S&P Global SPGI

Reference table

1. The separation, and the five basis defects it created

On 1 July 2026, S&P Global completed the separation of Mobility Global Inc. The 8-K filed 2026-07-02 states it plainly:

> "...at 12:01 a.m. New York City time, the previously-announced separation (the 'Separation') of Mobility Global Inc. ('Mobility Global') from S&P Global Inc. ('S&P Global') became effective. The separation of Mobility Global, which comprises ... vehicle history data for the automotive sector, which operated under the S&P Global Mobility division (the 'Spin Business'), was achieved through S&P Global's distribution ... of 100% of the shares of Mobility Global common [stock] ... S&P Global stockholders of record received one share of Mobility Global common stock for every share of S&P Global common stock. Following the Distribution, Mobility Global became an independent, publicly-traded company with its common stock listed under the symbol 'MBGL' on the New York Stock Exchange, and S&P Global retains no ownership interest in Mobility Global."

And the 10-Q filed 2026-07-28 states the accounting consequence:

> "Effective July 1, 2026, our operations consist of four reportable segments: S&P Global Ratings ('Ratings'), S&P Dow Jones Indices ('Indices'), S&P Global Energy ('Energy') and S&P Global Market Intelligence ('Market Intelligence'). The results of Mobility are included through June 30, 2026. Beginning with the third quarter of 2026, the historical financial results of Mobility through June 30, 2026 will be reflected in our consolidated financial statements as discontinued operations in accordance with U.S. GAAP for all periods."

> "Additionally, beginning with the third quarter of 2026, results will reflect product transfers of 451 Research and Maritime & Trade from Market Intelligence to Energy ... and a small portion of expenses associated with the transfer of Credit Analytics products from Market Intelligence to Ratings."

So as of this dive, S&P Global has: separated a segment one day after the quarter close, not yet restated its history, renamed another segment, and announced product transfers between two more. Every one of the five defects below follows from that.

Defect 1 — inc_q and earn_cal report the same quarter on different bases

We proved the reconciliation from the 10-Q's own segment table: Mobility's June-quarter revenue was exactly $468 million, and $4,146M − $468M = $3,678M. Both figures are right; they are answers to different questions; and the payload gives no indication which is which.

Defect 2 — earn_cal compares a continuing-operations actual against a total-basis estimate

earn_cal for 2026-07-28 reports revenueActual: 3,678,000,000 against revenueEstimated: 4,093,128,000. That is a 10.1% "revenue miss." It did not happen. The actual is post-separation and the estimate is pre-separation. The next row, for 2026-10-29, carries revenueEstimated: 3,636,628,000 — a continuing-operations figure. The estimate basis switches mid-series, which is precisely the PLD Core-FFO-to-GAAP failure named in the data contract, and it manufactures a large miss out of nothing. Note that the EPS rows do not have this problem: $4.83 actual against $4.81 estimated is a genuine like-for-like beat.

Defect 3 — est.revenueAvg for 2026 is below 2025 actual

Consensus revenue for the year to December 2026 is $14,906M, against $15,336M actually delivered in 2025. A reader would conclude S&P Global is shrinking 2.8%. It is not: the estimate blends six months with Mobility and six months without, against a full prior year with it. The 2027 estimate of $15,665M cannot be compared to either. No revenue growth rate derived from the est block is used anywhere in this dive.

Defect 4 — seg_prod omits the entire Energy segment for FY2025

The FY2024 vendor block contains "Commodity Insights $2,142M". The FY2025 block contains no equivalent line. The company renamed Commodity Insights to Energy, and the vendor lost the segment in the rename. The gap of $2,099M is exactly Energy's $2,299M less the $200M intersegment elimination. This is the MTZ vanished-segment failure mode, and it is unusual to be able to name the missing amount to the dollar. We use the 10-K's segment table throughout and do not use seg_prod for FY2025.

Defect 5 — the street price target may be pre-separation

pt.targetConsensus is $541.22, 31.2% above a price of $412.53, with a low of $498 — itself 20.7% above spot. S&P Global shareholders received a separate listed security on 1 July. A target struck before that date values S&P Global plus Mobility Global and is not comparable to a post-separation share price. The payload contains no lastUpdated field on pt, so we cannot establish the basis. We report the consensus, flag the uncertainty, and decline to lean on it as evidence of upside.

seg_geo is clean and is used. FY2025: United States $9,322M, Europe $3,531M, Asia $1,640M, Rest of World $843M = $15,336M, matching reported revenue exactly, with the 60.8% home market present. The 10-Q's June-quarter equivalent — US $2,530M, European region $952M, Asia $440M, Rest of the world $224M = $4,146M — also reconciles exactly.

2. The business that remains

FY2025 segment results, from the 10-K filed 2026-02-11 — the authoritative table, since seg_prod is defective:

Three things follow and they are the whole quality case.

First, Ratings and Indices are 65.0% of segment operating profit on 43.0% of gross revenue. At 63.8% and 68.7% operating margins, these are among the most profitable operating segments disclosed by any company in the S&P 500. They are also the two most defended: ratings by regulation and mandate embedding, indices by benchmark network effect.

Second, the company separated its worst segment. Mobility earned a 21.6% margin against a 43.0% company average. Post-separation, on fiscal 2025 figures, the remaining four segments earned $6,218M of operating profit on $13,589M of revenue — a 45.8% margin against 43.0% including Mobility. The separation is accretive to margin by roughly 280 basis points before any cost reallocation.

Third, Market Intelligence is the weak point and it is large. $4,916M of revenue — 36.2% of post-separation revenue — at a 20.2% margin, contributing only 15.9% of post-separation segment profit. It grew 6.0% in the June quarter, the second-slowest of the five. It is also precisely the business the knowledge base's only recent claim identifies as being disrupted.

The June 2026 quarter, from the 10-Q's disaggregation table:

Every segment grew. The 10-Q attributes Ratings' growth to "higher corporate bond ratings revenue primarily driven by strong investment grade issuance" and to "an increase in surveillance revenue, higher Ratings Evaluation Service activity and an increase in revenue at our Crisil subsidiary." Indices grew on "an increase in asset linked fees revenue driven by higher levels of assets under management for ETFs and mutual funds."

Within Ratings, note the composition: non-subscription/transaction revenue was $746M against $597M, up 25.0%. That is bond-issuance-linked and it is running hot. A reader should treat the Ratings segment's 16.6% growth as containing a cyclical component that will not repeat in a weaker issuance year — the same line fell in 2022 and 2023.

And within Indices: asset-linked fees were $348M against $286M, up 21.7%. That is the closest thing to a structural compounder in the company.

Management's own half-weighted guidance on what the post-separation basis looks like, from the 10-Q: "Had the impact of the product transfers discussed above been effective for the three and six months ended June 30, 2026, revenue would have increased 6% and 7% ... operating profit would have increased 16% and 64%." That "6%" is the most useful single number available for what growth looks like on the go-forward basis, and it is materially below the reported 10.4%.

3. Balance sheet, cash flow and the enterprise-value question

Filing-verified, 30 June 2026 (10-Q filed 2026-07-28):

From the fiscal 2025 balance sheet: goodwill $36,475M and intangibles $16,271M — $52,746M against total assets of $61,200M. intangiblesToTotalAssetsTTM is 82.7%, the highest in this batch, and tangibleAssetValueTTM is minus $15,403M. That is the IHS Markit merger, and it is why price-to-book of 3.85x is a number about accounting rather than about value.

The enterprise value, and a correction the vendor does not make. km_ttm.enterpriseValueTTM is $133,220,844,000, being market capitalisation of $121,613,844,000 plus $11,607,000,000 of net debt. We reconstructed that figure: total debt $15,170M plus lease liabilities of roughly $571M less cash of $4,134M = $11,607M. It is arithmetically defensible.

What it omits is $5,024 million of redeemable noncontrolling interests, which sit in the mezzanine between liabilities and equity on the 30 June balance sheet (and $4,917M at year-end, which bal_a correctly reports as minorityInterest). A minority interest is a third-party claim on the enterprise and belongs in enterprise value. Including it, and the $115M of ordinary noncontrolling interests:

The understatement is $5.14 billion, 3.9%. It matters more than the percentage suggests, because the minority interest is a claim on the group's equity, and a reader comparing S&P Global's EV/EBITDA to a wholly-owned peer's would be comparing 16.2x to a number that should be 16.8x.

Cash flow and capital returns:

Capital expenditure of $195 million on $15,336 million of revenue — 1.3% — is the lowest in this batch and is the defining financial characteristic of the business. Free cash flow of $5,456M is a 35.6% margin and a 4.6% yield on the current market capitalisation.

The buyback is the EPS engine. $5,001M in fiscal 2025 is 4.1% of the current market capitalisation in a single year, and H1 2026 added $1,933M. Diluted shares went from 318.9 million (FY2023) to 311.9 million (FY2024) to 305.1 million (FY2025) to 295.5 million in the June quarter — a 7.3% reduction in two and a half years, roughly 3% a year. That alone converts mid-single-digit revenue growth into high-single-digit EPS growth before any margin expansion.

Dividend: dividendPerShareTTM and lastDividend are both $3.86, yielding 0.94% with a 23.5% payout ratio. Consistent with every other name checked in this batch, the vendor's lastDividend equals its trailing-twelve-month figure exactly, which means it is a TTM total rather than the last declared quarterly dividend — implying roughly $0.965 a quarter. This is not an income stock; the return comes from the buyback.

4. Valuation — priced in or room?

At $412.53 (market cap $121.61B, ~294.8M shares, corrected enterprise value $138.36B):

Estimate coverage is thin for a $122 billion company and gets worse fast: 9 analysts on FY2026 EPS, 13 on FY2027, 5 on FY2028, and ONE each on FY2029 and FY2030. The grades block shows 28 analysts covering the stock, so the est block is capturing a third of them. The FY2029 and FY2030 rows, each resting on a single analyst, are excluded from every conclusion.

est.ebitdaAvg and est.ebitAvg are fixed-ratio fabrications and are REJECTED. ebitdaAvg is exactly 49.33% of revenueAvg in every year from FY2026 to FY2030; ebitAvg is exactly 42.07%. The ratios happen to be close to reality (actual FY2025 EBITDA margin 50.2%), which makes them more dangerous rather than less. Worse, netIncomeAvg EXCEEDS ebitAvg in both FY2029 ($7,639M against $7,438M) and FY2030 ($8,209M against $7,830M) — arithmetically impossible. The data-contract defect is present. All forward valuation runs on epsAvg alone.

A note on what the EPS estimates cover. The FY2026 epsAvg of $18.079 against first-half earn_cal actuals of $9.53 and a Q3 estimate of $4.42 implies a fourth quarter of $4.13 — lower than either reported half-year quarter, which is consistent with Mobility being removed for the second half. The FY2026 EPS estimate is therefore a blended pre- and post-separation number, and only the FY2027 estimate of $20.388 is a clean full post-separation year. That is why the fair values below are anchored on FY2027 rather than FY2026.

Peer context. The vendor peer set contains Moody's (MCO) — the direct comparable and the other half of the ratings duopoly — alongside Bank of America, BlackRock, Capital One, Interactive Brokers, Santander and Toronto-Dominion, five of which are banks or brokers and are not comparable at all. No peer multiples are supplied in the file, so no peer-multiple comparison is drawn. The knowledge base repeatedly pairs S&P Global with Moody's and, in one claim, prefers MSCI's index business to S&P's.

4a. What today's price assumes (the inversion)

At $412.53 — 20.2x the FY2027 consensus — the price embeds:

• Adjusted EPS reaches $20.388 in 2027 and $23.229 in 2028 — +12.8% then +13.9% — on a clean post-separation basis. (Consensus; 13 and only 5 analysts.) Roughly 3 points of that comes from the buyback alone, given a 3%-a-year share retirement rate, which means the operating business needs to deliver only high single digits.
• Ratings transaction revenue does not mean-revert. (Our number.) It grew 25.0% in the June quarter on "strong investment grade issuance". Ratings is 45.7% of segment operating profit. A normal issuance year would remove several points of company growth, and this line has fallen outright in past cycles.
• Indices asset-linked fees keep compounding near 20%. (Filing-confirmed at +21.7%.) This is the structural leg and the one we are most comfortable with — it tracks assets under management in index products, not the issuance cycle.
• Market Intelligence, 36% of revenue at a 20.2% margin, does not deteriorate. (Our number; it grew 6.0%.) This is the most fragile assumption in the price, because it is both the largest revenue segment and the one directly named by the only recent knowledge-base claim as a disruption target — a claim whose evidence is that comparable data-services incumbents have already suffered "brutal drawdowns".
• The market pays roughly 20-24x forward earnings for a business it has historically paid more for. (Our number.) At 17x FY2027E the stock is $347; at 26x it is $530.

4b. The return bridge (why the multiple moves)

Expected return over the next twelve months decomposes as: EPS growth (+13.9%, from FY2027E $20.388 to FY2028E $23.229) + multiple drift (EXPANSION, from 20.2x to roughly 21.5x on the respective forward year, +6%) + shareholder yield (+5.0%, being 0.94% dividend plus roughly 4.1% buyback) ≈ +13% to +16%.

Our base assumes modest multiple EXPANSION, and it is the fragile leg — stated explicitly. The argument for it is that the post-separation company is structurally better (a 45.8% segment margin against 43.0%, having shed the 21.6%-margin division), that the current 20.2x is below where a ratings-and-indices duopolist has historically traded, and that a restated, clean reporting basis on 2026-10-29 removes a discount that exists partly because the numbers are currently confusing. The argument against it is that the market may be pricing a genuine structural threat to the data-services half of the company rather than a temporary reporting fog, in which case the multiple stays where it is or falls.

Notably, an unusually large share of the expected return here is shareholder yield: roughly 5.0%, of which 4.1% is buyback. That is the highest in this batch and it is the reason this name does not need multiple expansion to work — if the multiple simply HELD at 20.2x on FY2028E, the price would be $469 (+13.7%), which is essentially our base case. The base does not actually require the re-rating; the re-rating is what takes it to the bull case.

4c. Variant perception (where we differ, what would surprise)

• Our largest divergence is from the DATA rather than from the crowd, and it is unusually severe. Five separate defects in this payload — two revenue bases in the same quarter, an estimate compared against the wrong basis, a consensus revenue line below last year's actual, a missing $2.3 billion segment, and a price target of unverifiable vintage — mean that a reader screening S&P Global today would see a company shrinking 2.8%, missing revenue by 10.1%, and trading 31% below target. All three are artefacts. The company grew 10.4%, beat on EPS, and its target basis is unknown.
• We differ from the street on magnitude while agreeing on direction, and we say why. Consensus is $541.22, +31.2%, with all 28 targets above spot. Our base is $470, +13.9%. The gap is 15%, and a material part of it may simply be that the consensus target has not been restruck for the separation — which we cannot verify and therefore do not treat as either evidence for or against our number.
• We think the market is under-differentiating between the two halves of this company, and that is our only genuine analytical claim. Ratings (63.8% margin) and Indices (68.7%) are protected by regulation and benchmark network effects that no language model can replicate. Market Intelligence (20.2% margin, 36% of revenue) and Energy are data-and-analytics businesses with materially less protection. The stock has been sold as one thing. The knowledge base's single recent claim is aimed precisely at the vulnerable half, and it is right to be — but that half is 15.9% of post-separation segment operating profit, not 36%.
• Positive surprise that would force a re-rate: the 2026-10-29 restatement showing post-separation operating margins materially above 46%; Market Intelligence growth accelerating above 8%; index asset-linked fees holding above 20%; or the buyback being raised above the $5 billion fiscal 2025 rate.
• Negative surprise that would break the thesis: Market Intelligence growth falling below 4% or its margin below 18%; Ratings transaction revenue turning negative as issuance normalises; any impairment against $52,746M of goodwill and intangibles; or the restated post-separation earnings base coming in materially below the $20.388 the 2027 consensus implies.

Synthos fair values

All three anchors are multiples of the FY2027 consensus adjusted-EPS distribution — mean $20.388, low $19.863, high $22.355 — which at 13 analysts is both the best-covered forward year and the first clean full post-separation year. The FY2026 row is deliberately not used, because it blends two reporting bases.

• Bear ~$335 — 16.9x the FY2027 consensus LOW of $19.863. Cross-check: 14.4x FY2028E; roughly 10% below the 52-week low of $373.75. The scenario: the AI-disruption claim proves right about Market Intelligence, that segment's revenue stalls and its margin compresses from 20.2%; Ratings transaction revenue mean-reverts as investment-grade issuance normalises from a strong year; and the market applies a data-services multiple to 55% of the revenue. −18.8%.
• Base ~$470 — 23.1x the FY2027 consensus MEAN of $20.388. Cross-check: 20.2x FY2028E; a 4.9% free-cash-flow yield; corrected enterprise value of $190B on FY2028E revenue. Sensitivity, stated openly: 21x gives $428, 25x gives $510. The whole answer lives in a 21-25x band. The scenario: the restated basis shows a cleaner, higher-margin company; Ratings and Indices carry the growth; Market Intelligence grinds along at mid single digits; the buyback retires 3% a year; and the multiple recovers modestly toward where a duopolist trades. +13.9%, plus roughly 5.0% of shareholder yield.
• Bull ~$600 — 26.8x the FY2027 consensus HIGH of $22.355. Cross-check: 25.8x FY2028E; 6.4% above the 52-week high of $564.15 and 4% below the street's high target of $625. The scenario: the separation is recognised as accretive, index asset-linked fees keep compounding above 20%, the AI fear directed at Market Intelligence proves misplaced, Ratings benefits from a sustained issuance cycle, and the market restores the premium multiple a regulated duopoly historically earns. +45.4%.

Base is 13.9% above spot; asymmetry roughly 2.42:1 to the upside (18.8% down, 45.4% up) — the best skew in this batch. What stops it being a Buy is not the arithmetic; it is that the arithmetic rests on an estimate block with nine analysts on the current year, a revenue series that cannot be used at all, and a company that will restate its entire history in eighty-six days. Acting before that restatement is acting on numbers we have just spent a section demonstrating are unreliable.

5. Knowledge base — four distinct claims, three of them old, one of them recent and bearish

Raw entity hits: 5. Distinct after de-duplication: 4. Supplementary text sweep: 18 raw, roughly 11 discarded as homographs, 3 retained as competitor context.

The four distinct claims, verbatim, oldest first:

> 2021-10-22 · bullish · conviction 80 · horizon: thesis · entities: MCO, SPGI · speaker: Dev Kantesaria · channel: we_study_billionaires · speaker_role: independent

> "Moody's and S&P Global, his longest-held names, are a natural duopoly rating 90% of world debt, now insulated from underwriting litigation and helped by regulation; he thinks they have a lot more to go."

> (Recorded twice on the same date from the same speaker with near-identical wording; counted once.)

> 2023-05-29 · bullish · conviction 85 · horizon: principle · entities: FICO, Moody's, S&P Global · channel: business_breakdowns · no named speaker

> "Like Moody's/S&P in ratings, FICO is the accepted securitization standard — using an alternative costs ~0-50bps more interest, so the standard is entrenched."

> 2024-02-23 · bullish · conviction 65 · horizon: thesis · entity: SPGI · speaker: Brian Yakman · channel: business_breakdowns · speaker_role: independent

> "S&P Global shares the same unassailable ratings network effect as Moody's at similar size, though he's mildly concerned about its culture versus Moody's."

> 2026-02-03 · BEARISH · conviction 75 · horizon: thesis · entities: Thomson Reuters, FactSet, S&P Global, Verisk · channel: compound_and_friends · no named speaker

> "Legal/data-services incumbents that endlessly raised prices are being directly disrupted by [an AI vendor's legal tools — the vendor name is redacted per house rules]; Thomson Reuters and Verisk saw brutal drawdowns."

What this lane is, stated without inflation. The bullish half is excellent in quality and poor in freshness. The 2021 claim carries a named independent speaker with a disclosed long-held position and states the structural case as clearly as anyone could; the 2023 claim gives the mechanism — a 0-50 basis point interest penalty for using a non-standard rating, which is why the standard cannot be dislodged; the 2024 claim confirms it and adds a mild caveat about culture. But the most recent of the three is thirty months old, and the horizon tags are "principle" and "thesis", which decay over years rather than never.

The bearish claim is the only one written after the stock began falling, and it is the one that matches the price action. It names S&P Global inside a set of information-services incumbents — Thomson Reuters, FactSet, Verisk — and argues that AI tools are directly disrupting businesses whose model has been annual price increases on entrenched data subscriptions. It offers falsifiable evidence: Thomson Reuters and Verisk have already had "brutal drawdowns." S&P Global has since had one too: minus 26.9% from its high, minus 24.4% over twelve months.

Our reading, stated as a judgement rather than as a finding. The bearish claim is directionally right about the market's behaviour and imprecise about the company. S&P Global is not one information-services business; it is a regulated ratings duopoly (45.7% of segment profit at a 63.8% margin), a benchmark index franchise (19.3% at 68.7%), and two data businesses (30.2% combined). The AI-disruption argument is a good argument against the last group and a weak argument against the first two — you cannot replace a Nationally Recognized Statistical Rating Organization or the S&P 500 index with a model. The market appears to have applied the argument to the whole company. That is the variant perception in Section 4c and it is the only genuine one this dive has.

Homograph discipline — roughly 11 discards from the 18 text hits, and one is new to the catalogue.

Three further claims are retained as competitor context rather than discarded, all from Brian Yakman (business_breakdowns, 2024-02-23, named independent speaker): that Moody's has an unassailable protocol network effect and 30-plus-year terminal-value certainty (conviction 82); that Moody's should grow GDP-plus with 3-4% annual price increases (conviction 70); and — notably — that he prefers MSCI's index business even over S&P Global's, calling MSCI "the gold standard for international indices" (conviction 65). That last is a mild negative for S&P Global specifically and is included for that reason.

Conclusion. Conviction Low, breadth 3, net positive-but-stale-and-turning. The structural case is well-evidenced and old; the only fresh claim is bearish and is aimed at the weakest third of the company. The lane supports the qualitative case in Section 2 and does not justify a valuation premium, and its recency tilt is a direct input to the Watch verdict.

6. Data integrity — what we rejected and why

Ten findings. This is the most basis-defective file in the batch, and every defect traces to one corporate action.

1. inc_q and earn_cal report the same quarter on different bases — both correct, neither labelled. inc_q Q2 2026 revenue is $4,146M, matching the 10-Q income statement exactly and including Mobility. earn_cal.revenueActual for 2026-07-28 is $3,678M, which is $4,146M less Mobility's disclosed segment revenue of $468M — continuing operations. No comparison between the two blocks is made in this dive.

2. earn_cal compares a continuing-operations ACTUAL against a total-basis ESTIMATE — a manufactured 10.1% miss. revenueActual: 3,678,000,000 against revenueEstimated: 4,093,128,000. The next row (2026-10-29) carries a continuing-operations estimate of $3,636.6M, so the basis switches mid-series. This is the PLD failure mode named in the data contract. The EPS rows are unaffected and remain like-for-like: $4.83 against $4.81 is a genuine beat.

3. est.revenueAvg for FY2026 ($14,906M) is BELOW FY2025 actual ($15,336M) — the entire revenue series is REJECTED. The estimate blends six months with Mobility and six without. No revenue growth rate from the est block appears anywhere in this dive, and the fair values are built on EPS only.

4. seg_prod OMITS the entire Energy segment for FY2025 — $2,299M of revenue, 15% of the company. The four supplied lines sum to $13,237M against reported revenue of $15,336M, a gap of $2,099M — exactly Energy's $2,299M less the $200M intersegment elimination. The company renamed Commodity Insights (present in the FY2024 block at $2,142M) to Energy, and the vendor lost the segment. This is the MTZ vanished-segment defect. The 10-K's segment table is used throughout instead.

5. pt.targetConsensus of $541.22 may be struck on a pre-separation basis — flagged, not used as evidence. S&P Global shareholders received one Mobility Global share per S&P Global share on 1 July 2026. A target set before that date values a company that no longer exists in that form. The pt block carries no lastUpdated field. We report the consensus and explicitly decline to treat its 31.2% implied upside as evidence.

6. est.netIncomeAvg EXCEEDS est.ebitAvg in two years — REJECTED. FY2029: net income $7,639M against EBIT $7,438M. FY2030: $8,209M against $7,830M. Arithmetically impossible for a taxpaying company. Both rows also carry only one analyst on EPS and are excluded entirely.

7. est.ebitdaAvg and est.ebitAvg are fixed-ratio fabrications — REJECTED. ebitdaAvg is exactly 49.33% of revenueAvg and ebitAvg exactly 42.07% in every year FY2026 to FY2030. The ratios happen to sit near the true margins, which makes them more dangerous, not less — a reader checking one year would conclude they were sound.

8. km_ttm.enterpriseValueTTM omits $5,139M of noncontrolling interests — corrected. The vendor's $133.221B is market capitalisation plus $11,607M of net debt (total debt $15,170M plus roughly $571M of lease liabilities, less $4,134M of cash) — arithmetically defensible. But the 30 June balance sheet carries $5,024M of redeemable noncontrolling interests in the mezzanine plus $115M of ordinary noncontrolling interests, and a minority claim belongs in enterprise value. Corrected EV $138.360B, 3.9% higher; EV/sales 8.26x → 8.58x; EV/EBITDA 16.20x → 16.82x.

9. returnOnEquityTTM is null in ratios_ttm — the km_ttm figure of 15.47% is used with a caveat. As with two other names in this batch, the two blocks disagree; here one is simply absent. The figure is in any case distorted by $38,307M of treasury stock and $52,746M of goodwill. returnOnInvestedCapitalTTM of 10.2% is quoted alongside and is also flattered by the intangible-heavy capital base.

10. quote.yearHigh/yearLow disagree with tech.hi52/lo52, and unusually the TECH figures are wider. Quote: $547.82 / $361.03. Tech: $564.15 / $373.75. The tech high is 3.0% ABOVE the quote high, which is the reverse of the usual intraday-versus-close pattern seen elsewhere in this batch and suggests the two series are adjusted differently for the 1 July distribution. We use tech for internal consistency with the moving averages and returns, and flag that no return figure in this file can be confidently described as spin-adjusted or not.

Not defects, correctly reported and independently confirmed: inc_q Q2 2026 revenue, operating income, pre-tax income and net income all match the 10-Q income statement; seg_geo sums to reported revenue exactly in every year with the 60.8% United States home market present, and the 10-Q's quarterly geographic table also reconciles exactly; bal_a FY2025 minorityInterest of $4,917M matches the balance sheet's redeemable noncontrolling interests line to the dollar; capexToRevenueTTM of 0.97% reconciles to $195M on $15,336M; interestCoverageRatioTTM of 23.6x is consistent with the debt and income figures; profile.currency and inc_a[0].reportedCurrency are both USD — no currency mixing; and the earn_cal EPS basis is consistent across all reported rows, so the beat series (+7.0%, −0.7%, +2.6%, +0.4%) is like-for-like and shows uniformly small beats.

8-K scan for corporate actions — completed, and it found the largest event in this dive. The three retained 8-Ks are 2026-07-02 (Item 1.01 — completion of the Mobility Global separation, the document quoted at length in Section 1), 2026-07-06 (a forward-looking-statements and risk-factor update devoted substantially to the separation) and 2026-07-28 (Item 2.02, second-quarter results, cover only). The separation is a filed, dated, fully documented corporate action that no vendor field reflects, and it is the reason for five of the ten findings above.

Non-equity tripwire — checked and passed. SPGI is common stock, $1 par value, NYSE-listed (10-Q balance sheet: 600 million shares authorised, 415 million issued). The $412.53 price is not par-like; beta is 1.079; the dividend is variable and has risen; volume was 1.83M shares (~$754M of turnover); and the 52-week band of $373.75 to $564.15 is a 51% range. This is common equity.

7. Technicals

• Price $412.53, down 0.89% on a broadly higher day. −26.9% from the 52-week high of $564.15; +10.4% above the low of $373.75. Position within the annual range: 20.3rd percentile.
• Moving averages, both overhead: −0.5% below a 50-day of $414.42; −9.1% below a 200-day of $453.90, which is falling. The stock is pinned at its 50-day with the 200-day 10% above.
• RSI 34.9 — the lowest reading in this batch, approaching but not at the conventional oversold threshold. MACD −0.55 — marginally negative.
• Maximum drawdown from the six-year peak: −26.9%, identical to the distance from the 52-week high, meaning the 52-week high is also the multi-year peak.
• Relative performance, uniformly negative: 3-month −2.9% vs SPY +7.6%; 6-month −21.8% vs SPY +11.1%; 12-month −24.4% vs SPY +24.3% and QQQ +30.8%. A 48.7-point underperformance against the S&P over twelve months and a 32.9-point gap over six.
• Sentiment, and it is one-sided: 24 buy, 4 hold, ZERO sell across 28 analysts. Consensus $541.22 (+31.2%), median $546, high $625 (+51.5%), low $498 — still 20.7% above spot. Basis unverified, per Section 6 finding 5.

What the technical picture does to the entry. On the surface this is the most attractive setup in the batch: the most oversold reading, the largest underperformance, unanimous analyst support, and a base fair value 13.9% above the price. The problem is that a portion of the decline is the distribution of a separate security rather than a de-rating, and the file does not let us determine how much. Until the 2026-10-29 restatement, "down 24% over twelve months" is a number of unknown composition, and building a position on it would be building on exactly the kind of measurement this dive has spent a section discrediting.

8. Insiders

The insider block for S&P Global contains no transactions that support inference in the records available. This is reported as an absence of data rather than as a neutral signal.

The company's own buying is the substitute and it is very large: $5,001 million of repurchases in fiscal 2025 — 4.1% of the current market capitalisation — plus a further $1,933 million in the first half of 2026, taking treasury stock at cost from $36,374M to $38,307M. Diluted shares have fallen from 318.9 million to 295.5 million since fiscal 2023, a 7.3% reduction. At the current price that programme is retiring roughly 3% of the company a year, and it is the single largest contributor to the consensus EPS growth path.

9. Verdict, kill-criteria and flip conditions

Watch.

What is genuinely excellent: two of the best operating segments disclosed by any company in the index — Ratings at a 63.8% operating margin and Indices at 68.7%, together 65% of fiscal 2025 segment operating profit on 43% of revenue; a June quarter in which every segment grew and total revenue rose 10.4%, with index asset-linked fees up 21.7%; capital expenditure of 1.3% of revenue and free cash flow of $5,456M at a 35.6% margin; a $5,001 million annual buyback retiring roughly 3% of the company a year; the separation of the group's worst-margin segment, lifting the remaining margin from 43.0% to 45.8%; interest coverage of 23.6x; and a price 26.9% below its high with the most oversold technical reading in this batch and all 28 analyst targets above it.

What holds us back, and it is not the business: five separate basis defects that make the current numbers unusable — two revenue bases for one quarter, an estimate compared against the wrong basis producing a false 10.1% miss, a consensus revenue line below last year's actual, a missing $2,299 million segment, and a price target of unverifiable vintage; an estimate block with nine analysts on the current year and one on each of the last two; a balance sheet 82.7% goodwill and intangibles with tangible asset value of minus $15,403 million; $5,024 million of noncontrolling interests the vendor's enterprise value ignores; a Ratings segment whose transaction revenue grew 25.0% off a strong issuance year and will not always; and a Market Intelligence segment at 36% of post-separation revenue and a 20.2% margin that is exactly the target of the knowledge base's only recent claim.

The distinction that matters. We think this is a good business at a reasonable price and we are declining to act on that because we cannot currently measure it. That is a different objection from every other Watch in this batch. There is a specific, dated event — the 2026-10-29 third-quarter report, which restates every historical period to continuing operations and incorporates the announced product transfers — after which the five defects in Section 6 resolve and the company becomes analysable. Acting eighty-six days early, on figures we have just spent a full section demonstrating are internally contradictory, would be the opposite of the discipline this process exists to enforce.

Pre-registered KILL criteria — what would take this to Avoid:

• Market Intelligence revenue growth below 4% or its segment margin below 18%, which would confirm the AI-disruption claim against the largest revenue segment.
• Ratings transaction revenue turning negative as investment-grade issuance normalises from the 25.0% growth just printed.
• Any impairment against $52,746 million of goodwill and intangibles, 82.7% of assets.
• The 2026-10-29 restated post-separation earnings base coming in materially below the $20.388 implied by the 2027 consensus, which would mean the estimate block never adjusted for the spin.
• The buyback falling below $3 billion a year, which would remove roughly a third of the consensus EPS growth path.
• Index asset-linked fee growth falling below 10%, which would break the one structurally compounding line.

Pre-registered FLIP conditions — what would take this to Buy — Tactical:

• The 2026-10-29 restatement showing a clean post-separation operating margin above 46% with revenue growth above 6%, at a price still below roughly $450. This is by far the most likely path and it is eighty-six days away.
• Price below approximately $370 (18.1x FY2027E) with no change in the fundamental picture, which would put the base case at +27% against a bear case only 10% lower.
• Market Intelligence growth accelerating above 8%, which would directly refute the only recent bearish claim in the knowledge base.
• Sell-side targets being restruck on a post-separation basis and still landing above $500, which would convert an unverifiable data point into a usable one.
• Genuine fresh knowledge-base conviction appearing — the three bullish claims here are 30 to 58 months old, and the only 2026 claim is bearish.

Where SPGI fits in the Synthos Framework Portfolio. The financial-data and exchange sleeve, at 0% today with a 2% target on the 2026-10-29 restatement or a fill near $370. Batch overlap note: there is no direct comparable in this batch — the vendor's peer set offers Moody's, which is not here — so the position competes for the quality-compounder allocation with SPGI's own closest analogue in spirit, which in this batch is nothing; the nearest is Fortinet on the "high-quality franchise, contested multiple" axis and they share no driver. Logged as a tracked Synthos call (Watch) as of 2026-08-04 at $412.53.

Single biggest risk: Market Intelligence. It is 36.2% of post-separation revenue and only 15.9% of post-separation segment operating profit, earning a 20.2% margin against 63.8% at Ratings and 68.7% at Indices. It grew 6.0% in the June quarter, the second-slowest of the five segments. And it is a subscription data-and-analytics business built substantially from the IHS Markit merger, whose model is annual price increases on entrenched enterprise contracts — which is exactly what the knowledge base's only 2026 claim, at conviction 75, says is being directly disrupted, citing "brutal drawdowns" at Thomson Reuters and Verisk as evidence. If that claim is right, the segment does not merely decelerate; its margin, already the lowest in the group, compresses further, and $52,746 million of goodwill and intangibles concentrated in exactly that acquisition comes under pressure. The protection is that 65% of the profit comes from two franchises that regulation and benchmark network effects make close to unassailable. The risk is that the market keeps pricing the whole company on the vulnerable third.

Provenance & disclosures

• Traceability: 5 raw knowledge-base entity hits name S&P Global, of which two are the same claim from the same speaker on the same date — 4 distinct claims (entity matches 5, distinct 4, discarded for collision ~11; breadth 3, net conviction positive-but-stale-and-turning). Three are bullish and old: Dev Kantesaria (we_study_billionaires, 2021-10-22, conviction 80, named independent speaker, disclosing a long-held position, on the ratings duopoly rating 90% of world debt); an unattributed business_breakdowns claim (2023-05-29, conviction 85, on the ratings standard's entrenchment via a 0-50 basis point interest penalty for alternatives); and Brian Yakman (business_breakdowns, 2024-02-23, conviction 65, named independent, on the same network effect with a caveat about culture). The fourth is the only claim written in 2026 and it is BEARISH — compound_and_friends, 2026-02-03, conviction 75, that legal and data-services incumbents which endlessly raised prices are being directly disrupted by AI tools, citing "brutal drawdowns" at Thomson Reuters and Verisk. An AI vendor name inside that claim is redacted per house rules and the redaction is disclosed. A supplementary text sweep on S&P Global, SPGI, credit rating agency and Moody's returned 18 hits, of which roughly 11 are discarded as homographs — and one collision is new to the catalogue: "S&P Global" meaning the PMI purchasing-managers survey rather than the equity (3 claims), plus "Moody's" meaning the US sovereign credit downgrade (3 claims) or a source of private-credit default statistics (3 claims), plus two analogical uses ("Moody's for crypto", "ratings agency" in an exchange-listing sense). Three further claims from Brian Yakman are retained as competitor context — two bullish on Moody's, and one stating he prefers MSCI's index business to S&P Global's, which is a mild negative for this name and is included for that reason. No claim in the lane carries a management speaker_role, so the half-weight management discount does not apply. No concentration test was required: no channel supplies more than two of the four distinct claims. All quotes are verbatim apart from the disclosed redaction.
• Data as-of: most recent filing-verified financials 2026-06-30, from the 10-Q filed 2026-07-28 (income statement, balance sheet, full segment revenue disaggregation, geographic table, segment assets, and the separation accounting disclosure) · annual segment and financial detail for fiscal 2025 ended 2025-12-31 from the 10-K filed 2026-02-11, used in preference to the vendor's defective seg_prod block · estimates 2026-08-04 · prices 2026-08-04, quote timestamp 1785873602 = 2026-08-04T20:00:02Z ($412.53, −0.89%; 50-DMA $414.42; 200-DMA $453.90; RSI 34.9; MACD −0.55) · knowledge-base claims 2026-08-04. Fiscal note: S&P Global's fiscal year is the calendar year. CRITICAL BASIS NOTE: the Mobility Global separation became effective 1 July 2026 — one day after the June quarter close and thirty-four days before this dive — and the company has NOT yet restated its history. Mobility is INCLUDED in all reported figures through 30 June 2026 and becomes discontinued operations for all periods beginning with the third quarter. All figures come from the Synthos vendor data file for SPGI or from the SEC filings in the SPGI archive; no figure comes from memory, recall or external retrieval.
• Filing archive contents: 10-K filed 2026-02-11 (fiscal 2025, containing the authoritative five-segment revenue and operating-profit tables used throughout); 10-Q filed 2026-04-28 (March 2026 quarter); 10-Q filed 2026-07-28 (June 2026 quarter, reviewed by Ernst & Young); 8-K filed 2026-07-02 (Item 1.01 — completion of the Mobility Global separation, quoted at length in Section 1, the single most consequential document in this dive); 8-K filed 2026-07-06 (forward-looking-statement and risk-factor update devoted substantially to the separation); 8-K filed 2026-07-28 (Item 2.02, second-quarter results — cover document only, Exhibits 99.1 and 99.2 not extracted).
• Where the filings contradicted or corrected the vendor (detailed in Section 6): seg_prod for FY2025 OMITS the entire Energy segment — $2,299M of gross revenue, leaving four lines summing $2,099M short of reported revenue — and the 10-K's segment table is used throughout instead; earn_cal.revenueActual of $3,678M identified as CONTINUING OPERATIONS by reconciling it against the 10-Q's disclosed Mobility segment revenue of $468M ($4,146M − $468M = $3,678M exactly), while inc_q's $4,146M is the total and matches the income statement — the two blocks are on different bases and no comparison between them is made; earn_cal compares that continuing-operations actual against a $4,093M total-basis estimate, manufacturing a 10.1% revenue "miss" that did not occur, with the estimate basis then switching to continuing operations in the next row; the entire est.revenueAvg series rejected because the FY2026 figure of $14,906M sits below the FY2025 actual of $15,336M purely on basis; est.netIncomeAvg rejected for exceeding est.ebitAvg in FY2029 and FY2030; est.ebitdaAvg and est.ebitAvg rejected as fixed-ratio fabrications at exactly 49.33% and 42.07% of revenue in every year; km_ttm.enterpriseValueTTM corrected from $133.221B to $138.360B by adding the $5,024M of redeemable noncontrolling interests and $115M of ordinary noncontrolling interests that the balance sheet carries and the vendor omits; pt.targetConsensus of $541.22 flagged as of unverifiable vintage and explicitly not used as evidence of upside; returnOnEquityTTM noted as null in ratios_ttm against 15.47% in km_ttm; quote.yearHigh/yearLow rejected in favour of tech.hi52/lo52, with the unusual observation that the tech high is 3.0% ABOVE the quote high, the reverse of the normal pattern, suggesting the two series are adjusted differently for the 1 July distribution. Where vendor and filing AGREED — worth recording: inc_q June-quarter revenue, operating income, pre-tax income and net income all match the 10-Q; seg_geo sums to reported revenue exactly in every year with the 60.8% United States home market present, and the 10-Q's quarterly geographic table reconciles exactly; bal_a minorityInterest of $4,917M matches the balance sheet's redeemable noncontrolling interests line to the dollar; the earn_cal EPS series is consistently adjusted and like-for-like.
• Basis note: analyst estimates and earn_cal EPS actuals are adjusted (non-GAAP); reported diluted EPS is GAAP. Trailing adjusted EPS is $18.56 against trailing GAAP diluted of $16.42, a 13.0% gap driven principally by acquisition-related intangible amortisation from the IHS Markit merger. The EPS basis is consistent across all earn_cal rows, so the beat series (+7.0%, −0.7%, +2.6%, +0.4%) is like-for-like and shows uniformly small deviations — a well-guided company. Forward multiples use consensus adjusted EPS; the trailing multiple is shown on both bases (22.2x adjusted, 25.1x GAAP diluted). The revenue basis is NOT consistent and is documented at length above.
• Estimate coverage: only 9 analysts on FY2026 EPS and 10 on FY2026 revenue, against 28 analysts in the grades block — the est block is capturing barely a third of the coverage. 13 on FY2027 EPS (the year the fair values use), 5 on FY2028, and ONE each on FY2029 and FY2030, both of which are excluded outright. This is thin coverage for a $122 billion company and the thinness is disclosed rather than buried.
• Peer note: the vendor peer set contains Moody's (MCO) — the direct comparable and the other half of the ratings duopoly — alongside Bank of America, BlackRock, Capital One, Interactive Brokers, Santander and Toronto-Dominion, six of which are banks, brokers or asset managers and are not comparable. The set has been matched on the "Financial Services" sector code. No peer multiples are supplied in the file, so no peer-multiple comparison is drawn.
• Fair-value caveat: the $335 / $470 / $600 anchors are multiples of the FY2027 (calendar 2027) consensus adjusted-EPS distribution — 16.9x the low of $19.863, 23.1x the mean of $20.388, and 26.8x the high of $22.355 — deliberately anchored on FY2027 rather than FY2026 because FY2026 blends pre- and post-separation reporting, and cross-checked against FY2028E of $23.229. Stated arithmetic, not a discounted cash flow. The base is sensitivity-disclosed: 21x gives $428, 25x gives $510. The base assumes modest multiple expansion and Section 4b flags that as its fragile leg — but also notes that a HELD multiple of 20.2x on FY2028E gives $469, which is essentially the base case, so the re-rating is not actually required.
• Timing: second-quarter results were released 2026-07-28, seven days before this dive, and the Mobility Global separation became effective 1 July 2026, thirty-four days before. The next print is 2026-10-29, 86 days away, and is unusually important: it is the first report in which Mobility appears as discontinued operations for ALL periods, restating the entire historical series and resolving five of the ten data-integrity findings above. The verdict is set with that date as an explicit input, and the dive states plainly that acting before the restatement would mean acting on figures it has just demonstrated to be internally contradictory.
• Accessibility note: no information in this dive is conveyed by colour. All emphasis is carried by bold text, table structure and explicit labelling.
• Not investment advice. Independent research, educational and informational only, never personalised. No recommendation to buy, sell or hold any security is made to any person.
• Version: 2026-08-04-full.