Intercontinental Exchange ICE

Reference table

1. The MarketAxess acquisition, and everything the file does not say about it

From the 8-K filed 2026-07-30, and this is the whole of what the archive contains:

> "On July 29, 2026, Intercontinental Exchange, Inc.... entered into an Agreement and Plan of Merger... with MarketAxess Holdings Inc.... and Igloo Merger Sub II, Inc.... Merger Sub will be merged with and into MarketAxess, with MarketAxess surviving as a wholly owned subsidiary of ICE."

> "...each share of common stock, par value $0.003 per share, of MarketAxess... will be converted into the right to receive $167.00 in cash per share, without interest."

> "ICE expects to finance the payment of the aggregate Merger Consideration... through a combination of available cash and incremental debt financing. However, there can be no assurance that such incremental debt financing will be completed... On July 29, 2026, ICE, Bank of America, N.A. and BofA Securities, Inc. entered into a financing commitment letter... for a 364-day senior unsecured bridge facility in an aggregate principal amount not to exceed $6.25 billion... The receipt or availability of any financing by ICE, including the Bridge Facility, is not a condition to ICE's obligation to consummate the Merger."

What that establishes. An all-cash acquisition at $167.00 per share, with no financing condition — ICE is obliged to close whether or not the debt markets cooperate — backstopped by a bridge facility of up to $6.25 billion. MarketAxess options and time-based restricted stock units convert into ICE equivalents. The 10-Q's XBRL tagging carries a ScenarioForecast period of 2027-04-01 to 2027-06-30 for MarketAxess Holdings, which implies management expects to close around the second quarter of 2027 — roughly three quarters away.

What it does not establish, and this is the reason for the verdict. The archive contains no aggregate purchase price, no expected synergies, no accretion or dilution estimate, no pro-forma leverage guidance, and no MarketAxess financial statements. The earnings-release exhibits for this name are not in the extraction. The 8-K is checked as "Soliciting material pursuant to Rule 14a-12", which means a proxy process is under way that this archive does not reach.

What can be inferred, labelled as inference. A $6.25 billion bridge sized to backstop the full consideration implies an aggregate price in that neighbourhood — the facility exists "to fund, in part, the acquisition and to pay related fees, commissions and expenses" if permanent financing is unavailable. On ICE's trailing EBITDA of approximately $6.98 billion, adding $6.25 billion of debt to $20,279 million existing takes gross leverage from roughly 2.9x to roughly 3.8x, and net leverage from 2.79x toward roughly 3.6x, before any MarketAxess EBITDA. That is our arithmetic on the vendor's balance sheet and the filing's facility size; it is not the company's guidance, and the company has published none this file can reach.

Two things a reader should hold alongside it. Jeffrey C. Sprecher has built this company by serial acquisition since 2000 — NYSE, Interactive Data, Ellie Mae, Black Knight — and that record argues for more trust in the decision than a stranger's would deserve. And the balance sheet shows what it has cost: $30,646 million of goodwill and $15,353 million of intangibles against $28,915 million of equity, so tangible equity is negative by approximately $17 billion.

The other 8-K in the record is routine: an Item 5.07 annual-meeting report (2026-05-19, meeting held 2026-05-15) recording eleven directors elected against 566,430,761 outstanding voting shares at the March record date.

2. The June quarter — and the revenue gross-up the payload commits

From the 10-Q filed 2026-07-30 ($M). Note the two revenue lines: the company's headline measure is total revenues LESS transaction-based expenses.

Four observations.

First, the headline growth rate decelerated sharply and the year-to-date number conceals it. The first half grew net revenue 12.5%. The March quarter grew 20.6%. The June quarter grew 4.8%. That is the sharpest sequential deceleration in this batch, and it is entirely invisible in the half-year figure most commentary will quote.

Second, cost control was excellent and is the reason operating income still grew. Total operating expenses rose 2.3% against net revenue up 4.8%; compensation rose 2.0%; depreciation and amortisation actually fell 1.5%, which is the Black Knight intangible amortisation beginning to roll off.

Third, the first half contains a $452 million gain on investments that appears in no vendor field. "Other income, net" went from $24 million to $485 million across the half, and the cash-flow statement's reconciling line reads "Gain on investments (452)". Almost all of it landed in the March quarter — Q2 other income was only $74 million — which is why the vendor's inc_q shows March-quarter net income of $1,413 million, up 77.3%. That figure is not a run rate and the payload gives no indication of it.

Fourth, and this is the confirmed defect: the vendor reports revenue on the GROSS basis while consensus is struck on the NET basis. inc_q.revenue for the June quarter reads $3,611M — the total-revenues line — while the company's headline and the analyst estimates use $2,666M. For FY2025, inc_a.revenue reads $12,638M against est.revenueAvg of $9,896.5M, a 27.7% gross-up. This is exactly the documented tell — "est.revenueAvg matches the published figure while inc_a.revenue does not" — that flagged USB at 50.2%, BNY at 101%, IBKR at 70%, COF and SCHW, and ICE was named as a candidate. Confirmed. Every revenue and margin figure in this dive uses the net basis, and the corroboration is that seg_geo ties to the net basis: FY2025 United States $6,316M plus United Kingdom/Continental Europe/Canada $3,615M equals $9,931M against est.revenueAvg of $9,896.5M — a 0.35% match — and against inc_a.revenue of $12,638M it would be 78.6%.

One further vendor error worth naming: inc_q.operatingIncome for the June quarter reads $1,987M against the filing's $1,391M — a 42.8% overstatement — while incomeBeforeTax ($1,287M), netIncome ($958M) and eps ($1.69) all match the filing exactly. The March-quarter figure of $1,665M is correct (the filing's H1 $3,056M less Q2 $1,391M). So one field in one quarter is wrong by 43% while its neighbours are right — the same partial-row corruption pattern documented on Robinhood in this batch.

3. Cash flow, the balance sheet and capital return

The capital-expenditure check: reconciles internally, with a two-line structure

ICE capitalises investment across two cash-flow lines, both visible in the 10-Q: "Capital expenditures" ($208M H1 2026, $145M H1 2025) and "Capitalized software development costs" ($230M H1 2026, $211M H1 2025) — $438M and $356M respectively.

The vendor's FY2025 investmentsInPropertyPlantAndEquipment of −$580M appears to be the SUM of both lines rather than one of them, and the trailing arithmetic confirms it: FY2025 $580M less H1 2025 $356M plus H1 2026 $438M = $662M, against a trailing figure of approximately $666M implied by capexToDepreciationTTM of 0.4300 against trailing depreciation and amortisation of $1,549M. A 0.6% reconciliation.

This is the opposite of the Robinhood case in this batch, where the same two-line structure produced a 72% understatement because only one line was captured. Here both are captured and the field checks out. Recorded as a clean verification, with the caveat that the FY2025 10-K cash-flow statement itself was not directly located in the extraction and the confirmation is arithmetic rather than line-matched.

A capex-to-depreciation ratio of 0.43 looks low against most industrials and is correct here: ICE's $1,549 million of trailing depreciation and amortisation is dominated by intangible amortisation from Black Knight, Ellie Mae and NYSE rather than by physical assets, and an exchange operator's true maintenance capital intensity is genuinely small. The ratio is explained rather than flagged.

Balance sheet and capital return

Total assets swing violently — $174,247 million at June 2026 against $136,887 million at December — because clearing-house margin deposits sit on the balance sheet and are not ICE's money. financialLeverageRatioTTM of 5.90x and returnOnAssetsTTM of 2.31% are computed against that gross figure and are not meaningful measures of this business. returnOnInvestedCapitalTTM of 8.42% is depressed by the same effect plus $46 billion of goodwill and intangibles; the operating margin of 44.9% and the free-cash-flow yield of 5.79% are the measures used here.

Capital return, and the acquisition puts part of it at risk:

The buyback restarted in FY2025 after two years of near-dormancy and accelerated in the first half of 2026 to $1,202 million — an annualised 2.9% of the market capitalisation. Combined with the 1.34% dividend that is roughly 4.3% of shareholder yield, and it is precisely the line an all-cash $6.25 billion acquisition is most likely to interrupt. Treasury stock rose from $7,792 million to $9,100 million across the half and the share count fell from 567 million to 561 million, −1.1%.

4. Valuation — priced in or room?

At $149.20 (market cap $83.760B, 561M shares):

Estimate coverage is thin on earnings and thinner beyond. 7 analysts on FY2026 EPS, 9 on FY2027 — the anchor — and only 4 on FY2028. FY2029 and FY2030 rest on TWO analysts each and are excluded from every conclusion. The FY2027 range of $8.276 to $9.264 is an 11.9% spread. Against 36 analysts issuing ratings, EPS coverage of 7-9 is notably thin and is stated as such.

est.ebitdaAvg and est.ebitAvg are REJECTED with the signature stated. In every year from FY2024 to FY2030, ebitdaAvg is exactly 53.50% of revenueAvg and ebitAvg is exactly 41.62% — except FY2023, where ebitdaAvg of $6,944.8M is 85.9% of revenueAvg, an obvious break. ICE's actual trailing EBITDA margin on the net revenue basis is 52.0%, so the fabricated ratio is close in level and is still a fixed ratio applied mechanically across seven years with one impossible outlier. All forward valuation uses epsAvg, which is verified: FY2025's $6.916 against a reported GAAP diluted $5.77 reflects a consistent adjusted-basis wedge (the FY2024 pair is $6.047 against $4.78), so the field is sound and the basis stable.

Enterprise value is broadly reliable and the NCI omission is immaterial. enterpriseValueTTM of $103,220M against a market capitalisation of $83,760M implies $19,460M of net claims, against FY2025 net debt of $19,442M — an $18 million residual. Noncontrolling interests of $54 million (0.05% of enterprise value) appear omitted, which is the KKR / MPC / FCX class in a form small enough to be irrelevant. evToEBITDATTM of 14.8x is quoted without correction.

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

At $149.20 — 18.5x FY2026 consensus and 17.0x FY2027 — the price embeds:

• Net revenue grows 11.2% in FY2026 and 5.8% in FY2027. (Consensus; 10 analysts each, on the revenues-less-transaction-based-expenses basis.) The first half delivered 12.5% — but the June quarter delivered 4.8%. The FY2026 consensus requires the second half to hold the first half's pace from a quarter that ran at less than half of it.
• EPS reaches $8.07 in FY2026 and $8.78 in FY2027. (Consensus; 7 and 9 analysts.) The FY2026 figure is flattered by a $452 million first-half gain on investments that the consensus may or may not exclude — the payload gives no adjusted-versus-GAAP breakdown and the earnings releases are absent. Stripping it at a 23% tax rate is roughly $0.62 per share, 7.7% of the FY2026 estimate.
• The MarketAxess acquisition is at worst neutral. (Our derivation; the company has published nothing this archive can reach.) This is the most fragile assumption in the price, not because the deal is necessarily bad but because it is entirely unquantified here: no aggregate price, no synergies, no accretion, no pro-forma leverage, no target financials. A $6.25 billion cash commitment with no financing condition, on a company with $20.3 billion of debt at 2.79x EBITDA, is a material change to the capital structure being underwritten on trust.
• The buyback continues near $2.4 billion a year. (Our number, annualising the $1,202M first half.) At 2.9% of market capitalisation it is roughly a third of the expected return. It is the most likely casualty of the acquisition financing.
• The market keeps paying 15-20x forward earnings. (Our number.) At 14x FY2027E the stock is $123 — essentially the 52-week low; at 21x it is $184, roughly the street's consensus target. Every point of multiple is $8.78 per share, 5.9% of the price.

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

Expected return over the next twelve months decomposes as: EPS growth (+8.9%, FY2026E $8.066 to FY2027E $8.780) + multiple drift (EXPANSION, from 18.5x on today's forward year to roughly 19.0x on the then-forward year) + shareholder yield (+4.3%, at risk from the acquisition) ≈ +13%.

This is the only name in this batch whose base case requires multiple EXPANSION, and the reason is where the multiple already sits. At 17.0x FY2027 ICE trades below the range a regulated exchange with a 52% EBITDA margin and a 5.79% free-cash-flow yield has occupied for most of the last decade, and below the 21-24x the shares themselves commanded twelve months ago at $188.45. Our base of $165 is 18.8x FY2027E — still below where this stock traded a year ago.

Note what that makes the return: modest earnings growth, a large and threatened shareholder yield, and a re-rating that has to happen for the base case to work. A thesis that requires the crowd to change its mind is a weaker thesis than one that does not, and this one does — which is a second reason, alongside the acquisition, that the verdict is Watch.

At 14x FY2027E the price is $123 (−17.6%). At 21x it is $184 (+23.5%).

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

• We think the June-quarter deceleration is the most under-discussed number in this file. Net revenue grew 4.8% against 20.6% in March. The half-year figure of 12.5% will be quoted everywhere and it averages the two. Watchable number: net revenue growth in the third-quarter release. Above 8% and the June quarter was a comparison effect; at or below 5% and the FY2026 consensus of 11.2% needs cutting by several points.
• We think the $452 million first-half gain on investments is not in most models as a one-off. It is 18.8% of first-half net income and appears in no vendor field; the payload's March-quarter net income of $1,413 million (+77.3%) is the visible consequence and is presented with no explanation. Watchable line: "Other income, net" in the third quarter, against $74 million in June and $5 million a year earlier.
• We agree with the street on direction and disagree on how much the acquisition matters. Consensus target $183.25, with 33 of 36 analysts positive, ZERO sells, and a LOW target of $163 that is 9.2% ABOVE the market price. Our base of $165 is 10.0% below their consensus and one dollar above their most bearish analyst. As on Capital One, Williams, Robinhood and Moody's in this programme, we treat a covering community with no downside case as a caution rather than a confirmation — and here the caution has a specific object: a $6.25 billion cash deal signed six days ago with no published economics.
• The knowledge base's only recent claim is structurally negative and is tagged bullish, and we grade it on content. arthur_hayes, 2026-03-05, conviction 85, skill 1.2, entities Hyperliquid / CME / ICE: "Perps' edge is clearing, not product spec: loser-can-only-lose-margin + ADL lets exchanges offer real leverage; guaranteed-settlement TradFi clearing can't compete." The stance field says bullish. The content says the settlement guarantee that defines ICE's and CME's clearing houses is a competitive disadvantage against auto-deleveraging perpetuals venues. This is the confirmed stance-inversion defect — previously documented on CME and ABNB — appearing on a second exchange, in the same sector, on the same theme. Graded on content: bearish, conviction 85. It is one voice over an unspecified horizon and does not make the name-level conviction negative, but it is the opposite of support.
• Positive surprise that would force a re-rating: disclosed MarketAxess economics showing accretion in the first full year with pro-forma net leverage guided below 3.25x; net revenue growth above 8% in the third quarter; the buyback continuing at $600 million a quarter through the acquisition financing; or Mortgage Technology reaccelerating above 8% as US origination volumes recover.
• Negative surprise that would break the thesis: the buyback being suspended to fund the acquisition; pro-forma net leverage above 3.75x; a second quarter of net revenue growth below 5%; the MarketAxess deal being repriced upward or facing a regulatory challenge; or a further large acquisition before this one closes.

Synthos fair values

All three anchors are multiples of the FY2027 consensus EPS distribution (mean $8.780, low $8.276, high $9.264; 9 analysts), cross-checked against the free-cash-flow yield.

• Bear ~$116 — 14.0x the FY2027 consensus LOW of $8.276. Cross-check: 14.4x FY2026E; a 7.4% free-cash-flow yield; 5.6% below the 52-week low of $122.91. The scenario: net revenue growth stays near the June quarter's 4.8%, the MarketAxess deal proves expensive and dilutive, the buyback is suspended to fund it, leverage passes 3.75x, and an exchange with $46 billion of goodwill and intangibles on $29 billion of equity re-rates to a levered-acquirer multiple. −22.3%.
• Base ~$165 — 18.8x the FY2027 consensus MEAN of $8.780. Cross-check: 20.5x FY2026E; 16.8x FY2028E; a 5.2% free-cash-flow yield; 12.4% below the 52-week high of $188.45. Sensitivity, stated openly: 17x gives $149 — exactly today's price — and 21x gives $184, the street's consensus. The whole answer lives in a 17-21x band on one estimate row supported by nine analysts. The scenario: the deceleration proves temporary, MarketAxess closes on terms that are at worst neutral, the buyback survives, and the multiple recovers part of the 21% de-rating. +10.6%.
• Bull ~$222 — 24.0x the FY2027 consensus HIGH of $9.264. Cross-check: 22.6x FY2028E; 17.8% above the 52-week high. The scenario: MarketAxess proves accretive and strategically coherent in fixed-income electronification, net revenue growth returns above 8%, Mortgage Technology recovers with origination volumes, and the market restores an exchange multiple to a company with a 52% EBITDA margin. +48.8%.

Base is 10.6% above spot; asymmetry roughly 2.19:1 (22.3% down, 48.8% up), before a shareholder yield of approximately 4.3%. That clears the 2:1 bar and the verdict is still Watch. The reasons are set out in Section 8 and they are qualitative rather than arithmetic — which is itself the honest description of this call.

5. Knowledge base — two claims, and the recent one is inverted

Raw hits: 2. After a case-sensitive entity re-run: 2. Name-level claims on ICE: 2. Discarded as a name-level input: 1.

The case-sensitive entity sweep ran ICE and Intercontinental Exchange across all 52,021 distilled claims and returned two hits. Worth recording separately: a three-letter ticker that is also an ordinary English word produced NO homograph contamination — no frozen water, no immigration enforcement, no "ICE age". That is a materially better result than CEG, PWR, EMR, USB or MMM produced and it means both hits can be taken at face value.

Both claims, verbatim:

> 2022-08-31 · bullish · conviction 80 · horizon: thesis · channel: real_vision · no named speaker · entities: tZERO, Intercontinental Exchange, Overstock

> "tZERO is the dominant digital-securities secondary venue (~70% of volume); ICE/NYSE backing positions it as the NYSE of private markets."

> 2026-03-05 · stance tagged BULLISH · conviction 85 · SKILL 1.2 · horizon: principle · channel: arthur_hayes · no named speaker · entities: Hyperliquid, CME, ICE

> "Perps' edge is clearing, not product spec: loser-can-only-lose-margin + ADL lets exchanges offer real leverage; guaranteed-settlement TradFi clearing can't compete."

The first is discarded as a name-level input. It is four years old, its subject is tZERO rather than ICE, and ICE appears only as a backer. It says nothing about the company's economics.

The second requires care and is the reason this section exists.

The stance field reads BULLISH. The content is structurally BEARISH for ICE. The claim's subject is Hyperliquid and the perpetual-futures clearing model; its argument is that "loser-can-only-lose-margin + ADL" — auto-deleveraging, in which the clearing venue socialises losses across profitable positions rather than guaranteeing settlement — "lets exchanges offer real leverage"; and its explicit conclusion is that "guaranteed-settlement TradFi clearing can't compete." ICE and CME are named as the TradFi clearing houses that cannot compete.

This is the confirmed stance-inversion defect. The programme has previously documented stance tags inverted relative to the named entity on CME and on ABNB. Finding it again here — on the same sector, the same theme and a sibling exchange — suggests the inversion is systematic in this thematic neighbourhood rather than incidental. The claim is graded on content: BEARISH, conviction 85, skill 1.2 (a 20% weight premium), one voice, horizon "principle" — that is, no date attached.

How much weight it carries. Not enough to make the name-level conviction negative: it is a single claim about market structure over an unspecified horizon, from one channel, and it does not address ICE's revenue, earnings or valuation. Enough to be the opposite of support, and enough to be the most substantive long-run question in this file: ICE's regulatory moat is precisely the settlement guarantee this claim argues is now a handicap in the competition for leveraged speculative flow. The exponential score of 4 reflects it and the long-horizon watch list names it.

Attribution note: both claims carry speaker: null and are channel-attributed — the weakest sourcing the four-lane policy admits. No management voice appears. No concentration test is meaningful across two claims from two channels.

Conclusion. Breadth 1 (one usable channel), claim count 2, net conviction negative-low. The Synthos knowledge base has almost nothing to say about Intercontinental Exchange, and what it does say — once read on content rather than on its stance tag — is a structural argument against the durability of exchange clearing economics. Regulated exchange operation joins the list of thin sectors in this store.

6. Data integrity — the gross-up confirmed, plus a 43%-wrong operating income

Six findings.

1. The revenue GROSS-UP is CONFIRMED — inc_a and inc_q report a basis the company does not headline. inc_a.revenue for FY2025 reads $12,638M; est.revenueAvg for FY2025 reads $9,896.5M — a 27.7% gross-up. The 10-Q settles it: total revenues of $3,611M in the June quarter less transaction-based expenses of $945M (Section 31 fees $288M plus cash liquidity payments, routing and clearing $657M) gives total revenues less transaction-based expenses of $2,666M, which is the company's headline and the analysts' basis. This is precisely the documented tell — "est.revenueAvg matches the published figure while inc_a.revenue does not" — that flagged USB (50.2%), BNY (101%), IBKR (70%), COF and SCHW; ICE was named as a candidate and is confirmed. Consequential rejections: priceToSalesRatioTTM of 6.24x (correct on the net basis ≈ 8.4x), evToSalesTTM of 7.69x (correct ≈ 10.4x), netProfitMarginTTM of 30.0% (correct ≈ 38%) and operatingProfitMarginTTM of 44.9% — which, note, is ALREADY on the net basis, so the ratio block is internally inconsistent. Corroboration: seg_geo ties to the NET basis (FY2025 United States $6,316M plus United Kingdom/Continental Europe/Canada $3,615M = $9,931M against est.revenueAvg of $9,896.5M, a 0.35% match). All revenue and margin figures in this dive use the net basis.

2. seg_prod alternates between two different measures across years — REJECTED for FY2020, FY2021, FY2024 and FY2025. FY2022 and FY2023 tie exactly to inc_a revenue ($9,636M and $9,903M). But FY2025 reports Exchanges $1,031M, Fixed Income and Data Services $1,359M and Mortgage Technology $269M, summing to $2,659M — 21.0% of revenue — when the 10-Q shows June-quarter segment revenues of $2,409M, $645M and $557M, implying full-year figures roughly eight times, twice and eight times the vendor's respective numbers. FY2024 (21.3%), FY2021 (22.3%) and FY2020 (23.0%) show the same collapse. Whatever the low-year figures are — and $2,659M for FY2025 is suspiciously close to the implied transaction-based expense total — they are not segment revenue. A block that reports total segment revenue in two years and a ~22% sub-component in four is unusable; all segment figures in this dive come from the 10-Q.

3. inc_q.operatingIncome for the June 2026 quarter is 42.8% too high — REJECTED, filing substituted. The vendor reads $1,987M against the filing's $1,391M. The March-quarter figure of $1,665M is CORRECT (the filing's H1 $3,056M less Q2 $1,391M), and incomeBeforeTax ($1,287M), netIncome ($958M) and eps ($1.69) for the June quarter all match the filing exactly. One field, one quarter, wrong by 43%, surrounded by correct fields — the same partial-row corruption documented on Robinhood in this batch.

4. The MarketAxess acquisition and the $452 million gain on investments are both absent from every vendor field. The 2026-07-29 merger agreement at $167.00 per share in cash with a $6.25 billion bridge facility post-dates bal_a (which ends 2025-12-31) and cf_a (FY2025) and appears nowhere. And the first half of 2026 contains a $452 million gain on investments — visible in the 10-Q cash-flow reconciliation and in "Other income, net" rising from $24M to $485M — almost all of it in the March quarter, which is why the payload's March net income of $1,413M reads +77.3% with no explanation. This is the "whole transaction absent" class, twice.

5. est.ebitdaAvg and est.ebitAvg carry a fixed-ratio fabrication signature with one impossible outlier — REJECTED. From FY2024 to FY2030, ebitdaAvg is exactly 53.50% of revenueAvg and ebitAvg is exactly 41.62%. The FY2023 row breaks the pattern entirely at 85.9% and 51.7%, which is not a margin any exchange achieves. All forward valuation uses epsAvg, whose adjusted-basis wedge against GAAP is consistent (FY2025 $6.916 against $5.77; FY2024 $6.047 against $4.78).

6. returnOnAssetsTTM, financialLeverageRatioTTM and returnOnInvestedCapitalTTM are structurally misleading — NOT USED. Total assets of $174,247M at 2026-06-30 against $136,887M at December swing by tens of billions because clearing-house margin deposits sit on ICE's balance sheet and are not ICE's money. returnOnAssetsTTM of 2.31% and financialLeverageRatioTTM of 5.90x are computed against that gross figure. returnOnInvestedCapitalTTM of 8.42% is additionally depressed by $46.0 billion of goodwill and intangibles. The measures used here are the operating margin (44.9%), the EBITDA margin (52.0%) and the free-cash-flow yield (5.79%).

Clean verifications worth recording. The capital-expenditure check reconciles: ICE capitalises across two lines ("Capital expenditures" $208M and "Capitalized software development costs" $230M in H1 2026), the vendor's FY2025 −$580M appears to capture both, and the trailing rebuild ($580M − $356M + $438M = $662M) matches the $666M implied by capexToDepreciationTTM of 0.430 to 0.6% — the opposite of the Robinhood case in this batch, where only one of two lines was captured and the field understated by 72%. The share count is EXACT: the 10-Q shows 655 million issued less 94 million treasury = 561 million outstanding at 2026-06-30, against 561.4 million implied by market capitalisation ÷ price. enterpriseValueTTM of $103,220M implies $19,460M of net claims against FY2025 net debt of $19,442M, an $18 million residual, with the $54 million of noncontrolling interests omitted at 0.05% of enterprise value — immaterial. And seg_geo ties to the net revenue basis within 0.35%, with the home market PRESENT at United States $6,316M, 63.8% of net revenue — checked against the NEM / WM / MAR omission class and passed.

Two gaps. The earn_cal block is COMPLETELY ABSENT for this name — no actual-versus-estimate history and no forward date — so no beat-and-miss pattern could be assessed and the next reporting date is inferred from filing cadence. And the insider block returned no usable transactions.

Non-equity tripwire — checked and passed. ICE is common stock, $0.01 par value, NYSE-listed, per the 10-Q balance sheet (1,500 million shares authorised, 655 issued, 561 outstanding). Price of $149.20 is not par-like; beta is 0.948; the dividend is a declared and rising quarterly rate ($2.00 trailing, 1.34% yield); volume was 3.56M shares (~$531M of turnover); the 52-week band of $122.91 to $188.45 is a 53% range. Preferred stock: 100 shares authorised, NONE issued or outstanding. Noncontrolling interests are $69 million. This is common equity.

Vendor composite rating — noted, and not used. B / 3 overall, with 4 on return on equity, 4 on return on assets and 1 on price-to-book. The return-on-assets score is computed against a balance sheet inflated by clearing margin deposits. Not used.

7. Technicals

• Price $149.20. −20.83% from the 52-week high of $188.45; +21.39% above the low of $122.91. Position within the annual range: 40th percentile.
• Above the shorter average, below the longer: +6.4% above a 50-day of $140.21; 3.5% BELOW a 200-day of $154.65. The 50-day sits well below the 200-day — the shape of a stock that broke down, bottomed, and is now climbing back toward its falling long-term average, which is where such advances typically stall.
• RSI 64.0 — firm, approaching but not at overbought. MACD +4.03.
• Maximum drawdown from peak over the trailing year: −20.83%, identical to the distance from the high.
• Beta 0.948 — the second-lowest in this batch after Williams, and a reminder that this is infrastructure rather than a cyclical.
• Relative performance, and it is the worst record in this batch: 3-month −4.63% against SPY +7.59% — twelve points behind; 6-month −13.85% against SPY +11.09% — twenty-five points behind; 12-month −19.18% against SPY +24.26% and QQQ +30.80% — a FORTY-THREE-POINT and fifty-point deficit. No other name in this batch is close.

Today's move

ICE closed 2026-08-04 at $149.20, down 1.71% or $2.60 from $151.80. It opened at $150.04, traded $147.68 to $151.44, and closed near the low of a narrow 2.5% range on 3.56 million shares. No company-specific filing is dated 2026-08-04; the last events were the 2026-07-29 merger agreement and the 2026-07-30 8-K and 10-Q, five and six days earlier.

The honest read: a stock that has been sold heavily for a year, has stabilised, and fell again on the day — six days after committing $6.25 billion of cash to an acquisition. A 43-point twelve-month deficit on a 0.948-beta infrastructure business is not a market phenomenon; it is a company-specific de-rating, and this file contains only part of the explanation. The valuation cushion is genuine — 17.0x FY2027, a 5.79% free-cash-flow yield, 20.8% below the high — and it is the reason this is Watch rather than Hold. What it is not is a reason to buy something the company has not yet explained.

8. Verdict, kill-criteria and flip conditions

Watch.

This dive has to do something unusual, so it does it first: the asymmetry clears the bar and we are declining the tier anyway.

Base fair value $165 (+10.6%), bear $116 (−22.3%), bull $222 (+48.8%) — 2.19:1. On arithmetic alone that is a Buy — Tactical. Three reasons it is not, all stated so a reader can disagree with them.

First, the largest capital commitment this company has made since 2023 is six days old and unquantified in this archive. On 29 July 2026 ICE agreed to acquire MarketAxess Holdings at $167.00 per share in cash, with no financing condition, backstopped by a $6.25 billion 364-day bridge facility from Bank of America. The archive contains no aggregate purchase price, no synergy estimate, no accretion arithmetic, no pro-forma leverage guidance and no MarketAxess financials — the earnings-release exhibits are not in the extraction. Our own arithmetic puts pro-forma net leverage at roughly 3.6x against 2.79x today, and the $1,202 million half-year buyback — 2.9% of market capitalisation annualised, a third of the expected return — is the obvious source of funds. We do not think a Buy should be issued on a capital structure that is about to change by an amount we cannot size.

Second, the most recent quarter decelerated sharply and the half-year figure hides it. Net revenue grew 4.8% in the June quarter against 20.6% in March. The FY2026 consensus needs 11.2%. And the first half contains a $452 million gain on investments that no vendor field flags and that will not repeat.

Third, the base case requires multiple EXPANSION — the only one in this batch that does. At 17.0x FY2027 the shares are below their own year-ago multiple, and our $165 is 18.8x. A thesis that needs the crowd to change its mind is weaker than one that does not, and after a 43-point twelve-month relative decline the crowd has been changing its mind in the other direction for a year.

And the honest counterweight, which is substantial. Thirty-three of thirty-six analysts are positive, there is not one sell rating, and the LOWEST target on the street is $163 — 9.2% above the price and one dollar below our base. The business is excellent: a 52.0% EBITDA margin, a 44.9% operating margin, 13 regulated exchanges and 6 clearing houses, and $4.85 billion of annual free cash flow at a 5.79% yield. Jeffrey C. Sprecher has compounded this company by acquisition for twenty-six years and has earned more benefit of the doubt than a stranger would. If MarketAxess is disclosed as accretive with leverage guided below 3.25x, this becomes a Buy immediately and without further analysis.

Pre-registered UPGRADE conditions — either of these makes this Buy — Tactical:

• Disclosed MarketAxess economics showing accretion in the first full year AND pro-forma net leverage guided below 3.25x, with the buyback maintained. This is the primary trigger and it is a disclosure event, not a price.
• Or a price at or below $132 with the FY2027 consensus of $8.78 intact — 15.0x FY2027E, 6.0% below the 200-day average, restoring approximately 5.6:1 asymmetry against the $116 bear and the $222 bull and providing enough cushion to underwrite the acquisition uncertainty without disclosure.

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

• The buyback suspended to fund the acquisition — it is roughly a third of the expected return.
• Pro-forma net leverage above 3.75x, or the $6.25 billion bridge being drawn in full rather than refinanced.
• A second consecutive quarter of net revenue growth below 5%.
• The MarketAxess consideration being increased, or a regulatory challenge to the merger.
• Another acquisition announced before MarketAxess closes — the XBRL implies a Q2 2027 completion, and a second commitment before then would change the character of the balance sheet.
• Mortgage Technology revenue declining year on year, which would make the $10.2 billion Black Knight purchase a cyclical bet rather than a structural one.

Where ICE fits in the Synthos Framework Portfolio. No position today. Tracked on the financials / market-infrastructure watch list with two triggers: the MarketAxess disclosure condition above, and a $132 price. Initial size 2%, target 4%. Sizing note: this is the name in this batch where the gap between the arithmetic and the information is widest — the numbers say buy and the file does not contain the single most important fact about the company's next two years. The correct response to that gap is a written trigger tied to a DISCLOSURE rather than to a price, which is why the primary condition above is not a number. Logged as a tracked Synthos call (Watch) as of 2026-08-04 at $149.20, with the fair-value anchors, both upgrade conditions, the kill criteria and the 4a assumptions all gradeable. Also logged as a DATA flag: this payload reports revenue on a gross basis 27.7% above the company's headline, its seg_prod block alternates between two measures across years, its June-quarter operating income is 43% too high, and its earn_cal and insider blocks are empty.

Single biggest risk: a $6.25 billion cash acquisition, announced six days ago, that this file cannot evaluate. ICE has committed to buy MarketAxess Holdings at $167.00 per share in cash with no financing condition, backstopped by a $6.25 billion bridge facility, on a balance sheet already carrying $20,279 million of debt at 2.79x EBITDA and $46.0 billion of goodwill and intangibles against $28,915 million of equity. The company has published no economics this archive can reach. On our arithmetic the transaction takes net leverage toward 3.6x and puts a buyback worth 2.9% of the market capitalisation at risk, three quarters before it is expected to close. The company's acquisition record — NYSE, Interactive Data, Ellie Mae, Black Knight — is good, and the market has nonetheless marked the shares down 20.8% from their high and 43 points behind the index over twelve months. At 17.0x FY2027 consensus with a 5.79% free-cash-flow yield, most of a bad outcome is already in the price. What is not in this dive is any basis for judging whether the outcome is bad, and that is why the verdict is Watch and why the primary upgrade trigger is a disclosure rather than a number.

Provenance & disclosures

• Traceability: 2 name-level knowledge-base claims name Intercontinental Exchange out of 52,021 distilled claims (raw hits 2, surviving a case-sensitive entity re-run 2, name-level 2, discarded as a name-level input 1; breadth 1, claim count 2, net conviction negative-low). Recorded separately as a favourable result: a three-letter ticker that is also an ordinary English word produced ZERO homograph contamination — no frozen water, no immigration enforcement — a materially cleaner outcome than CEG (Constellation Software on 26 of 29), PWR (Quanta Computer at conviction 85), EMR (Emerson Collective), USB (Universal Serial Bus) or MMM ("$3M" strings). The first claim — real_vision, 2022-08-31, conviction 80, entities tZERO / Intercontinental Exchange / Overstock — is four years old, concerns tZERO rather than ICE, and is discarded. The second is graded on CONTENT rather than on its stance tag and is the substantive finding: arthur_hayes, 2026-03-05, conviction 85, SKILL 1.2, entities Hyperliquid / CME / ICE, stance tagged BULLISH — "Perps' edge is clearing, not product spec: loser-can-only-lose-margin + ADL lets exchanges offer real leverage; guaranteed-settlement TradFi clearing can't compete." The claim is bullish on Hyperliquid and on the perpetuals clearing model; its explicit conclusion is a structural argument AGAINST the economics of ICE's and CME's guaranteed-settlement clearing houses. This is the confirmed STANCE-INVERSION defect, previously documented on CME and ABNB, appearing here on a second exchange in the same sector on the same theme — which suggests the inversion is systematic in this thematic neighbourhood rather than incidental. Graded: BEARISH, conviction 85, skill 1.2, one voice, horizon "principle" with no date attached. It is not enough to make name-level conviction negative — one market-structure claim over an unspecified horizon is not a view on the security — and it is emphatically not support; the exponential score of 4 and the long-horizon watch list both reflect it. Both claims carry speaker: null and are channel-attributed. Regulated exchange operation joins the list of thin sectors in this store. Quotes are verbatim from the stored claim text.
• Data as-of: fundamentals — income statement, balance sheet, cash flow, segment revenue, share counts and the transaction-based-expense reconciliation through 2026-06-30, all filing-verified from the 10-Q filed 2026-07-30; the merger agreement and bridge facility from the 8-K filed 2026-07-30 (event date 2026-07-29); annual figures for FY2025 and earlier from the 10-K filed 2026-02-05 · estimates 2026-08-04 · prices 2026-08-04, quote timestamp 1785873603 = 2026-08-04T20:00:03Z ($149.20, −1.71%; 50-DMA $140.21; 200-DMA $154.65; RSI 64.0; MACD +4.03; beta 0.948) · knowledge-base claims 2026-08-04. ICE's fiscal year is the calendar year. All figures come from the Synthos vendor data file for ICE or from the SEC filings in the ICE archive; no figure comes from memory, recall or external retrieval.
• Filing archive contents: 10-K filed 2026-02-05 (fiscal 2025); 10-Q filed 2026-04-30 (March 2026 quarter); 10-Q filed 2026-07-30 (June 2026 quarter — the source of the income statement with the transaction-based-expense reconciliation, the segment revenue detail, the cash-flow statement with the two capitalised-investment lines and the $452 million gain on investments, the balance sheet and share counts, and the MarketAxess ScenarioForecast XBRL tagging implying a Q2 2027 close); 8-K filed 2026-05-19 (Item 5.07 — annual meeting held 2026-05-15, eleven directors elected against 566,430,761 outstanding voting shares at the March record date); 8-K filed 2026-07-30 (Item 1.01 — the Agreement and Plan of Merger with MarketAxess Holdings at $167.00 per share IN CASH via Igloo Merger Sub II, with a $6.25 BILLION 364-day senior unsecured bridge facility from Bank of America and BofA Securities, and NO financing condition; checked as "Soliciting material pursuant to Rule 14a-12"); a second 8-K filed 2026-07-30 (second-quarter earnings release; Exhibit 99.1 NOT in the extracted text). All carry preserved [TABLE] statement data (tables: true); the 10-Q's 65 preserved tables made every check in Section 6 possible. The archive contains NO aggregate MarketAxess purchase price, NO synergy estimate, NO accretion arithmetic, NO pro-forma leverage guidance and NO MarketAxess financial statements — a gap stated repeatedly in this dive because it is the reason for the verdict.
• Where the filings contradicted or corrected the vendor (detailed in Section 6): the REVENUE GROSS-UP, confirmed — inc_a.revenue of $12,638M for FY2025 against est.revenueAvg of $9,896.5M, a 27.7% gross-up, with the 10-Q showing June-quarter total revenues of $3,611M less transaction-based expenses of $945M equalling the company's headline $2,666M; this is exactly the documented tell that flagged USB (50.2%), BNY (101%), IBKR (70%), COF and SCHW, and ICE was named as a candidate; consequential rejection of priceToSalesRatioTTM (6.24x against approximately 8.4x on the net basis), evToSalesTTM (7.69x against approximately 10.4x) and netProfitMarginTTM (30.0% against approximately 38%), with the ratio block internally inconsistent because operatingProfitMarginTTM of 44.9% is ALREADY on the net basis; seg_prod, which alternates between two different measures across years — FY2022 and FY2023 tie exactly to revenue while FY2020, FY2021, FY2024 and FY2025 sum to only 21-23% of it, with the FY2025 total of $2,659M suspiciously close to the implied transaction-based expense figure; inc_q.operatingIncome for the June 2026 quarter, at $1,987M against the filing's $1,391M, 42.8% too high, while incomeBeforeTax, netIncome and eps for the same quarter all match exactly and the March-quarter figure is correct — the same partial-row corruption documented on Robinhood in this batch; the complete absence of the MarketAxess acquisition and of the $452 million first-half gain on investments from every vendor field, the latter explaining an unannotated +77.3% March-quarter net income figure; est.ebitdaAvg/ebitAvg, rejected for a fixed-ratio fabrication signature (exactly 53.50% and 41.62% of revenueAvg from FY2024 to FY2030) with an impossible FY2023 outlier at 85.9%; and returnOnAssetsTTM (2.31%), financialLeverageRatioTTM (5.90x) and returnOnInvestedCapitalTTM (8.42%), all computed against a balance sheet inflated by clearing-house margin deposits that are not ICE's money — total assets swing from $136,887M to $174,247M in six months for that reason. Where vendor and filing AGREED — recorded, because clean verifications are real findings: the capital-expenditure check RECONCILES — ICE capitalises across two cash-flow lines ("Capital expenditures" $208M and "Capitalized software development costs" $230M in H1 2026), the vendor's FY2025 −$580M captures both, and the trailing rebuild of $662M matches the $666M implied by capexToDepreciationTTM of 0.430 to 0.6% — the opposite of the Robinhood case in this batch, where only one of two lines was captured and the field understated by 72%; the share count is EXACT (10-Q 655 million issued less 94 million treasury = 561 million outstanding at 2026-06-30, against 561.4 million implied by market capitalisation ÷ price); enterpriseValueTTM ($103,220M implying $19,460M of net claims against FY2025 net debt of $19,442M, an $18M residual, with the $54M of noncontrolling interests omitted at 0.05% of enterprise value — immaterial); and seg_geo, which ties to the NET revenue basis within 0.35% with the home market PRESENT at United States $6,316M, 63.8% — checked against the NEM / WM / MAR omission class and PASSED.
• Basis note: the company's headline revenue measure is "total revenues, less transaction-based expenses", and every revenue, margin and multiple in this dive uses it. Transaction-based expenses — Section 31 fees and cash liquidity payments, routing and clearing — are pass-through amounts collected and remitted, and were $945 million in the June quarter alone, 26.2% of gross revenue. Consensus epsAvg is on an adjusted basis with a consistent wedge against GAAP (FY2025 $6.916 against a reported diluted $5.77; FY2024 $6.047 against $4.78). Forward multiples use epsAvg; the trailing multiple of 21.1x is computed on GAAP diluted EPS and is shown separately. A separate earnings-quality note: the first half of 2026 contains a $452 million gain on investments, approximately $0.62 per share after tax, which is 7.7% of the FY2026 consensus and which the payload does not flag.
• Estimate coverage: 7 analysts on FY2026 EPS, 9 on FY2027 — the anchor for all three fair values, with a range of $8.276 to $9.264 (11.9%) — and only 4 on FY2028. FY2029 and FY2030 rest on TWO analysts each and are excluded from every conclusion. Revenue coverage is 10, 10 and 8. Against 36 analysts issuing ratings, EPS coverage of 7-9 is notably thin and is stated as such.
• Peer note: the vendor peer set — Brookfield Asset Management, Bank of New York Mellon, Bank of Nova Scotia, CME Group, Coinbase, Moody's, Mizuho, Marsh & McLennan, Nasdaq and Nu Holdings — contains three banks, an insurance broker and an alternative-asset manager. CME, Nasdaq and Coinbase are the three genuine exchange comparables, and Tradeweb — MarketAxess's principal competitor and therefore directly relevant to the pending acquisition — is absent, as are London Stock Exchange Group and Deutsche Börse. Note that Moody's is also in this batch and is analysed separately, and that BNY appears elsewhere in this programme as a name whose payload contradicts itself on net cash versus net debt. No peer-multiple comparison is drawn.
• Price-target note: pt reports targetHigh $208, targetLow $163, targetConsensus $183.25 and targetMedian $180 — a genuine four-point distribution. The LOW target of $163 is 9.2% ABOVE the spot price of $149.20, and grades shows 1 strong buy, 32 buy, 3 hold and ZERO sell across 36 analysts — 33 of 36 positive. As on Capital One, Williams, Robinhood and Moody's in this programme, this dive treats a covering community with no downside case as a caution rather than a confirmation — and here the caution has a specific object, being a $6.25 billion cash acquisition signed six days ago with no published economics. Separately, quote.yearHigh/yearLow ($189.35/$121.79) disagree with tech.hi52/lo52 ($188.45/$122.91) by 0.5% and 0.9%; tech is used throughout.
• Fair-value caveat: the $116 / $165 / $222 anchors are multiples of the FY2027 consensus EPS distribution — 14.0x the low of $8.276, 18.8x the mean of $8.780, and 24.0x the high of $9.264 — cross-checked against the free-cash-flow yield (7.4%, 5.2% and 3.9%). Stated arithmetic, not a discounted cash flow. The base is sensitivity-disclosed: 17x gives $149 — exactly today's price — and 21x gives $184, the street's consensus. The whole answer lives in a 17-21x band on one estimate row supported by nine analysts. This is the only name in this batch whose base case requires multiple EXPANSION, and that is stated as a weakness of the thesis rather than concealed: at 17.0x FY2027 the shares sit below their own year-ago multiple, and the base of $165 at 18.8x is still 12.4% below the 52-week high.
• Timing: second-quarter 2026 results were filed 2026-07-30, five days before this dive; no actual-versus-estimate comparison is available because the earn_cal block is EMPTY for this name. The next print is expected late October 2026 on filing cadence, approximately 85 days away. The dominant event is the MarketAxess merger agreement dated 2026-07-29, six days before this dive, with the 10-Q's XBRL tagging implying an expected close around the second quarter of 2027 — roughly three quarters away. The insider block returned no usable transactions. 2026-08-04 carried no company-specific filing.
• 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.