S&P Global sells the yardstick.

In 1860, Henry Varnum Poor published a book about American railroads and canals. A railroad could be an engineering achievement and a bewildering investment. Investors needed a way to understand the business behind the tracks. Poor assembled information that helped them do it. The interesting part of this origin story is the customer’s problem: before you decide where money belongs, you need facts you can put beside one another.

That problem has outlived the railroad boom. Today it occupies bankers comparing borrowers, investors examining funds and energy traders negotiating cargoes. S&P Global supplies ratings, datasets, benchmarks and software for those decisions. Its particular appeal is the combination: the company behind a familiar stock-market number also operates in the less visible business of making financial information usable.

The original comparison problem

There is a practical difference between having information and being able to compare it. Two companies can report similar activities under different headings. A useful database has to connect identities, definitions and periods before an analyst can reach a sensible conclusion. Compustat, introduced in 1964, made standardized historical company data a product. Capital IQ Pro now puts financials, filings, deals and research into the analyst’s working day.

The clientele explains the product. An investment bank needs comparable companies for a valuation. A lender needs evidence about a borrower. An asset manager needs to test a thesis against financial statements and market expectations. In a February 2026 case study, an unnamed global commercial bank described using Capital IQ Pro to support a private-credit team, combining private-company intelligence, loan-market information and banking data.

The useful outcome is quite ordinary: less time hunting and rekeying, more time examining the deal. That is also a good way to evaluate the service. Pick a real research assignment and follow it from first question to final spreadsheet. A large dataset matters only when the relevant records, permissions and tools fit the work.

Four ways to sell a common language

S&P Global’s businesses meet customers at different points in a decision. Ratings offers opinions on creditworthiness. S&P Dow Jones Indices supplies rules for measuring markets and licensing indices. Energy, including Platts, provides commodity reporting and price assessments. Market Intelligence combines data, analytics and enterprise workflows. Each helps people discuss a complicated market using shared reference points.

Platts illustrates the distinction. Its Dated Brent methodology describes how bids, offers and transactions inform a price assessment, including periods of limited trading. The product gives participants a reference for discussion and contracts. It cannot remove differences in cargo quality, delivery terms or timing. Using the number well requires understanding which market the number describes.

Those reference points generate different kinds of fees. Data subscriptions pay for continuing access. Ratings brings transaction and surveillance revenue. Index licensing can link revenue to assets or trading activity. Software and services add another route into the customer’s processes. The company reported $15.336 billion in 2025 revenue, including Mobility, which subsequently became independent.

Competition follows the task. Moody’s and Fitch compete in ratings; financial-data buyers also consider Bloomberg, LSEG and FactSet. Index and commodity customers have their own alternatives. S&P Global’s proposition is the breadth of its specialist information and the ability to place it inside recurring work. That breadth does not make every product the right answer for every desk.

The bill for being believed

Building this collection has been expensive. The IHS Markit acquisition closed in February 2022; acquisition accounting recorded $43.5 billion in aggregate consideration. In November 2025, S&P Global spent $1.8 billion on With Intelligence, adding alternative-investment information and workflows. Private markets extend the old comparison problem into places where financial information is harder to assemble.

Authority carries another kind of cost. In 2015, S&P and its parent agreed to a $1.375 billion settlement with the US Justice Department and state partners over pre-crisis mortgage-related ratings. The agreed facts described business concerns influencing rating-model decisions and positive ratings continuing despite awareness of deteriorating securities. Commercial incentives had compromised the judgment investors relied upon.

“Every word or decision can have far-reaching impacts.”

S&P Global’s published Integrity value

That history belongs beside the company’s stated values of discovery, partnership and integrity. It identifies a condition under which the whole arrangement becomes dangerous: users treat an authoritative opinion as certainty while the provider’s incentives weaken its independence. Better distribution cannot repair bad judgment. A rating still requires scrutiny, and a benchmark still depends on its construction.

Taking the data to the question

Under Martina Cheung, CEO since November 2024, the company has been changing both its boundaries and its interfaces. It completed the Mobility spin-off in July 2026. The same month, it reorganized Market Intelligence around Kensho Data & Platforms and Enterprise Solutions, explicitly responding to changing customer needs in an AI-driven market.

The delivery strategy is concrete. An August expansion with Microsoft connects licensed intelligence to Microsoft 365 Copilot workflows. October brought broader ChatIQ research capabilities and spreadsheet assistance in Capital IQ Pro; presentation tools remained in Labs. On October 15, Visible Alpha’s broker-model data joined the AI Data Portal. Its Adaptive Retrieval support was still expected later in the year.

There is a lesson here for a much smaller business: start with a repeated, costly question, standardize the evidence and put the answer where the customer already works. The conditions matter. Coverage must match the question, licensing must permit the workflow and someone must check the result. An elegant AI answer built on mismatched definitions is simply a quicker way to be mistaken.