S&P Global Puts Its AI Bridge on the Auction Block

A strategic review of Capital IQ and Kensho data lands beside the first hard AI revenue numbers the company has ever disclosed — and a stock still 20% below its peak.

SPGI · Earnings Call · 2026-10-27

The review is the story

S&P Global delivered a clean third quarter: revenue up 7%, adjusted operating margin at 53.5% (60 basis points better), EPS up 9%, and full-year organic growth guidance raised to 7.5%–8.5%. Yet the shares sit roughly 20% below the August 2025 peak even after a 10% bounce over the last 90 trading days. The reason is not the quarter. It is the sentence management buried in the release: a strategic review of alternatives for Capital IQ and its associated data businesses, including a potential separation.

Read the perimeter carefully. In scope: Capital IQ, the Kensho data and platforms business, Visible Alpha, and the financials/F&L data sets. Explicitly retained: RatingsDirect, RatingsExpress and Enterprise Solutions, the WSO/ClearPar lending suite that management calls its credit anchor. That is not a random divestiture. It is the piece of Market Intelligence that management spent the last two years positioning as the company's bridge to an AI-first market — and the piece whose users number more than 355,000 across 15,000 clients.

The reversal matters because the prior calls built the opposite case. In July, Eric Aboaf described AI monetization as multifaceted, noting that “MCP Connectors are up to 500 clients. Data usage is up 5x relative to the prior quarter” — Eric Aboaf, Chief Financial Officer · 2026-07-28. In February, Martina Cheung put the desktop at “about 6% of our enterprise revenue” — Martina Cheung, President and Chief Executive Officer · 2026-02-10 while arguing its strategic value ran well beyond that number. Now the company is testing whether separate ownership creates more value than integration.

The perimeter is the answer, and nobody has it yet

Management was admirably direct about how little is settled. Asked about size and structure, Aboaf demurred:

it is premature at this point. We are working down. Working through the perimeter of potential review and potential separation as we learn more, we will certainly come back to you.
Eric Walter Aboaf, Chief Financial Officer · 2026-10-27

The ambiguity extends to Kensho, the AI shop acquired in 2018 for roughly $550 million and since folded into Market Intelligence. Cheung said “we would look at the Kensho team as being very important to the perimeter” — Martina L. Cheung, President and Chief Executive Officer · 2026-10-27, which is a careful way of saying the people who built the AI retrieval layer may leave with the business. She also insisted that “we did not announce a change in the strategy” — Martina L. Cheung, President and Chief Executive Officer · 2026-10-27. Both things can be true, but they sit awkwardly together: the strategy was built around flexible distribution of proprietary data, and the distribution engine is now under review.

This is also not the first pruning. Mobility was spun on July 1; EDM and thinkFolio were sold; With Intelligence was bought. On the prior call Cheung hinted at exactly this logic, saying the company would consider where “subscale products will create greater value outside of our portfolio” — Martina Cheung, President and Chief Executive Officer · 2026-07-28. The difference is scale. CapIQ is not subscale; it is the front door to Market Intelligence.

The AI receipts are real — and portable

What makes the timing odd is that the AI monetization story finally has numbers. Management disclosed that AI users are spending faster than everyone else: ACV growth is 1.8x the non-AI rate in Market Intelligence and roughly 3x in Energy. Discrete AI products added about 50 basis points of growth in each of Energy and Kensho data and platforms. API call volume rose nearly 50% sequentially and more than 4,000% year over year. The Adaptive Retrieval product and AI product sales are no longer slideware.

That creates the central tension of the call. If the data portal is the best AI distribution asset in the portfolio, why sell it? The bull answer is that the market is not paying for it inside a $132 billion conglomerate, and a separation forces both halves to be valued on their own merits. The bear answer is that S&P is retreating to the businesses it can defend — benchmarks and ratings — just as the AI-native competition arrives. Management clearly wants the first reading: keep the benchmark IP, keep the credit workflows, and let someone else fund the desktop arms race.

The sales-cycle noise that dogged the prior quarter has also cleared. Cheung said renewals in Energy are back to pre-war levels and that AI-rights negotiations are normalizing with standard terms across the business. That removes a bear case that had been building for two quarters.

What remains is a benchmark compounding machine

Strip out the review and the quarter is a reminder of what S&P owns. Ratings revenue grew 2%, with build issuance up 8% and investment grade up 28%; hyperscaler issuance has reached roughly $240 billion year to date against a $250–$300 billion full-year expectation. Indices grew 20% — the 14th consecutive record quarter — with ETF AUM at $6.5 trillion and more than $600 billion of net inflows. The CBOE extension through 2051 adds a couple of points of index revenue growth in 2027.

The next optionality is prediction markets. Kathy Clay called the Kalshi relationship a preferred partnership, not exclusive, and stressed it is “very early days, but we are really excited to participate in these new markets in a meaningful way” — Catherine R. Clay, CEO, S&P Dow Jones Indices · 2026-10-27. This is a genuinely nascent theme globally — other reporters are only beginning to mention prediction markets — and S&P is positioning as both an index provider and a data source for resolution mechanisms. It will not move 2026 numbers, but it is the kind of adjacency that justifies the remainco multiple.

The contrast with the tape is stark. The global AI data center theme is deeply negative over 90 days across dozens of tickers, yet S&P's exposure is through data and benchmarks (Data Center Hawk, 451 Research) rather than capital equipment. Its shares are up 10% in the same window. The market appears to be treating S&P as an AI beneficiary without treating it as an AI capital-spending proxy.

The valuation already assumes something goes wrong

At 7.4x revenue, down about 32% year over year, the stock is closer to a cyclical industrial than a monopoly data compounder. The earnings power is intact: $4.1 billion of quarterly revenue, up 10%, and a 43.7% operating margin. Free cash flow was $1.3 billion in the quarter, flat year over year, and the company is on track for more than $7 billion of buybacks this year — double the original plan — funded partly with about $2 billion of new debt.

That debt sits against an $11 billion net debt position, so the review is also a capital-allocation event. A separation would give the remainco a cleaner balance sheet, a higher growth rate and, management argues, more flexibility to return capital. The risk is that the sum-of-the-parts argument only works if the discontinued piece fetches a full price; earnings calls are not auctions, and the NDAs signed do not guarantee a bid.

The quarter itself was good. The strategic question is better. S&P Global is simultaneously arguing that trusted data matters more than ever and preparing to sell the platform through which much of that data reaches customers. That is either a sharp focus on the moat or an admission that the AI bridge was more expensive to own than to rent. The market has 20% of drawdown already voting on the second interpretation — and a February update that will decide whether the first one gets the benefit of the doubt.