.

The second is indices. S&P Global owns the S&P 500. Not a fund tracking it — the index itself. Every exchange-traded fund, every pension mandate, every derivative written against that benchmark pays a licence fee, forever, calculated on the assets tracking it. ★ There is no substitute, because the substitute would be a different index, and the whole point of a benchmark is that everybody agrees on which one it is.

The third is Commodity Insights — the Platts price assessments that the world's oil and gas contracts are settled against. The same idea in a different market.

And the fourth is Market Intelligence, the data and analytics division, whose flagship is the Capital IQ terminal. This is the ordinary business: it competes with Bloomberg, FactSet, MSCI and a great many others, on features and price.

★★ That is the entire argument of this analysis, and it fits in one line. The market has taken about a fifth off a company whose two irreplaceable franchises are untouched, because it is worried about a terminal that is roughly six per cent of revenue.

One housekeeping note before we go further. S&P Global spun off its Mobility division on 1 July 2026, leaving the four divisions above. That is why the consensus revenue estimate for 2026 ($14.86bn) is lower than 2025's reported $15.34bn — the business did not shrink; a division left. Anyone reading the estimate line without knowing that will draw exactly the wrong conclusion.

Let us state the bear case at full strength, because it is not silly.

Capital IQ is a screen that a junior analyst uses to look things up. Company financials, comparable transactions, ownership records, filings. It is expensive, it is sold by subscription per seat, and a great deal of what it does is retrieval and summarisation — which is precisely what large language models have become very good at. If an AI assistant can pull a company's filings, build the comparable table and write the summary, the argument goes, why would a firm keep paying for thousands of terminal seats?

That fear is legitimate and we are not going to dismiss it. Nor is it abstract: the head of Market Intelligence, Saugata Saha — who had been rolling out the division's own AI tools, ChatIQ and Document Intelligence — announced his departure in May 2026. Analysts at RBC noted that Market Intelligence revenue and margins fell short in the quarter, alongside softer ratings transaction revenue. This is a division under genuine pressure with a leadership hole in it.

Now the size of it.

So the honest conclusion is a split one. About a third of this company faces a real technological threat and its margins are already showing it. About two thirds faces essentially none, because its moat is written into regulation and market convention rather than into software. The share price has been marked down as though the first number were the whole company.

Having defended the company against the fashionable objection, here is the unfashionable one, and we think it is considerably more serious.

In February 2022 S&P Global completed a $44 billion all-share merger with IHS Markit — one of the largest deals in the history of financial information. It issued roughly 77 million new shares to pay for it, raising the count by about a quarter. Look at what has happened since.

Four years. Revenue up eighty-five per cent. Earnings per share up seventeen — about four per cent a year. And free cash flow per share, which is the fairer measure because it excludes the non-cash amortisation of what was bought, up twenty-one per cent, or about five per cent a year. Either way, a merger that nearly doubled the company has so far delivered mid-single-digit progress for the people who own it.

And look at what it cost to hold even that line. In 2022 alone S&P Global spent $12.0 billion repurchasing stock — buying back the shares it had just issued. Since then it has spent a further $3.3bn, $3.3bn and $5.0 billion in 2025, with more than $7 billion planned for 2026. A great deal of the last five years' free cash flow has gone on undoing the dilution the company created.

There is a fair defence and we should give it. Integrations of that size take years; the amortisation of acquired intangibles depresses GAAP earnings without consuming cash; adjusted earnings per share grew 23% in the second quarter; and the Mobility spin-off is exactly the kind of pruning a management team does when it has decided to focus. The 2024 and 2025 free cash flow figures — $5.57bn and $5.46bn against $3.56bn in 2021 — show the machine working.

But an investor should hold both facts at once. This is a company with an extraordinary moat and an ordinary record of converting acquisitions into per-share value. The first is why we are constructive. The second is why we are not more so.

Three numbers in that panel are worth dwelling on.

A 46.3% operating margin is not something a competitive business earns. It is what a licensor earns. For comparison, the best consumer brands we have written about run in the twenties and thirties; Zoetis, one of the highest-margin businesses in medicine, earns 37%.

Free cash flow per share exceeding earnings per share tells you the accounting is conservative rather than flattering — the gap is amortisation of the IHS Markit intangibles, which is a real historical cost but not a current one. ★ When a company's cash exceeds its reported profit, the reported profit is usually understating the business.

And the divisional guidance for 2026 shows where the growth actually is:

★ Note the shape of that table, because it is the opposite of what the share price implies. The fastest-growing division is the one with the strongest moat, and even the division everybody is worried about is guided to grow at five and a half to seven per cent. Nothing here is shrinking.

⚠️ The genuine soft spot is ratings, and it is not AI. Ratings revenue moves with how many bonds get issued, and issuance moves with interest rates and corporate confidence. Transaction revenue growth weakened in the quarter. That is ordinary cyclicality in a business with an extraordinary moat — the fee per issue is safe; the number of issues is not.

The intangibles figure needs a word, because 82.7% of assets sounds alarming and mostly is not.

When you buy a business for $44 billion whose value lies in customer relationships, data and brands rather than in factories, accounting requires you to record that value as goodwill and intangible assets. S&P Global's real assets — the ratings franchise, the S&P 500, the Platts assessments — barely appear on the balance sheet at all, because they were built rather than bought. The balance sheet records what was purchased, not what is valuable.

★ Which is why price-to-book of 4.14× tells you almost nothing here — and why we lean on free cash flow instead. This is the mirror image of Chubb, published this morning, where price-to-book is the single best measure. The right metric depends entirely on what kind of thing the company is.

The risk that is real: goodwill of this size can be written down if the acquired business underperforms, and a large impairment would be an admission about the merger. Nothing in the current numbers suggests one is coming. ⚠️ But it is why we did not score financial strength a nine.

S&P Global yields 0.88%. Nobody owns it for income, and the payout ratio of 23.5% says the board agrees. The dividend is in no danger whatsoever — but the increases have done something we have now seen five times this month, and it is worth putting on the record.

A one per cent increase is not a capital-allocation decision. It is a streak being maintained. It was declared in February 2026 — the same month the company issued 2026 guidance that came in below expectations and the shares began their fall. The board knew what the year looked like before the market did, and it said so with a penny.

★ This is now the fifth time this month we have found the same signal. Zoetis raised 15% for four years and then 6%, five months before cutting guidance. Rexford went from 31% to 1.2%. T. Rowe Price from 11% to 2.4%. Chubb, by contrast, is accelerating — and Chubb is the one whose results are strongest. The character of a dividend increase is one of the most reliable pieces of forward information a board publishes, and almost nobody reads it.

What would force a cut. Nothing we can construct. Free cash flow would have to fall by nearly 80%. ★ The correct reading of this dividend is that it is a rounding error the board maintains for the sake of a fifty-year record, while the actual capital return happens through a buyback five times its size — and that buyback is now being executed at a share price a fifth below February's.

So what does 24.7× assume? It assumes Market Intelligence keeps growing at five to seven per cent rather than rolling over, that ratings issuance does not freeze, and that indices carry on compounding at twelve to fourteen. Two of those three are close to structurally assured. The first is the open question, and it is the one you are being paid a fifth off to take a view on.

★ This is not a cheap stock and we are not going to call it one. Twenty-five times forward earnings is a full price in absolute terms. But it is roughly a market multiple for one of the widest moats in existence, at a twenty per cent discount to where the same business traded seven months ago, on a fear about six per cent of revenue. We have seen worse trades.