The Williams Companies WMB

> ## EARNINGS BANNER — REPORTED YESTERDAY

> Williams released second-quarter 2026 results on 2026-08-03, one day before this dive.

> Adjusted EPS $0.50 against a $0.502 estimate — in line to a fifth of a cent. Revenue $3,053M against $2,828M estimated, +7.9%. GAAP diluted EPS was $0.68; the gap is unrealized commodity-derivative marks.

> The shares rose 1.53% to $71.51 on 1.85x normal volume, having traded a $68.52-$72.53 range.

> Two July 2026 transactions absent from every vendor field dominate this dive: a $5.34 billion sale of a 49% interest in five power projects, and a pending $5.5 billion acquisition of Momentum.

> ## DATA BANNER — THIS FILING SET IS PROSE-ONLY

> The SEC extraction for Williams contains ZERO [TABLE] markers in the 10-K and ZERO in the 10-Q ("tables": false in the manifest, confirmed by direct inspection).

> No income statement, no balance sheet and no cash-flow statement is available to check the vendor payload against. Every narrative disclosure quoted below is filing-sourced; every number in the financial tables below is vendor-sourced and UNVERIFIED, including capital expenditure. This is stated rather than worked around, and it is why the confidence ratings in this dive are lower than the business quality alone would justify.

Reference table

1. What the business is, and the segment that was renamed

Four reportable segments, per the 10-Q filed 2026-08-03: "Williams' four reportable segments are Transmission, Power & Gulf; Northeast G&P; West; and Gas & NGL Marketing Services. All remaining business activities are included in Other."

Note the first name. The vendor's own seg_prod block for FY2024 calls it "Transmission And Gulf Of America", and the vendor's profile description calls it "Transmission & Gulf of Mexico" and says the company "operates through four key segments" under the old naming. The segment has been renamed to insert "Power", and the 10-Q explains why: Transmission, Power & Gulf "also includes power innovation projects under development that will deliver speed-to-market solutions". A segment rename that installs an entirely new business line into the company's largest reporting unit is a corporate action, and no vendor field records it.

The segment comprises "interstate natural gas pipelines and their related natural gas storage facilities, including Transco, NWP, MountainWest Pipelines Holding LLC (MountainWest), and a 50 percent equity-method investment in Gulfstream Natural Gas System", plus Gulf Coast gathering, processing and crude-oil production handling, plus storage and pipelines in north Texas, Louisiana and Mississippi.

Segment profit is measured as "Modified EBITDA", which the 10-Q defines as income before income taxes excluding upstream/corporate contributions, depreciation, equity earnings, other investing income, interest expense and ARO accretion, plus Williams' proportionate share of Modified EBITDA from equity-method investments. The chief operating decision maker is the chief executive officer. The segment Modified EBITDA figures themselves are in tables that did not survive extraction, so this dive discusses segment direction from the management commentary and does not quote segment profit numbers.

The power innovation business, in the company's own words

From the 10-K filed 2026-02-24:

> "Williams is investing in construction projects to support the power demands created by new data center and industrial development in power grid-constrained markets, including agreements with a large, investment-grade company to provide onsite natural gas and power generation infrastructure. The projects, located in Ohio and Utah, represent a combined 1.9 gigawatts of total capacity and are backed by up to 12.5 year, primarily fixed-price agreements, with an option for the customer to extend the term of the agreements. The projects will require additional capital to fund construction until the projects are placed in-service. Williams plans to place the projects into service during 2026 through 2028, assuming timely receipt of permits, and continues to pursue additional projects to support the power demands created by new data center and industrial development."

And on the lead project:

> "Williams has received approval from the Ohio Power Siting Board for the power generation facilities and is expecting final approval for the associated gas pipeline infrastructure in the first half of 2026. The Socrates project involves the construction of the Socrates North and South power generation facilities in New Albany, Ohio. Williams has agreed to provide committed power generation and associated gas pipeline infrastructure for the project, which is expected to provide a combined 400 megawatts of onsite power generation capacity to the customer. The project is backed by a 10 year [contract]..."

Four things to take from that. The scale is 1.9 gigawatts. The contracts are long and fixed-price, which makes them financeable and caps the upside. The counterparty is "a large, investment-grade company" and is not named — a 1.9-gigawatt, 12.5-year concentration to an undisclosed single customer. And the company told the market in February that the projects "will require additional capital" — which is precisely what the July transaction supplied.

The two July 2026 transactions, verbatim from the 10-Q

The Power Innovation joint venture:

> "In July 2026, Williams sold a 49 percent noncontrolling interest in five power innovation projects, Socrates, Apollo, Aquila, Socrates the Younger, and Neo, to an investor in exchange for $5.34 billion of committed capital. The initial July 2026 contribution of approximately $3.75 billion is expected to increase both Capital in excess of par value and Noncontrolling interests in consolidated subsidiaries, reflecting the change in Williams' ownership interest while retaining control as an equity transaction. The balance of the committed capital is expected to be received through early 2027. Cash distributions will generally align with ownership percentages and distributions to the investor in excess of a target return will serve to..."

The Momentum acquisition:

> "In July 2026, Williams agreed to acquire M6 Midstream LLC (Momentum) for total consideration up to $5.5 billion, including approximately $2 billion of Williams common stock, subject to certain holding restrictions. Momentum's assets in the Haynesville Shale region include 6 Bcf/d of gathering capacity and 4 Bcf/d of pipeline capacity. The transaction is expected to close later this year, subject to customary closing conditions and regulatory approvals."

How to read the $5.34 billion, carefully. It is tempting to annualise it into a valuation — 49% for $5.34 billion implies roughly $10.9 billion for the five projects, about 12% of Williams' entire market capitalisation for a business that generates almost nothing today. We do not draw that conclusion and here is why. The disclosure calls it "committed capital" delivered in instalments through early 2027, credited partly to "Capital in excess of par value", with distributions "in excess of a target return" subject to a mechanism the extracted text truncates. That is the structure of a development-capital partnership with a preferred return, not a clean sale of a half-interest at a mark. What it does establish, and this is worth a great deal on its own, is that a sophisticated third party has underwritten the asset class and the counterparty at a scale of $5.34 billion. It also means Williams' noncontrolling interests — $2,188 million at 2025-12-31 — are about to rise by something on the order of $3.75 billion, which has direct consequences for enterprise value discussed in Section 4.

Two further filing-sourced items with no vendor trace: the 8-K of 2026-07-01 records the appointment of Robb E. Turner and Lloyd W. (Billy) Helms, Jr. to the board, expanding it from ten to twelve directors, both independent; and the 10-Q records that in December 2025 management "approved a plan to sell certain gas gathering assets in the Mid-Continent region", designated held for sale at 31 December 2025. The 10-Q also notes the Cogentrix equity-method investment "purchased in March 2025" as a driver of Gas & NGL Marketing Services results, and the acquisition of the remaining interest in Gulfstar One in December 2025.

2. The trajectory

Two observations that shape everything else.

First, the earnings series is volatile for a pipeline company and the cause is derivative accounting. FY2023's $2.60 and FY2024's $1.82 differ by 30% on essentially flat revenue. The 10-Q is explicit: "Williams experiences significant earnings volatility from the fair value accounting required for the derivatives used to hedge a portion of the economic value of the underlying transportation and storage capacity portfolios as well as upstream-related production. However, the unrealized fair value measurement gains and losses on the derivatives are generally offset by valuation changes in the economic value of the underlying production or transportation and storage capacity contracts, which are not recognized until the underlying transaction occurs." est.epsAvg tracks the adjusted series and matches FY2025 almost exactly ($2.12 against a reported $2.14), which is why this dive values on epsAvg.

Second, free cash flow has collapsed as capital expenditure has doubled — from $2,283M in FY2022 to $4,999M in FY2025, with the trailing figure now running near $6.2 billion. FY2025 free cash flow of $899 million did not come close to covering the $2,442 million of dividends paid, and the trailing figure is negative.

The quarterly record, and note the two EPS bases:

The June quarter was in line to a fifth of a cent on earnings and 7.9% ahead on revenue. The GAAP-to-adjusted gap of $0.18 in that quarter is the largest in the series and runs the favourable way, which means the reported profit was flattered by unrealized derivative gains that consensus correctly excluded. A reader looking only at the vendor's inc_q would see EPS of $0.68 against $0.45 a year earlier — a 51% increase — and conclude the quarter was spectacular. It was fine.

The one filing-verifiable number in this entire dive, and it checks: the 10-Q states that "Net income (loss) attributable to The Williams Companies, Inc. for the six months ended June 30, 2026, increased $455 million compared to the six months ended June 30, 2025." The vendor's quarterly series gives $865M + $827M = $1,692M for the 2026 half against $691M + $546M = $1,237M for 2025 — a difference of exactly $455 million. That is the only cross-check the prose-only extraction permits, and the payload passes it.

3. Cash flow — the check that cannot be completed, and what the direction says anyway

The capital-expenditure check CANNOT BE PERFORMED against the filing, and this dive states that rather than working around it.

The manifest records "tables": false for all six Williams filings. Direct inspection confirms it: the 10-K contains zero [TABLE] markers across 619,320 characters, and the 10-Q zero across 238,802. Searches for the FY2025 capital-expenditure figure ($4,999), the operating-cash-flow figure ($5,898) and the year-end cash balance ($2,845 — itself a figure the payload does not carry) all return no hits. The consolidated statements of income, financial position and cash flows are referenced by name in the extracted text and their contents are absent.

So the following is entirely vendor-sourced and labelled as such:

Trailing figures derived from operatingCashFlowPerShareTTM ($4.897) and capexToOperatingCashFlowTTM (1.0357) against 1,223.0 million shares; they reproduce freeCashFlowPerShareTTM of −$0.1748 and freeCashFlowYieldTTM of −0.2447%.

Now the important analytical point, and it is the mirror image of the usual finding.

The capex diagnostic trips the defect screen. capexToDepreciationTTM of 2.744 and a swing from 1.21x to 2.74x in two years is exactly the "capex/D&A swings more than 3x across three years" signature that flags 919 files in the programme's census. On most names that signature accompanies an understated capex and a flattered free-cash-flow yield.

Here it accompanies the opposite: a capex figure that is, if anything, corroborated by the narrative, and a free-cash-flow yield that is NEGATIVE. The 10-K states the power projects "will require additional capital to fund construction until the projects are placed in-service". The 10-K's list of named projects placed in service during 2025 — Commonwealth Energy Connector, Texas Louisiana Energy Pathway, Southeast Energy Connector, Alabama Georgia Connector, Shenandoah, Overthrust Westbound compression — is a heavy construction programme. And in July 2026 the company raised $5.34 billion of third-party committed capital explicitly to fund the same build. A company that has just sold 49% of five projects to finance their construction is not a company whose capital expenditure is being overstated.

This is the MPC lesson in reverse and it is worth recording as such: on MPC the diagnostic read impossibly while the datum was correct. Here the diagnostic reads suspiciously and the datum is corroborated by the narrative — so the suspicion is dismissed on evidence rather than on the diagnostic's own authority. The honest caveat is that "corroborated by the narrative" is weaker than "matched to the filing", and the filing is unavailable. The dive therefore quotes a recomputed free-cash-flow yield of −0.24% while stating that the underlying capex figure is unverified.

The consequence for the dividend is not a nuance. Williams paid a quarterly dividend of $0.525 per share in June 2026 — a figure the 10-Q states directly — which annualises to $2.10, a 2.94% yield, and costs roughly $2.57 billion a year on 1,223 million shares. Trailing free cash flow is approximately negative $214 million. The dividend is currently funded by borrowing and by the proceeds of the joint-venture sale, not by operations. For a growth-capex-heavy midstream that is a normal and defensible financing structure during a build cycle, and it is also the single thing that makes this equity fragile if the build slips. dividendPaidAndCapexCoverageRatioTTM of 0.688 says the same thing in one number: operating cash flow covers 68.8% of dividends plus capital expenditure.

One qualification in the company's favour, stated because it is real: midstream capital expenditure divides into maintenance and growth, and only maintenance is a genuine call on distributable cash. The vendor does not split it and the filing extraction does not permit the split. With depreciation running at $2,347 million in FY2025, maintenance capex is plausibly a small fraction of the $6.2 billion trailing figure — which would make the underlying free cash flow substantially positive. We cannot demonstrate that from this file and therefore do not claim it.

4. Balance sheet and enterprise value — and an NCI omission about to get much larger

Three observations.

Goodwill went to zero in FY2025 — $466 million at the end of 2024, nil at the end of 2025. The payload gives no reason and the prose-only extraction offers none. The 10-Q's only adjacent disclosure is that gas gathering assets in the Mid-Continent region were designated held for sale in December 2025. An unexplained disappearance of the entire goodwill balance is flagged rather than interpreted.

Net debt of $29,330 million at 4.10x netDebtToEBITDATTM is high, and the trajectory is the wrong way: up $5.0 billion in two years while property, plant and equipment rose $7.7 billion. That is a company borrowing to build, which is what it says it is doing.

Enterprise value cannot be rebuilt from a filing and the NCI question is live. enterpriseValueTTM reads $118,047M against a market capitalisation of $87,457M, implying $30,590M of net debt and other claims. Against the December 2025 balance sheet, net debt of $29,330M plus preferred of $35M is $29,365M, leaving a $1,225M residual — plausibly the increase in borrowings across the first half of 2026, which the prose-only extraction cannot confirm.

What is clear is that the $2,188 million of noncontrolling interests is NOT in it. Adding it gives approximately $120,235M, 1.9% above the vendor's figure. That is small — and it is about to stop being small. The July 2026 joint venture "is expected to increase both Capital in excess of par value and Noncontrolling interests in consolidated subsidiaries" by approximately $3.75 billion immediately, rising toward $5.34 billion through early 2027. On completion, Williams' noncontrolling interests will be on the order of $6-7.5 billion — roughly 6% of enterprise value — and a payload that omits NCI will understate enterprise value by that amount. This is the KKR / MPC / FCX defect class caught in advance rather than after the fact, and it is registered here as a forward-dated data flag: any Williams enterprise-value multiple computed from this vendor after the third-quarter balance sheet publishes should be treated as understated by approximately 5-6%.

5. Segments and geography — the worst blocks in this batch

seg_prod fails in two different ways in adjacent years.

FY2025 drops Transmission, Power & Gulf and Northeast G&P entirely — including the segment that contains Transco, the largest asset in the company, and the power business that is the entire growth story. FY2024 contains all four but has no eliminations line, so intersegment revenue is double-counted and the block sums to 136% of revenue — the CDNS/CSX over-count class. Both defect classes, same block, adjacent years, on one name. No segment revenue figure from this block is used anywhere in this dive.

seg_geo is a category error. The single FY2025 entry reads:

`

{"West": 2847000000}

`

"West" is a reportable SEGMENT of Williams, not a geography. It is also the identical value to the West line in seg_prod. The geography block has been populated with a segment name and a segment number. The United States line is absent, and in this case there is nothing else it could be: the 10-Q states plainly that Williams' "operations are located in the United States". Williams is a 100% domestic company; the correct geography block is a single United States line equal to total revenue. This is a more severe failure than the NEM / WM / MAR home-market omissions, because the block does not merely omit the home market — it substitutes a different dimension entirely. The block is rejected in full.

6. Valuation — priced in or room?

At $71.51 (market cap $87.457B, 1,223.0M shares):

Estimate coverage is thin and one row is corrupt. 9 analysts on FY2026 EPS, 10 on FY2027, 8 on FY2028, and only 6 on FY2029 and FY2030, both excluded from every conclusion. The FY2027 epsLow reads MINUS $3.62863 — an impossible figure for a fee-based pipeline company whose worst annual EPS in the last decade was $0.17, and one that makes the FY2027 range ($−3.63 to $3.74) unusable. Only the FY2027 epsAvg and epsHigh are used; the low anchor for the bear case is derived from a multiple rather than from the corrupt estimate row, and that substitution is disclosed.

est.ebitdaAvg and est.ebitAvg are REJECTED with the signature stated. In every year from FY2026 to FY2030, ebitdaAvg is exactly 59.04% of revenueAvg and ebitAvg is exactly 39.96%. Williams' actual FY2025 EBITDA margin on the vendor's own income statement was 62.0%, so the fabricated ratio is not wildly wrong in level — it is wrong in kind, being a fixed ratio applied mechanically across five forward years in a business whose segment mix is changing materially. All forward valuation uses epsAvg.

A note on why a 28.6x P/E is not the shock it appears. Midstream companies carry very heavy depreciation against long-lived assets — Williams' FY2025 D&A of $2,347 million is 19.6% of revenue — so GAAP earnings understate cash generation and P/E multiples run structurally high. The relevant cross-check is EV/EBITDA at 15.8x, and that IS a premium: large-cap North American midstream typically trades in a 9-12x band. Williams' premium is the power business, and the question this dive has to answer is whether roughly four turns of EBITDA — call it $30 billion of enterprise value — is the right price for 1.9 gigawatts of contracted generation plus the option on more.

6a. What today's price assumes (the inversion)

At $71.51 — 28.6x FY2027 consensus and 15.8x trailing EV/EBITDA — the price embeds:

• Adjusted EPS reaches $2.38 in FY2026 and $2.50 in FY2027, then steps 21.7% to $3.04 in FY2028. (Consensus; 9, 10 and 8 analysts.) The first half delivered $1.23 of adjusted EPS ($0.73 plus $0.50), so FY2026 needs $1.15 in the second half — achievable. The FY2028 step is where the power projects are assumed to commission, consistent with the 10-K's "in service during 2026 through 2028".
• The 1.9 gigawatts is delivered on schedule and on budget. (Our derivation from the 10-K's project descriptions.) This is the most fragile assumption in the price. Gas turbines are the industry's binding supply constraint, the 10-K explicitly lists "delays in capital project execution, including increases from inflation or delays caused by supply chain disruptions" and "risk of delay or denial in permits and approvals" among its risks, and the Socrates gas pipeline infrastructure was still awaiting final approval as of the February 2026 filing.
• The market keeps paying roughly 15-16x EV/EBITDA — four turns above the midstream norm. (Our number.) Every turn of EBITDA is roughly $7.5 billion of enterprise value, about 8.5% of the market capitalisation. A reversion to 13x, still a premium, is roughly $21 billion — 24% of the equity.
• The dividend keeps growing while free cash flow is negative. (Our derivation; the June 2026 quarterly rate is $0.525 per the 10-Q, and the payload's dividendPerShareTTM of $2.05 implies a rising series.) At roughly $2.57 billion a year against approximately negative $214 million of free cash flow, the dividend is a financing decision, not a distribution of surplus.
• Momentum closes at or near $5.5 billion and earns its price in the Haynesville. (Company disclosure; our characterisation.) Approximately $2 billion of it is Williams stock — roughly 28 million shares, 2.3% dilution at the current price — and the balance is cash added to $29 billion of net debt.

6b. The return bridge (why the multiple moves)

Expected return over the next twelve months decomposes as: adjusted EPS growth (+5.0%, FY2026E $2.380 to FY2027E $2.500) + multiple drift (roughly HELD, from 30.0x on today's forward year to about 28.5x on the then-forward year) + dividend yield (+2.94%) ≈ +7%.

Note what that decomposition reveals: on the FY2026-to-FY2027 step, consensus earnings grow only 5.0%, and the dividend is more than half the expected return. The whole equity case is the FY2028 step of 21.7% as the power projects commission — which means this is a two-to-three-year investment being asked to justify a 28.6x multiple on a one-year view.

Our base assumes the multiple is roughly held rather than expanded, and the reason is that a 15.8x EV/EBITDA already prices a successful power business. We are not underwriting further expansion. If the projects commission on schedule and the company announces a second tranche of comparable size, expansion is the bull case, not the base. If they slip, the compression is severe and immediate, because 15.8x EV/EBITDA has no support from the pipeline assets alone.

At 22x FY2027E the price is $55 (−23.1%). At 32x it is $80 (+11.9%). Our bull case of $96 is 31.6x the FY2028 consensus of $3.042 — that is, it requires the power projects to be in service and earning before the multiple is applied.

6c. Variant perception (where we differ, what would surprise)

• We think the July transactions are the most important thing about this company and are not yet legible in any screen. A $5.34 billion third-party commitment for 49% of five power projects, and a $5.5 billion acquisition, both agreed in July 2026, appear in no vendor field: bal_a ends 2025-12-31, cf_a ends FY2025, seg_prod has FY2025 as its latest year and drops two segments, and profile still describes the old four-segment structure with "Transmission & Gulf of Mexico". Anyone screening Williams on this payload is looking at a company that no longer exists. Watchable event: the third-quarter balance sheet on or about 2026-11-02, which will show noncontrolling interests roughly tripling.
• We differ from the street on the level, and modestly. Consensus target is $83.54 (+16.8%) against our $76 (+6.3%) — a 9.0% gap. More striking than the level is the shape of the street's distribution: the LOW target is $75, 4.9% ABOVE the current price, and there is not a single sell rating among 34 analysts. We treat unanimity about the absence of downside as a caution rather than a comfort, exactly as on Capital One, and more so here because the balance sheet carries 4.10x leverage and negative free cash flow.
• We think the negative free cash flow is being under-weighted and the dividend coverage question deserves a direct answer nobody is giving. Trailing free cash flow of approximately −$214 million against dividends of roughly $2.57 billion is a gap of about $2.8 billion a year, currently bridged by borrowing and by the joint-venture proceeds. Watchable number: dividendPaidAndCapexCoverageRatioTTM, currently 0.688 — operating cash flow covers 68.8% of dividends plus capex. A midstream in a build cycle can run this for two or three years. It cannot run it indefinitely, and the projects must commission for it to stop.
• The knowledge base has NOTHING on this name and we treat the void as an input rather than a neutral. Zero entity matches in 52,021 claims. The only adjacent material is a coherent sector thesis from a single voice — doomberg — arguing for exactly this asset class, and its most specific expression names Energy Transfer, a competitor, as the preferred vehicle for "a US midstreamer leveraged to increasing natural gas volumes with an AI arc and a nice dividend", while noting he does not own it. That is one independent voice describing the Williams thesis and choosing someone else. It carries no weight in the conviction score and it is not nothing.
• Positive surprise that would force a re-rating: a second Power Innovation tranche of comparable scale, or the naming of the investment-grade counterparty, or a disclosed Power Innovation backlog in gigawatts and contracted dollars — Williams currently publishes capacity but not contracted revenue; or Socrates entering service ahead of schedule; or free cash flow turning positive while the dividend keeps growing.
• Negative surprise that would break the thesis: any project slipping beyond 2028 on turbine availability or permitting; the Momentum acquisition closing above $5.5 billion or with more stock; leverage rising through 4.5x; the dividend growth rate slowing, which would be management's own signal that the build is not self-funding; or a write-down at Momentum in a Haynesville basin whose volumes are themselves a forecast.

Synthos fair values

Anchored on the FY2027 consensus adjusted EPS mean of $2.500 (10 analysts) and the FY2028 mean of $3.042 (8 analysts). The FY2027 epsLow of −$3.62863 is corrupt and is NOT used; the bear anchor is derived from a multiple instead, and that substitution is disclosed rather than hidden.

• Bear ~$55 — 22.0x the FY2027 consensus MEAN of $2.500, i.e. a de-rating rather than an earnings miss. Cross-check: 23.1x FY2026E; approximately 13x EV/EBITDA, still a midstream premium; 2.7% below the 52-week low of $56.51. The scenario: power projects slip past 2028 on turbine or permitting delay, leverage rises through 4.5x with Momentum, dividend growth stalls, and the market prices Williams as the pipeline company it currently is. −23.1%.
• Base ~$76 — 30.4x the FY2027 consensus MEAN of $2.500. Cross-check: 31.9x FY2026E; 25.0x FY2028E; approximately 16.5x EV/EBITDA. Sensitivity, stated openly: 28x gives $70 — essentially spot — and 33x gives $82.50, the street's median. The entire answer lives in a 28-33x band on one estimate row supported by ten analysts, with a corrupt low. The scenario: the projects commission across 2026-2028 broadly on schedule, Momentum closes and is neutral to modestly accretive, the dividend keeps growing, and the multiple holds. +6.3%.
• Bull ~$96 — 31.6x the FY2028 consensus of $3.042, i.e. the projects in service and earning before the multiple is applied. Cross-check: 38.4x FY2027E; 20.9% above the 52-week high of $79.40. The scenario: 1.9 gigawatts delivered, a second tranche announced, the counterparty named and creditworthy, free cash flow turning positive, and Williams re-rated as a contracted power-infrastructure company rather than a midstream. +34.2%.

Base is 6.3% above spot; asymmetry roughly 1.48:1 (23.1% down, 34.2% up), before a dividend yield of 2.94%. That is a real payoff and it is below the 2:1 bar, on a file where not one financial-statement number could be checked against a filing. Watch.

7. Knowledge base — zero name-level claims, reported as zero

Raw entity hits: 0. Free-text hits: 11. Name-level claims on Williams: 0. Discarded: 11.

The case-sensitive entity sweep ran WMB, Williams and Williams Companies across all 52,021 distilled claims. It returned NOTHING. No homograph collision, no partial match, no substring artefact — and no coverage. US natural gas midstream is a genuine sector void in this store, and it joins insurance brokerage, environmental services, diversified industrials and architectural coatings on the list of categories the ingest pipeline does not reach.

A free-text sweep on midstream returned 11 hits. Every one is sector-level and not one names Williams. The lane carries ZERO weight in the conviction rating and is reproduced below only because a reader is entitled to know what was searched and what came back.

The thematic material, quoted separately and at zero weight:

> 2025-12-22 · neutral · conviction 75 · horizon: principle · channel: doomberg · speaker: doomberg · role: independent · skill 1.0

> "Don't own commodities or price-taking producers; own companies leveraged to volume — service providers, enablers and midstream — since energy demand always grows regardless of price."

> 2026-05-27 · bullish · conviction 60 · horizon: thesis · channel: doomberg · speaker: Doomberg

> "Prefer midstreamers, service providers and royalty owners that profit as energy consumption rises over capital-intensive E&P price-takers, which get low multiples and make money only in bursts."

> 2026-05-23 · bullish · conviction 48 · horizon: thesis · entities: ET · channel: doomberg · speaker: Doomberg

> "As a $100k trade, buy Energy Transfer — a US midstreamer leveraged to increasing natural gas volumes with an AI arc and a nice dividend; leveraged to energy volume, not price. (Notes he doesn't own it.)"

The remaining eight hits are further from the mark still: three doomberg claims about Hormuz, Middle East reconstruction and the "great reconciliation" against renewables; two macrovoices claims about critical-metals refining using "midstream" in an entirely different sense (rare earths, China's processing chokepoint); one about desalination "midstreamers"; one about Permian produced water; and one from interviewers favouring "capital-light royalty/streaming and midstream toll-booth models". Two of the eleven are not about energy midstream at all.

What this lane is, stated without inflation. One voice — doomberg — holds a consistent, repeatedly expressed, directionally favourable view of the asset class Williams occupies. It is thematically correct about the mechanism: fee-based, volume-leveraged infrastructure rather than commodity price exposure, which is precisely how Williams' revenue works. And its single most specific expression names Energy Transfer, not Williams, as the vehicle — with an explicit "AI arc" reference that is the Williams thesis exactly — and the speaker notes he does not own it. One independent voice described this trade and picked a competitor. We record that plainly and give it no weight in the score.

Attribution note: all eleven hits carry a channel; three carry speaker: doomberg or Doomberg with speaker_role: independent. Under the standing rule that speaker_role is unreliable, the tag is noted rather than relied upon. No management voice appears anywhere in this lane. No concentration test is meaningful across zero name-level claims.

Conclusion. Breadth 0, claim count 0, net conviction none. The Synthos knowledge base has nothing to say about The Williams Companies, and this dive reports that as the absence it is. It is a real input to the Watch verdict: a name at a 15.8x EV/EBITDA premium, with unverifiable financials and negative free cash flow, has nothing but its own filings to lean on.

8. Data integrity — the least verifiable file in this batch

Seven findings. Williams' file is the weakest in this batch, and the principal finding is not any single defect but the fact that none of them could be settled against a filing.

0. THE FILINGS ARE PROSE-ONLY — the capex check CANNOT BE COMPLETED. The manifest records "tables": false for all six Williams filings, and direct inspection confirms zero [TABLE] markers in the 10-K (619,320 characters) and zero in the 10-Q (238,802 characters). Searches for the FY2025 capital expenditure ($4,999), operating cash flow ($5,898) and year-end cash ($2,845) all return no hits. The consolidated statements of income, financial position and cash flows are named in the extracted text and their contents are absent. Every figure in the financial tables of this dive is vendor-sourced and unverified, and is labelled as such. The only cross-check the extraction permits — the 10-Q's statement that first-half net income attributable to Williams rose $455 million — the payload PASSES exactly ($1,692M against $1,237M). That is one data point of corroboration and it is reported as one data point.

1. capitalExpenditure — the diagnostic trips, the datum is corroborated by narrative, and the direction is UNFAVOURABLE rather than flattering. capexToDepreciationTTM reads 2.744 and the ratio swung from 1.21x (FY2024) to 2.13x (FY2025) to 2.74x trailing — the "swings more than 3x" suspicion signature. But the resulting trailing free-cash-flow yield is NEGATIVE 0.24%, evToFreeCashFlowTTM is −551.6x and priceToFreeCashFlowRatioTTM is −408.7x. The programme's census establishes that the capex error "always flatters"; this file does the opposite. The narrative corroborates the level: the 10-K states the power projects "will require additional capital to fund construction", lists six named projects placed in service during 2025, and the company raised $5.34 billion of third-party capital in July 2026 to fund the same build. The suspicion is dismissed on evidence rather than on the diagnostic's authority — and the caveat is stated: "corroborated by narrative" is weaker than "matched to filing", and the filing is unavailable.

2. seg_prod fails in TWO different ways in adjacent years — REJECTED in full. FY2025 contains only two of four segments (Gas & NGL Marketing Services $7,175M and West $2,847M, summing to $10,022M against revenue of $11,950M — 83.9%, with Transmission/Power/Gulf and Northeast G&P both absent, including the segment holding Transco and the entire power business). FY2024 contains all four but no eliminations line, summing to $14,299M against revenue of $10,503M — 136.1%, an over-count of $3,796M of the CDNS/CSX class. FY2023 is missing one segment at 64.6%. Both confirmed defect classes, on one name, in consecutive years. No figure from this block is used.

3. seg_geo contains a SEGMENT name in place of a geography — REJECTED in full. The sole FY2025 entry is {"West": 2847000000} — identical in name and value to the West line in seg_prod. The United States is absent, and the 10-Q states that Williams' "operations are located in the United States", so the correct block is a single domestic line equal to revenue. This is more severe than the NEM / WM / MAR home-market omissions: the block does not omit the home market, it substitutes a different reporting dimension entirely.

4. Two July 2026 transactions totalling roughly $10.8 billion are absent from every vendor field. The $5.34 billion Power Innovation joint venture (49% of five named projects, ~$3.75 billion received July 2026) and the up to $5.5 billion Momentum acquisition (including ~$2 billion of Williams stock). bal_a ends 2025-12-31, cf_a ends FY2025, and profile still describes the pre-rename four-segment structure with "Transmission & Gulf of Mexico". This is the "whole transaction absent" defect class, twice, on one name, within one month.

5. est.epsAvg FY2027 epsLow reads MINUS $3.62863 — CORRUPT, and the affected anchor is rebuilt. A negative low estimate for a fee-based pipeline company whose worst annual GAAP EPS in the last decade was $0.17 is not a bearish view; it is a broken number. It makes the FY2027 range ($−3.63 to $3.74) unusable. The bear-case anchor in this dive is therefore derived from a multiple applied to the consensus MEAN rather than from epsLow, and that substitution is disclosed in Section 6.

6. est.ebitdaAvg and est.ebitAvg carry a fixed-ratio fabrication signature — REJECTED. In every year from FY2026 to FY2030, ebitdaAvg is exactly 59.04% of revenueAvg and ebitAvg is exactly 39.96%. Williams' actual FY2025 EBITDA margin was 62.0%, so the level is not absurd — the defect is that a fixed ratio is applied mechanically across five forward years in a business whose segment mix is being deliberately changed. All forward valuation uses epsAvg.

7. Enterprise value omits noncontrolling interests, and the omission is about to grow roughly threefold. enterpriseValueTTM of $118,047M implies $30,590M of net debt against a December-2025 net debt of $29,330M plus $35M preferred — a $1,225M residual plausibly explained by first-half 2026 borrowing that the prose-only extraction cannot confirm. The $2,188M of noncontrolling interests is not included; adding it gives approximately $120,235M, 1.9% higher. The July 2026 joint venture will increase noncontrolling interests by approximately $3.75 billion immediately and toward $5.34 billion by early 2027, taking the balance to roughly $6-7.5 billion. Registered as a forward-dated flag: from the third-quarter balance sheet onward, any Williams enterprise-value multiple from this vendor should be treated as understated by approximately 5-6%.

Two further items, both flagged and neither used. goodwill went from $466M at 2024-12-31 to ZERO at 2025-12-31 with no explanation available in the payload or the extracted prose; the only adjacent disclosure is the December 2025 designation of Mid-Continent gathering assets as held for sale. And five of the eight insider records — all dated 2026-08-04 — carry securitiesTransacted: 0, price: 0 and securitiesOwned: 0 on formType: 4, which is not a transaction but an empty row; see Section 10.

Non-equity tripwire — checked and passed, with one note. WMB is common stock listed on the NYSE; the 10-Q states "Williams is a Delaware corporation whose common stock is listed and traded on the New York Stock Exchange." Price of $71.51 is not par-like; beta is 0.608, the lowest in this batch and consistent with regulated infrastructure rather than a fixed-income instrument; the dividend is a declared and raised quarterly rate ($0.525 in June 2026), not a coupon; volume was 13.09M shares (~$936M of turnover) against a 7.06M average; the 52-week band of $56.51 to $79.40 is a 40% range, far too wide for a debt-like security. Note the separate $35 million of preferred stock and $2,188 million of noncontrolling interests, neither of which is this security. This is common equity.

Vendor composite rating — noted, and not used. B− / 3 overall, with 5 on return on equity, 4 on return on assets, and 1, 1 and 1 on debt-to-equity, price-to-earnings and price-to-book. The discounted-cash-flow sub-score of 3 is built on a free-cash-flow figure that is negative; whatever it means, it is not a valuation.

9. Technicals

• Price $71.51. −9.94% from the 52-week high of $79.40; +26.54% above the low of $56.51. Position within the annual range: 65th percentile.
• Below the shorter average, above the longer: 2.5% BELOW a 50-day average of $73.37; 4.3% ABOVE a 200-day of $68.58. The 50-day remains above the 200-day.
• RSI 39.7 — the second-weakest in this batch, and below the 50 midline. MACD −0.885.
• Maximum drawdown from peak over the trailing year: −9.94%, identical to the distance from the high, so the recent low is the trough of the drawdown.
• Beta 0.608 — the lowest in this batch by a wide margin, and the single most important technical fact about this name: it is a defensive-profile security carrying an offensive-profile capital programme.
• Relative performance: 3-month −5.17% against SPY +7.59% and QQQ +7.67% — thirteen points behind; 6-month +7.79% against SPY +11.09%; 12-month +18.65% against SPY +24.26% and QQQ +30.80%. Williams has underperformed the index over every window measured here, which for a name executing a transformation this large is either an opportunity or a verdict, and the dive does not claim to know which.

Yesterday's print and today's move

WMB closed 2026-08-04 at $71.51, up 1.53% or $1.08 from $70.43. It opened at $69.57 — below the previous close — traded $68.52 to $72.53, and closed within 1.4% of the day high on 13.09 million shares against a 7.06 million average, 1.85 times normal. The second-quarter release and 10-Q were both filed 2026-08-03, one day earlier.

The honest read: a mildly positive reaction to an in-line quarter, on elevated volume, in a stock that has been drifting. Adjusted EPS of $0.50 matched the $0.502 estimate to a fifth of a cent; revenue beat by 7.9%. There is no valuation signal in a 1.53% move. What the setup does offer is a technically unexcited entry — below the 50-day average, RSI under 40, 9.9% off the high — which is materially better than the entries available on Sherwin-Williams or Capital One in this batch, and is why the verdict is Watch rather than Avoid despite the leverage and the unverifiable data.

10. Insiders — one real transaction and five empty rows

One genuine transaction: Terrance Lane Wilson, Senior Vice President and General Counsel, sold 2,000 shares at $70.65 on 2026-08-03 — the day the second quarter was reported — retaining 281,159 shares. The sale is 0.7% of his holding and the retained position is worth approximately $20.1 million. That is a routine disposal by an executive with a very large remaining stake, and it carries essentially no signal.

Five of the eight rows are EMPTY — a data defect, not an insider event. The records for Payvand Fazel, Glen G. Jasek, Thomas F. McCoy, Eric J. Ormond and Todd J. Rinke, all dated 2026-08-04 and all coded formType: 4, carry zero shares transacted, zero price, zero securities owned and a blank transaction type. A Form 4 with no transaction is not a filing state that exists. These are most likely truncated captures of routine filings; whatever they are, they contain no information and are not interpreted.

The two Robb E. Turner rows dated 2026-07-01 are Form 3 initial statements of beneficial ownership, consistent with his appointment to the board on that date per the 8-K, showing indirect holdings of 84,000 and 6,000 shares. A new director arriving with 90,000 shares already held is mildly notable and is recorded as such.

What the file does not contain: any transaction by the chief executive or chief financial officer, and no purchases by anyone. Across the largest strategic transition in this company's recent history — a $5.34 billion joint venture and a $5.5 billion acquisition both agreed in July 2026 — the insider file records one 2,000-share sale. That is uninformative in both directions and is reported as such.

11. Verdict, kill-criteria and flip conditions

Watch.

What is genuinely interesting here, stated first because it is real. Williams is building 1.9 gigawatts of onsite natural-gas power generation for data centres in grid-constrained Ohio and Utah, "backed by up to 12.5 year, primarily fixed-price agreements" with "a large, investment-grade company", in service "during 2026 through 2028", with Socrates North and South (400 MW, New Albany, Ohio, ten-year contract) already approved by the Ohio Power Siting Board. It has renamed its largest segment "Transmission, Power & Gulf" to hold this business. And in July 2026 a third party committed $5.34 billion for a 49% interest in five of the projects, delivering approximately $3.75 billion immediately — an outside underwriting of the asset class that no amount of company narrative could substitute for. Alongside it, an agreed $5.5 billion acquisition of 6 Bcf/d of Haynesville gathering that fits the 10-K's own forecast of "expected increases in Haynesville Shale volumes". None of this is in the vendor payload. Anyone screening Williams on that data is looking at a company that no longer exists.

Why it is not a Buy, in four numbers.

First, the asymmetry is 1.48:1. Base $76 (+6.3%), bear $55 (−23.1%), bull $96 (+34.2%), plus a 2.94% dividend. The bar is 2:1 and this does not clear it.

Second, free cash flow is negative and the dividend is not covered. Trailing operating cash flow of approximately $5,989 million against trailing capital expenditure of approximately $6,203 million gives free cash flow of about minus $214 million, against dividends costing roughly $2.57 billion a year. dividendPaidAndCapexCoverageRatioTTM is 0.688. The gap is bridged by borrowing and by the joint-venture proceeds. That is a defensible financing structure during a build cycle and it is a fragile one if the build slips.

Third, leverage is 4.10x net debt to EBITDA before a $5.5 billion acquisition, roughly $3.5 billion of it cash, on top of $29.4 billion of existing debt.

Fourth, and this is the reason the confidence ratings throughout this dive are low: the file cannot be verified. The SEC extraction for Williams is prose-only — zero table markers in either the 10-K or the 10-Q — so not one financial-statement line could be checked against a filing. The capex figure that drives the negative free-cash-flow conclusion is unverified. The balance sheet that drives the leverage conclusion is unverified. The one cross-check available — the 10-Q's statement that first-half net income rose $455 million — the payload passes exactly, which is encouraging and is one data point. We do not think a Buy should be issued on a file in this condition, and we would say the same if the arithmetic were better.

And the honest counterweight. The street disagrees with us and disagrees confidently: a $83.54 consensus target, 27 buys, 7 holds, ZERO sells among 34 analysts, and a LOW target of $75 that is 4.9% above the current price. Not one covering analyst sees downside. We treat that as a caution rather than a confirmation — unanimity is a risk marker on a levered name — but a reader should know that our $76 is 9% below the street's number and that we are the outlier.

Pre-registered BUY trigger — the specific conditions that would make this Buy — Tactical:

• A price at or below approximately $63 with the FY2027 consensus intact — that is 25.2x FY2027E, roughly 8% below the 200-day moving average, and restores about 2.2:1 asymmetry against the $55 bear and the $96 bull.
• Or, at any reasonable price, confirmation that the first Power Innovation projects have entered service on schedule with the contracted economics disclosed, plus free cash flow turning positive on a trailing basis. That combination removes both the execution risk and the funding risk in one disclosure.

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

• Any Power Innovation project slipping beyond 2028, on turbine availability, permitting or interconnection.
• netDebtToEBITDATTM above 4.5x after the Momentum closing.
• Dividend growth stalling — the clearest signal management could send that the build is not self-funding.
• Momentum closing above $5.5 billion, or with materially more than $2 billion of stock, which at $71.51 is already 2.3% dilution.
• A second consecutive quarter of negative trailing free cash flow with capital expenditure still rising, absent a corresponding third-party funding commitment.
• Any impairment at Momentum or in the Haynesville gathering assets.
• Disclosure that the unnamed investment-grade counterparty has reduced or restructured its 12.5-year commitments.

Where WMB fits in the Synthos Framework Portfolio. No position today. Tracked on the energy-infrastructure / AI-power watch list at the $63 buy trigger above, with a 2% initial size if triggered. Sizing note: the correct way to hold this name, if it is held, is small and early rather than large and late — the payoff is a 2028 commissioning event, the funding risk is 2026-2027, and the dividend pays you 2.94% to wait but is itself the thing at risk if the build slips. Logged as a tracked Synthos call (Watch) as of 2026-08-04 at $71.51, with the fair-value anchors, the buy trigger, the kill criteria and the 6a assumptions all gradeable. Also logged as a forward-dated DATA flag: from the 2026-11-02 balance sheet onward, this vendor's enterprise value for WMB should be treated as understated by approximately 5-6% because of the noncontrolling-interest omission documented in Section 8.

Single biggest risk: a large construction programme funded with debt while free cash flow is negative and the dividend is not covered. Trailing capital expenditure of roughly $6.2 billion against roughly $6.0 billion of operating cash flow, dividends of roughly $2.6 billion, net debt of roughly $29 billion at 4.10x EBITDA, and a $5.5 billion acquisition pending — against a project schedule the 10-K itself hedges with "assuming timely receipt of permits" and a risk list that leads with "unexpected significant increases in capital expenditures or delays in capital project execution". Williams has bought itself two to three years of funding with the July joint venture, and it needs the 1.9 gigawatts to be earning inside that window. If the projects commission on time, the 15.8x EV/EBITDA will look cheap and the street's $83.54 will look low. If they slip, a 0.608-beta infrastructure stock will discover that 4.10x leverage and an uncovered dividend do not have a low beta at all.

Provenance & disclosures

• Traceability: ZERO name-level knowledge-base claims name The Williams Companies out of 52,021 distilled claims. The case-sensitive entity sweep on WMB, Williams and Williams Companies returned nothing — no homograph collision and no coverage (raw entity hits 0, free-text hits 11, name-level 0, discarded 11; breadth 0, claim count 0, net conviction none). US natural gas midstream is a genuine sector void in this store, alongside insurance brokerage, environmental services, diversified industrials and architectural coatings. The 11 free-text hits on "midstream" are all sector-level and carry zero weight in the conviction rating; they are quoted separately in Section 7 because a reader is entitled to know what was searched. The coherent thematic material is one voice, doomberg (skill 1.0, speaker: doomberg/Doomberg, speaker_role: independent), arguing repeatedly for volume-leveraged midstream over price-taking producers — and its single most specific expression, 2026-05-23 conviction 48, names ENERGY TRANSFER rather than Williams as the vehicle for "a US midstreamer leveraged to increasing natural gas volumes with an AI arc and a nice dividend", while noting the speaker does not own it. One independent voice described this exact thesis and chose a competitor; that is recorded and given no weight. Two of the eleven hits ("midstream refining" of rare earths; desalination "midstreamers") are not about energy midstream at all. Under the standing rule that speaker_role is unreliable, the "independent" tags are noted rather than relied upon. No management voice appears. No concentration test is meaningful across zero name-level claims. Quotes are verbatim from the stored claim text.
• Data as-of: fundamentals — quarterly income data through 2026-06-30 and annual balance-sheet and cash-flow data through 2025-12-31, all vendor-sourced and UNVERIFIED for the reason in the next bullet; narrative disclosures (the Power Innovation joint venture, the Momentum acquisition, the segment rename, the 1.9-gigawatt project description, the Socrates detail, the June 2026 dividend rate and the first-half net-income change) filing-verified from the 10-Q filed 2026-08-03 and the 10-K filed 2026-02-24 · estimates 2026-08-04 · prices 2026-08-04, quote timestamp 1785873602 = 2026-08-04T20:00:02Z ($71.51, +1.53%; 50-DMA $73.37; 200-DMA $68.58; RSI 39.7; MACD −0.885; beta 0.608) · knowledge-base claims 2026-08-04. Williams' fiscal year is the calendar year. All figures come from the Synthos vendor data file for WMB or from the SEC filings in the WMB archive; no figure comes from memory, recall or external retrieval.
• FILING EXTRACTION LIMITATION — stated prominently because it caps the confidence of this entire document: the manifest records "tables": false for all six Williams filings, and direct inspection confirms ZERO [TABLE] markers in the 10-K (619,320 characters) and ZERO in the 10-Q (238,802 characters). No income statement, balance sheet or cash-flow statement survived extraction. Targeted searches for the FY2025 capital expenditure ($4,999), operating cash flow ($5,898) and year-end cash ($2,845) return no hits. The capital-expenditure check — the highest-yield check in this programme, and one that fires on roughly four files in ten — therefore COULD NOT BE PERFORMED on Williams. The single cross-check the prose permits is the 10-Q's statement that first-half 2026 net income attributable to Williams rose $455 million against the first half of 2025; the payload's quarterly series gives exactly $455 million ($1,692M against $1,237M), and that is reported as one data point of corroboration, not as verification.
• Filing archive contents: 10-K filed 2026-02-24 (fiscal 2025 — the source of the 1.9-gigawatt project description, the 12.5-year fixed-price contract language, the Socrates detail, the 2026 outlook and the named 2025 in-service projects); 10-Q filed 2026-05-04 (March 2026 quarter); 10-Q filed 2026-08-03 (June 2026 quarter — the source of the four-segment structure including the renamed Transmission, Power & Gulf, the Power Innovation joint-venture note, the Momentum acquisition note, the $0.525 June dividend, the derivative-volatility language and the $455 million first-half statement); 8-K filed 2026-07-01 (Item 5.02 — Robb E. Turner and Lloyd W. Helms, Jr. appointed to the board, expanding it from ten to twelve directors, both independent, with $130,000 cash and $200,000 equity annual retainers); 8-K filed 2026-07-13 (Item 7.01 — a press release and investor presentation concerning "a joint venture financing in support of the Company's announced behind-the-meter Power Innovation projects"; Exhibits 99.1 and 99.2 are NOT in the extracted text); 8-K filed 2026-08-03 (Item 2.02 second-quarter earnings release; Exhibit 99.1 NOT in the extracted text). None carries preserved table data. Because no earnings-release exhibit is present, no management guidance is quoted anywhere in this dive.
• Where the vendor could not be checked, and where it failed on its own internal evidence (detailed in Section 8): the capex check could not be completed for the reason above, though the diagnostic capexToDepreciationTTM of 2.744 trips the suspicion screen while producing an UNFAVOURABLE result — a negative trailing free-cash-flow yield of −0.24% — which inverts the programme's finding that the capex error always flatters, and which the 10-K narrative corroborates ("The projects will require additional capital to fund construction", plus a $5.34 billion third-party funding commitment in July 2026); seg_prod, which drops two of four segments in FY2025 (83.9% of revenue) and over-counts to 136.1% of revenue in FY2024 with no eliminations line — both confirmed defect classes on one name in adjacent years; seg_geo, whose sole entry is {"West": 2847000000} — a segment name substituted for a geography, on a company the 10-Q states operates entirely in the United States; the absence of both July 2026 transactions totalling roughly $10.8 billion from every vendor field, including a profile block still describing the pre-rename "Transmission & Gulf of Mexico" segment; est FY2027 epsLow of MINUS $3.62863, an impossible figure for a fee-based pipeline company, which makes the FY2027 range unusable and forced the bear anchor in this dive to be derived from a multiple instead — a substitution disclosed rather than hidden; est.ebitdaAvg/ebitAvg, rejected for a fixed-ratio fabrication signature (exactly 59.04% and 39.96% of revenueAvg in every year FY2026 to FY2030); enterpriseValueTTM, which omits $2,188M of noncontrolling interests (1.9% of EV) and which will understate enterprise value by approximately 5-6% from the third-quarter balance sheet onward as the July joint venture raises noncontrolling interests toward $6-7.5 billion — registered as a forward-dated data flag; goodwill falling from $466M to ZERO between the 2024 and 2025 balance sheets with no available explanation; and five of eight insider records carrying zero shares, zero price and zero holdings on formType: 4, which is not a filing state that exists.
• Basis note: consensus epsAvg is on an adjusted basis that excludes unrealized commodity-derivative marks, and the divergence is material: in the June 2026 quarter GAAP diluted EPS was $0.68 while the adjusted actual was $0.50 — an $0.18 gap running in the flattering direction. A reader taking the vendor's inc_q at face value would see 51% year-on-year EPS growth in a quarter that was in line with consensus. The 10-Q explains the mechanism directly: "Williams experiences significant earnings volatility from the fair value accounting required for the derivatives used to hedge... the unrealized fair value measurement gains and losses on the derivatives are generally offset by valuation changes in the economic value of the underlying production or transportation and storage capacity contracts, which are not recognized until the underlying transaction occurs." All forward multiples in this dive use epsAvg; the trailing GAAP multiple of 28.4x is shown separately.
• Estimate coverage: 9 analysts on FY2026 EPS, 10 on FY2027 — the anchor for all three fair values — and 8 on FY2028. FY2029 and FY2030 rest on 6 analysts each and are excluded from every conclusion. The FY2027 epsLow is corrupt (−$3.63) and the FY2027 range is therefore not quoted.
• Peer note: the vendor peer set — Canadian Natural Resources, Enterprise Products Partners, Energy Transfer, Kinder Morgan, Marathon Petroleum, MPLX, ONEOK, Petrobras (both share classes) and TC Energy — is materially better constructed than most in this batch, containing five genuine North American gas-midstream comparables (EPD, ET, KMI, MPLX, OKE). It also contains a Canadian E&P, a US refiner and the same Brazilian integrated oil company twice under two tickers. No peer-multiple comparison is drawn, because the July transactions make Williams non-comparable to any of them on a trailing basis. Note that Energy Transfer, a peer in this set, is the name the knowledge base's only relevant claim actually recommends.
• Price-target note: pt reports targetHigh $99, targetLow $75, targetConsensus $83.54 and targetMedian $82 — a genuine four-point distribution, not a WM-class artefact. The notable feature is that the LOW target of $75 is 4.9% ABOVE the spot price of $71.51, and that among 34 rated analysts there are 27 buys, 7 holds and ZERO sells. Not one covering analyst currently sees downside in a company with 4.10x leverage and negative free cash flow. This dive treats that unanimity as a caution rather than a comfort and says so. Separately, quote.yearHigh/yearLow ($80.08/$55.82) disagree with tech.hi52/lo52 ($79.40/$56.51) by roughly 0.9% and 1.2%; tech is used throughout.
• Fair-value caveat: the $55 / $76 / $96 anchors are 22.0x and 30.4x the FY2027 consensus adjusted EPS mean of $2.500 and 31.6x the FY2028 mean of $3.042 respectively. The FY2027 epsLow of −$3.62863 is corrupt and was NOT used; the bear anchor is a multiple applied to the mean instead, and that substitution is disclosed rather than concealed. Cross-checked against EV/EBITDA (approximately 13x, 16.5x and 20x) and the 2.94% dividend yield. Stated arithmetic, not a discounted cash flow. The base is sensitivity-disclosed: 28x gives $70 — essentially spot — and 33x gives $82.50, the street's median. The entire answer lives in a 28-33x band on one estimate row supported by ten analysts.
• Timing: second-quarter 2026 results were released 2026-08-03, one day before this dive, with the 10-Q filed the same day. Adjusted EPS of $0.50 matched the $0.502 estimate to a fifth of a cent; revenue of $3,053M beat the $2,828M estimate by 7.9%; GAAP diluted EPS was $0.68. The next print is 2026-11-02, 90 days away (consensus adjusted EPS $0.57, revenue $2,947M). The Momentum acquisition is expected to close "later this year". The most recent insider filing is dated 2026-08-04 and is one of five empty rows; the last genuine transaction was 2026-08-03, a 2,000-share sale by the general counsel at $70.65.
• 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.