The Williams Companies, Inc. (NYSE: WMB) — The Backlog Got Signed; The Premium Got Richer

Report date: 2026-08-15 · Price reference: $75.20 (close, 2026-08-14) Coverage status: Update — follow-up to my note of 2026-06-14 Sector: Energy · Oil & Gas Midstream

⚡ Claude’s Take

This block is Claude’s own subjective opinion. It is general information, not investment advice, and should not be relied on as such. The analysis in the numbered sections that follow deliberately takes no position and carries no price target — that discipline is absolute everywhere outside this clearly-labeled block.

Verdict: HOLD — direction unchanged, level raised, conviction raised. Accumulate on weakness toward roughly 12.5–13.5x forward EV/Adjusted-EBITDA (~$60–67/share, a ~3.1–3.5% entry yield). Not a short. Conviction: medium-high. Two months ago the bear case had two legs: the multiple was rich, and the growth being paid for was not yet contracted. The second leg has been cut off. Williams delivered Socrates Phase 1 — a 200 MW utility-scale behind-the-meter power plant — on time and on budget in roughly fourteen months from site handover; funded 49% of its five flagship power projects with $5.34 billion of Blackstone-led partner equity at a capped 6.35% cost, taking $900 million of additional consideration in the process; bought Momentum Midstream at ~8.5x forward EBITDA; and, decisively, raised the contracted portion of its 2025–2030 growth algorithm from ~9% to 11%. Leverage went the right way too — 3.67x at quarter end against a June expectation of drift toward 4.1x. Management did, item by item, what it said it would do. My prior report’s single named bull-killer was “an equity raise to fund the capex program”; what actually happened was better in kind, and I should say so plainly.

What has not improved is the price. At $75.20 the stock trades at ~14.9x the 2026 guide and ~14.6x pro forma for Momentum, ~32x adjusted earnings, ~14.6x AFFO, on a 2.79% yield — and at the 91st percentile of its own ten-year composite valuation range, the 99th on price-to-book and the 99th on price-to-sales. It is the second-richest and second-lowest-yielding name in its peer set on any basis I can construct. Run the arithmetic forward and it is unsentimental: delivering the full contracted 11% CAGR gets 2030 Adjusted EBITDA to ~$13.5 billion, against which today’s pro-forma enterprise value is ~9.6x. Exit 2030 at 12x and you earn roughly 8% a year including the dividend; you need the premium multiple to hold at 13x-plus to earn low double digits. That is a fair bet, not a good one. There is also a per-share wedge worth naming: the 11% is a consolidated EBITDA CAGR, but 49% of the flagship power economics now belong to Blackstone, Apollo and KKR — guided 2026 AFFO per share grows 7.7%, not 11%, and the ~$2 billion of stock issued for Momentum adds another ~2.2% of dilution in 2027.

Framing — and this is where my read genuinely changes. In June I called this a crowded yield-plus-momentum re-rate. The factor tape no longer supports the momentum half. Williams’s Momentum loading has fallen (+0.315 → +0.20) while DividendYield has risen to +0.95, with Value at −0.43 and Quality at −0.27, beta 0.49, and a trailing three-month return of −2.5%. The AI-narrative buyer has largely finished; what holds the stock now is low-beta energy-and-dividend carry. That changes the shape of the risk: the de-rate scenario is less “the AI story breaks” and more “long rates back up and a 2.8%-yielding utility-grade cash flow gets repriced against them.” The single piece of evidence that would flip me bullish: a de-rate toward ~13x forward with the contracted book intact, or a 2027 bridge showing AFFO-per-share growth converging on the 11% headline rather than trailing it by three points. The single piece that would flip me bearish: a second power JV priced materially above the 6.35% cost of equity — which would mark the first one as a top-of-market print and re-expose the capex program to the common — or a hyperscaler counterparty restructuring a long-dated behind-the-meter contract. Tag: they earned the premium they were already charging.

🔄 Changes Since the 2026-06-14 Report

Nine weeks; the stock is +5.2% ($71.49 → $75.20); and an unusually eventful quarter. What moved:

Confirmed the thesis (bull-side).

1. Socrates Phase 1 in service, on time and on budget (late July 2026) — 200 MW of utility-scale behind-the-meter generation delivered in ~14 months from site handover. In June the power platform was an unproven capability being paid for as if proven. It is now demonstrated. Phase 2 is on track for Q4-2026.
2. The contracted growth base rose from ~9% to 11%. Management’s 2025–2030 target went from “10%+” to “11%+”, and — the part that matters — the CFO stated the 11% “continues to be centered on our existing contracted book of business,” excluding any further power or pipeline commercialization. The prior memo’s core bear leg, that the premium rested on un-signed optionality, is substantially falsified.
3. The financing question was answered without the common. The Blackstone-led consortium (with Apollo and KKR) committed $5.34 billion — $4.4 billion for 49% of the five power projects’ growth capex plus $900 million of additional consideration to Williams — at a capped 6.35% cost of equity, with Williams retaining operatorship, control and a 2033 buyout option.
4. Leverage improved rather than crept. 3.67x at 2026-06-30 (versus 3.80x a year earlier), guided to ~3.75x normalized. The June report expected drift to 4.0–4.1x.
5. New long-dated, contracted FIDs at scale: Delta Access ($1.5B, 2.25 Bcf/d expandable to 3.5, in service Q1-2029, fully contracted to LNG and power); Shelby Trough Connector (750 MMcf/d expandable to 1.5 Bcf/d, Q2-2028); Transco’s Leidy Access and Garden Connector signed; Power Express upsized to 800 MMcf/d; a Line 200 lateral into Lake Charles.
6. Momentum Midstream acquired for up to $5.5 billion (~$3.5B cash and debt, ~$2B WMB equity) at ~8.5x projected 2027 EBITDA — the cheapest headline multiple Williams has paid this cycle.
7. FY2026 guidance raised $200 million at the midpoint to $8.3–8.5 billion; Adjusted EPS $2.35; AFFO $6,375M / $5.15 per share; coverage 2.47x.

Cut against it (bear-side). 8. The valuation got richer, not cheaper. Own-history percentiles moved up across the board: composite 88.7 → 91.3, P/E 70.9 → 75.6, P/B 97.4 → 99.1, P/S 97.8 → 99.3. 9. The quarter itself decelerated and missed. Q2 Adjusted EBITDA +6% (versus +13% in Q1, +10% year-to-date); Adjusted EPS $0.50 against a $0.52 consensus; and operating cash flow fell $74 million year over year on Transco rate-refund payments. The +51% GAAP EPS headline is flattered by a $126M Brazos Permian II disposal gain and a +$106M derivative swing, both excluded from the adjusted figures. 10. Growth capex was raised to $7.3–7.9 billion and first-half capital investment nearly doubled year over year ($3,284M vs $1,709M). Free cash flow after capex and dividends was roughly −$1.6 billion in the half, debt-funded. 11. Common equity was issued after all — ~$2 billion of stock as Momentum consideration, ~2.2% dilution, released over 180 days with a volume-based trading restriction rather than a hard lock-up. 12. A structural dilution of the headline metric. 49% of the five flagship power projects now belongs to third parties, so the consolidated 11%+ EBITDA CAGR increasingly overstates what accrues per share: guided 2026 AFFO/share growth is 7.7%.

A correction to the prior report. The 2026-06-14 memo quoted an enterprise value of “~$104.9B” and derived ~13.5x trailing / ~12.8x forward EV/Adjusted-EBITDA. That figure was ROIC.ai’s FY2025 annual enterprise value, struck at the 2025-12-31 close of $59.24 — while the memo described the stock at ~$72. Recomputed at that report’s own price, enterprise value was ~$119.0 billion and the multiples were ~14.8x trailing and ~14.5x forward — roughly 1.3 to 1.7 turns richer than stated. The prior report’s direction was right and its conclusion is unaffected; its stated multiple was too low. Every multiple in this report is struck at the current price on the Q2-2026 balance sheet.

Scoring the prior memo’s falsification tests (the what-must-be-true section of the 2026-06-14 report). The bull case was to be falsified by project slippage (not triggered — Socrates delivered), an equity raise to fund capex (not triggered in kind — capex was funded with 6.35% partner equity; the common issued was acquisition currency, with a ~2.2% asterisk), or a CAGR guided below 9% (falsified in the opposite direction — guided up to 11%). The bear case was to be falsified by new long-dated FIDs at advertised multiples (hit), deleveraging below 3.8x (hit at 3.67x), and the multiple holding through a drawdown (partially — the stock fell 11% to $70.13 on 2026-07-29 and recovered, but that was an idiosyncratic pullback, not a market-wide test). On the operating scorecard the bear case lost. What survives of it is valuation, and only valuation.

📈 Stock Price Action — Five-Year Event Map

Williams has compounded from roughly $19 in August 2021 to $75.20 today — very nearly a four-bagger, with the entire step-change concentrated in the eighteen months from late 2024. The all-time high of $78.82 was set on 2026-05-19; the stock sits 4.6% below it, inside a 52-week range of $54.78–$78.82, above its 21-, 50- and 200-day exponential moving averages, with a published beta of 0.49. The worst five-year drawdown was −23.0% into September 2022. Price moves below are FACT; the attributed drivers are INTERPRETATION.

Cycle narrative. (1) The 2021–22 leg was a commodity-and-geopolitics repricing: US gas infrastructure was suddenly strategic, and Williams — sitting on the corridor from the Gulf to the eastern seaboard — was a direct beneficiary. (2) The 2022 drawdown was the mirror image: when the gas spike unwound, so did the multiple; a −23% fall in a business whose contracted cash flows barely moved is a useful reminder that this equity carries multiple risk, not cash-flow risk. (3) 2023 was the dead year — gas at cycle lows, the stock range-bound while management bolted on MountainWest and the DJ Basin systems. (4) 2024 is when the story changed: data-center power demand became the marginal driver of US gas consumption and Williams re-rated 62% in twelve months. (5) The two large 2025 single-day drops are the cleanest evidence available that the market treats WMB as an AI-adjacent security — a Chinese model-efficiency paper knocked 8.4% off an interstate pipeline in one session. (6) The 2026 run to the all-time high was driven by the February Analyst Day, which put a number (10%+) on the algorithm, and a Q1 print that beat it. (7) The May–July fade was valuation, not news; the stock reached the 99th percentile of its own price-to-sales history and stalled. (8) The recovery since late July is the market marking to the evidence — the JV, the completed power plant, the raised contracted CAGR — though notably the market did not pay up on the print day itself (3 August closed at $70.43); the re-rate took two weeks to arrive.

1. Executive Summary

The Williams Companies is a ~$92 billion-equity, ~$125 billion-enterprise-value US natural-gas midstream C-corporation built around Transco, the largest-volume interstate gas pipeline in the country, which moves roughly 16% of all US natural gas from the Gulf up the eastern seaboard. Cash flow is high-90s-percent fee-based and rate-regulated. FY2026 guidance is for $8.3–8.5 billion of Adjusted EBITDA, $2.35 of Adjusted EPS, and $6,375 million of available funds from operations — $5.15 per share, covering the $2.10 dividend 2.47 times.

The quarter changed the character of the investment case. Nine weeks ago the debate was whether Williams could convert a 6 GW behind-the-meter power backlog into signed, financed, built projects at advertised returns. It has now: Socrates Phase 1 delivered 200 MW on time and on budget; a Blackstone-led consortium funded 49% of the five-project program with $5.34 billion at a capped 6.35% cost of equity while Williams kept operatorship and took $900 million of additional consideration; Momentum Midstream was acquired at ~8.5x forward EBITDA to consolidate the Haynesville–to–Gulf Coast corridor; two large contracted pipelines (Delta Access, Shelby Trough Connector) were sanctioned; and the contracted portion of the 2025–2030 growth algorithm rose from ~9% to 11%. Leverage fell to 3.67x. On the operating scorecard this was close to a perfect quarter for the bull case.

The bear case has correspondingly narrowed — to valuation, where it has strengthened. At $75.20 Williams trades at ~14.9x the 2026 Adjusted-EBITDA guide (~14.6x pro forma for Momentum), ~17.7x trailing GAAP EBITDA, ~32x adjusted earnings and ~14.6x AFFO, on a 2.79% yield. Against its own ten-year history that is the 91st percentile on the composite measure, the 99th on price-to-book and the 99th on price-to-sales — richer on every one of those than at the June report. It is the second-most-expensive and second-lowest-yielding name in a peer set (KMI 13.6x/3.6%, TRGP 14.0x/1.6%, OKE 11.5x/4.5%, MPLX 11.3x/7.2%, ET 8.9x/6.4%) in which every member is trading near its own historical high on the same LNG-plus-AI-power thesis. Meanwhile the underlying economics remain utility-grade: consolidated return on invested capital was 7.44% in FY2025, and the growth is capital-hungry — first-half capital investment nearly doubled to $3.3 billion, free cash flow after capex and dividends was roughly −$1.6 billion, and full-year growth capex is guided to $7.3–7.9 billion.

Two second-order issues deserve naming. First, the quarter itself decelerated: Adjusted EBITDA +6% against +13% in Q1, an Adjusted-EPS miss versus consensus, and operating cash flow down year over year. Second, and more durable, the headline growth metric is becoming a less complete description of shareholder economics — 49% of the flagship power projects now belongs to third parties, and ~$2 billion of stock was issued for Momentum, so guided 2026 AFFO per share grows 7.7% against an 11%+ consolidated EBITDA target.

Synthesis. The risk in owning Williams has migrated from delivery risk to multiple risk. That is a better risk to be taking, and it is why the growth-adjusted valuation is arguably more attractive today than in June even though every percentile is higher. But it is still a risk that is being taken at the top of the stock’s own history. Delivering the full contracted 11% CAGR implies ~$13.5 billion of 2030 Adjusted EBITDA, against which today’s pro-forma enterprise value is ~9.6x — an exit at 12x produces roughly an 8% annual total return. The franchise is excellent, the execution has been proven, and the price already reflects both.

2. Business Overview

What Williams does (FACT). Founded in Tulsa in 1908, Williams gathers, processes, transports, stores and markets natural gas and natural gas liquids across roughly 33,000 miles of pipeline, 29 processing plants, seven fractionators and ~23 million barrels of NGL storage. Since the 2018 roll-up of Williams Partners it has been a single C-corporation issuing a 1099 rather than a K-1 — a structural advantage over MLP peers such as Energy Transfer and MPLX that materially widens its investable audience and is one reason it commands a premium multiple within its own comp set.

A renamed segment tells the story of the year (FACT). With the second-quarter 2026 report, the segment formerly called “Transmission & Gulf of Mexico” became “Transmission, Power & Gulf.” Power is now a named reporting pillar rather than a venture inside the pipeline segment. The four reportable segments and their first-half 2026 Adjusted EBITDA:

The crown jewel is unchanged (FACT). Transco is a ~10,000-mile interstate system running from South Texas and the Gulf Coast up the eastern seaboard to New York, carrying roughly 16% of all US natural gas consumption. It connects the lowest-cost supply in North America to the three highest-value demand sinks: the Southeast power-and-population corridor (now adding data-center load in Virginia and the Carolinas), the heating-constrained Northeast, and the Gulf Coast LNG-export terminals. As a FERC-certificated route monopoly it earns a regulated cost-of-service return on a large and growing rate base, and — critically — it can be expanded incrementally by looping and adding compression along an existing right-of-way at a fraction of the cost and permitting risk of a greenfield line.

What is genuinely new: a third leg (FACT/INTERPRETATION). Williams is no longer only a gatherer-and-transporter. It now builds and operates dedicated on-site generation for data centers, and as of late July 2026 it has done so at utility scale. The Power Innovation platform has five projects underway; Socrates Phase 1 (200 MW) is in service with Phase 2 due in Q4-2026; and 49% of the program’s economics have been sold to a Blackstone-led consortium. Management describes the strategic destination directly: “through the end of the decade and beyond, us continuing to grow pipeline business at a very healthy clip. The power business emerging and growing alongside it… Gathering and processing will shrink… relative to those other two.” That is a deliberate mix shift toward the two highest-barrier, longest-contracted parts of the business, and away from the most competitive part.

Revenue model (FACT/INTERPRETATION). The interstate pipelines recover the majority of their cost of service through firm capacity-reservation charges — customers pay for the right to move gas whether or not they flow it. The gathering and processing book runs on multi-year fee-based contracts, many with minimum-volume commitments. The behind-the-meter power projects are long-dated take-or-pay-equivalent contracts (Neo is 12.5 years). Management characterises the aggregate book as high-90s-percent fee-based; the only material direct commodity exposure is the ~5%-of-EBITDA marketing segment, plus a thin keep-whole sliver in the West and the residual upstream in Other. The Q2 numbers illustrate exactly this: Gas & NGL Marketing Services swung from −$30M to +$123M of Modified EBITDA on $124 million of unrealized derivative marks, while contributing −$1M of Adjusted EBITDA. All the noise, none of the substance.

Momentum, once closed, extends the same model (FACT). The acquisition adds more than 4,000 miles of pipe, over one million dedicated acres across four gathering areas with 6 Bcf/d of combined capacity, multiple processing and treating facilities, and three take-or-pay pipelines totalling 4.05 Bcf/d. Management describes the mix as “a balanced mix of gathering and take-or-pay earnings that fits squarely within the fairway of our existing core business mix” — i.e., it does not change the character of the company, it deepens it in the one basin (Haynesville) most likely to supply the one demand corridor (Gulf Coast LNG) Williams already serves.

Verdict. A high-quality, recurring-revenue infrastructure business with a clean C-corp structure, an irreplaceable core asset, and — new since the last report — a demonstrated third leg in dedicated power generation. The cash flows are about as predictable as energy gets, and the mix is shifting toward the more defensible end of the value chain. The quality is high; the return ceiling is still set by regulation and by the cost of the capital being deployed, a tension developed in the competitive-position section and the financial-quality section.

3. Industry Dynamics

A two-tier industry, now with a third tier emerging (FACT/INTERPRETATION). US natural-gas midstream remains two businesses bolted together. Interstate FERC pipelines are rate-regulated quasi-utilities: allowed cost-of-service returns on rate base, near-zero market-share volatility on certificated routes, demand-charge revenue insensitive to volume. Structurally excellent — stable, defensive, protected. Gathering and processing is competitive basin-by-basin, volume- and commodity-cyclical, with lower barriers to entry; structurally average, with returns that rise and fall with drilling activity. Williams is roughly 47% weighted to the good tier and 43% to the contracted G&P book. The third tier — dedicated behind-the-meter generation for hyperscalers — is genuinely new, and its structural attractiveness is not yet established: barriers are currently execution and equipment access rather than regulation, which is a weaker and more contestable form of protection than a FERC certificate.

Demand backdrop — quantified, and still the best in energy (FACT). Three drivers converge on the gas network at once, and management put numbers on two of them this quarter. LNG exports: US liquefaction runs at “about 18 Bcf a day” today with forecast models showing “above 40 [Bcf/d] in [the] next 10 years”; the earnings release cites Gulf Coast LNG demand rising ~20 Bcf/d over ten years. Supply to serve it: the Haynesville “is expected to grow by over 10 Bcf per day” over the same decade — which is precisely why Williams bought Momentum. Gas-fired power for AI and electrification: the marginal new demand source, now visible in Williams’s own contracts rather than only in third-party forecasts. Coal-to-gas switching remains a steadier baseline tailwind. This is a strong, corroborated demand picture; the honest caveat is that LNG demand is contracted under 15–20-year offtake agreements (durable) while data-center demand depends on the AI-capex cycle continuing (extrapolative), and the bull case routinely blurs the two.

How FERC rate-making protects and caps (FACT/INTERPRETATION). Interstate rates are set on a cost-of-service basis: recovery of operating costs, depreciation and taxes plus a regulated return on rate base at an allowed equity return typically in the low-to-mid teens. This bargain is double-edged. It is the source of the moat’s stability — demand-charge revenue is contractual and the return is protected from competition. It is also precisely why consolidated ROIC is only ~7.4%: regulators are statutorily required to allow a “just and reasonable” return, not an excess one, so a functional monopoly cannot earn monopoly rents. Rate cases can reset allowed returns — and did, visibly, this quarter: Transco’s April 2026 rate refunds were the specific cause of the year-over-year decline in second-quarter operating cash flow. Regulation is not an abstraction here; it shows up in the cash flow statement.

The capital cycle: the permitting moat holds, the sector-wide race is the risk (FACT→INTERPRETATION, Marathon lens). The single most important structural fact remains that new interstate pipeline construction has become extremely difficult. A FERC the capital-allocation section certificate plus multi-year permitting, rights-of-way, eminent domain and near-certain litigation stand between capital and a new greenfield line; Mountain Valley required six-plus years and an act of Congress, while Constitution and Atlantic Coast were cancelled after large sunk costs. The binding constraint is permitting, not money — so capital cannot flood in to compete away the returns of incumbents holding certificated corridors. Larry Larsen’s comment this quarter that the larger Transco opportunities are “more in the 2030-plus time frame… trying to line-out the regulatory certainty and timing for that is what’s taking a little bit of time” is the moat and its cost in one sentence: the permitting friction that protects Williams also slows Williams.

The Marathon caution has, if anything, sharpened. Reading across my own recent notes on Kinder Morgan, ONEOK, Enterprise Products, MPLX and Cheniere, all written within the last eight weeks, every single one is trading at or near the top of its own historical valuation range, on the same LNG-plus-AI-power thesis, while simultaneously accelerating capital deployment into it. That is the textbook set-up the capital cycle warns about: an entire industry investing into the same exciting end-market at once, with the supply arriving several years later compressing the returns that justified the investment. Williams’s existing corridors are protected from that; its new capital — $7.3–7.9 billion of it in 2026 alone — is not exempt. The one genuine mitigant is that a large share of Williams’s new capital is contracted before it is spent, which is a materially better position than building on spec.

Behind-the-meter power specifically deserves capital-cycle scrutiny (INTERPRETATION). Williams’s ~5x build multiple on power projects implies roughly a 20% gross EBITDA yield on invested capital. Returns like that, in a business whose barrier is turbine access and construction capability rather than a government licence, attract competition — and the competitors are not just other midstream companies but independent power producers, utilities and the hyperscalers themselves. Larsen’s own framing acknowledges the current constraint: “it’s not only just the turbines. We’ve highlighted that we’ve locked up the turbines to be able to support our backlog, but it’s the rest of the balance of plant.” Locked-up turbines are a real, and temporary, advantage. The durable question is whether the returns survive the arrival of the capital that a 20% EBITDA yield will inevitably summon.

Competitive set (FACT). The relevant peers are Kinder Morgan (the closest pure gas-pipeline-scale FERC comparable), Energy Transfer (larger, more diversified into crude and NGLs, K-1), ONEOK (NGL-centric), Enbridge (larger, slower), MPLX, Targa and DT Midstream. FactorsToday’s factor-similarity screen independently returns the same set — KMI 0.96, Antero Midstream 0.93, DTM 0.92, OKE 0.89, ET 0.88, Cheniere 0.87, MPLX 0.87, TRGP 0.87 — which is a useful confirmation that the comp set is not an analyst’s construction.

Verdict: structurally GOOD for the regulated tier, AVERAGE for G&P, and UNPROVEN for power. The interstate business Williams over-weights is one of the better structures in energy: permitting-protected, demand-charge-driven, route-monopoly, into a secularly growing end market. The G&P book is solid but ordinary, and Momentum deepens it. The power tier is the exciting one and the least structurally protected one — the current advantage there is execution and equipment access, which are real but replicable. “Favorable industry” here still means durable and defensive, not high-return; the regulation that protects the franchise also caps it.

4. Competitive Position

The moat, named (Greenwald taxonomy). Transco combines the strongest available advantage types: a government-licence / regulatory franchise (the FERC certificate), an irreplaceable physical right-of-way, economies of scale on a fixed network, and customer captivity (firm-reservation contracts at delivery points with no alternative route to market). You cannot build a competing line from the Gulf to the New York City load centre — not for permitting reasons, not for right-of-way reasons, not at any price. The customers at the end of the pipe are physically captive.

The share-stability test: PASS for Transco, WEAKER elsewhere. Greenwald’s empirical test of a moat is whether market shares are stable. Transco’s share of its served markets is fixed by physics and regulation; it does not lose volume to entrants because entrants cannot exist. The G&P business is contestable basin-by-basin, so the test is weaker there — which is precisely why consolidating the Haynesville via Momentum matters strategically: it converts a contested position into a dominant one. Management’s claim after the deal is that Williams becomes “the largest gatherer of Haynesville gas connected to Transco, the largest gas transmission pipeline system along the Gulf Coast corridor.” If that holds, it is a genuine local-scale advantage of the kind Greenwald favours — dominance in a defined geography beats scale in the abstract.

Now tie the moat to a financial outcome — the skeptical caveat, updated (FACT/INTERPRETATION). A moat that does not show up in returns is not a moat. Here it shows up in stability more than in level. The proof it is a moat: Transmission, Power & Gulf Adjusted EBITDA rose from $1,765M to $1,969M year over year, and has risen straight through gas-price cycles for a decade. The proof it is a utility-grade moat rather than a high-return compounding one: consolidated ROIC was 7.44% in FY2025 (ROIC.ai; 6.31% in 2024, 8.50% in 2023, 6.72% in 2022, 5.30% in 2021), even though return on capital reads 8.3% and ROE 20.8%. The gap between the levered headline and the unlevered economic return is the tell.

The correct characterisation is therefore unchanged from June, and it is the load-bearing judgment of this report: Williams owns a wide, durable moat that protects its cash flows and lets it grow cheaply along existing corridors — but that moat secures low-volatility, high-single-digit unlevered returns; it does not generate the high-ROIC compounding that a “wide-moat” label often implies. For context from my own recent coverage, that ~7.4% sits between Kinder Morgan (~5.8%, below its cost of capital) and Enterprise Products (~10.8%, clearly above), and roughly level with ONEOK (~8%). Williams is a mid-pack earner of capital in a group where nobody earns a lot.

What the power business does to the moat question (INTERPRETATION). The Power Innovation platform’s ~5x build multiples imply a ~20% gross EBITDA yield — roughly 2.7x the consolidated ROIC. If that scales, Williams’s blended return on incremental capital rises materially above its historical base, which would over time repair the ROIC criticism. Two things stand between here and there. First, the Blackstone JV sells 49% of the economics of the current five projects, which caps how much of that high return accrues to Williams shareholders on this tranche (though the $900 million of additional consideration is compensation for exactly that, and management argues it improves the cash-flow-to-invested-capital ratio ~56% over the primary term — an unaudited management figure, treated as hypothesis). Second, high returns in a non-licence-protected business are the thing capital hunts. The honest read: the power platform is the first credible path in a decade to lifting Williams’s return on incremental capital above utility levels, and it is also the part of the business with the least durable protection.

Network effects and switching costs (INTERPRETATION). Network effects are weak and routinely overstated for pipelines — the value is the regulated route monopoly and scale economics, not Metcalfe-style value that grows with each node. Switching costs are high and physical: a captive delivery point cannot be re-routed; a basin’s gathering system cannot be moved. They are strongest on Transco (no alternative route), moderate in G&P (minimum-volume commitments), high but time-limited in power (12.5-year contracts with a defined term), and weak in marketing.

Versus peers (FACT/INTERPRETATION). Against KMI (similar enterprise value, similar leverage, ~4% organic growth, 3.6% yield, ~5.8% ROIC), OKE (11.5x, 4.5% yield, NGL-levered), ET (8.9x, 6.4% yield, K-1, more commodity exposure) and MPLX (11.3x, 7.2% yield, sponsor-controlled), Williams’s edge is asset location and the credibility of its growth runway. The pattern is unchanged and internally logical: Williams trades at the top of the group on multiple and at the bottom on yield because it has the best asset and the fastest contracted growth. What changed this quarter is that the growth claim is far better evidenced than it was — which strengthens the case for a premium while doing nothing to establish the size of premium that is appropriate.

Verdict: a durable but utility-grade moat, now with a credible high-return extension attached. The competitive position is excellent and defensible, and the Momentum deal genuinely strengthens the weaker (G&P) half of it. The error to avoid remains conflating “best-positioned asset in the sector” with “high return on capital.” The first is true; the second is not yet, and the part of the business that could make it true is the part that is least protected.

5. Growth History and Forward Opportunities

History (FACT). Reported revenue is the wrong metric — it includes commodity pass-throughs and the volatile marketing line, and has been range-bound ($10.6B in 2021 → $12.0B in 2025). Adjusted EBITDA is the right lens and has compounded steadily: ~$5.0B (2021) → ~$6.4B (2022) → ~$6.8B (2023) → ~$7.1B (2024) → $7.75B (2025), and is tracking to $8.3–8.5B in 2026. First-half 2026 Adjusted EBITDA of $4,175M was up 10.0%. Diluted share count has been essentially flat over five years (1,218M → 1,225M) with no buybacks and, until now, no meaningful issuance.

The forward algorithm, and how it was rebuilt this quarter (FACT). The disclosure trail is unusually clean and worth laying out precisely, because the contracted number is the whole argument:

The CFO’s words: “after layering in the Momentum transaction as well as the other projects we’ve announced today, we feel confident in moving our target up to 11% plus… we’re really talking about a number here that continues to be centered on our existing contracted book of business. And so we’re excluding the commercialization of any additional power or pipes projects.” Treat that as management commentary requiring validation — but note that it is validated, in part, by the signed contracts and announced FIDs behind it, and by the fact that the same management team has raised this number twice in six months rather than defending a stale one.

The growth drivers (FACT):

• Power / data centers. Five behind-the-meter projects underway against a ~6 GW backlog at a stated ~5x build multiple. Socrates Phase 1 (200 MW) in service late July 2026 on time and on budget, ~14 months from site handover; Phase 2 due Q4-2026. Neo remains the largest project ever undertaken — 682 MW, 12.5-year contract, ~$2.3 billion, in service H2-2028. Zamarin says the pipeline “is supported by multiple customers beyond just our first primary customer” and that additional projects will be commercialised “between now and the end of the year.”
• Transco transmission. Southeast Supply Enhancement under construction with possible early in-service at the start of 2027 and full in-service targeted Q3-2027; NESE compression beginning late 2026 with the offshore build mostly in 2027; Power Express upsized to 800 MMcf/d into Virginia; Leidy Access and Garden Connector signed for Pennsylvania and New Jersey; Delta Access newly sanctioned at $1.5 billion for 2.25 Bcf/d (expandable to 3.5 Bcf/d) in service Q1-2029, fully contracted to LNG and power.
• Haynesville and LNG feedgas. Louisiana Energy Gateway in service since Q3-2025 and already visible in West segment growth; Line 200 (3.1 Bcf/d, 20-year take-or-pay to Woodside) plus a newly commercialised lateral into Lake Charles power demand; Shelby Trough Connector at 750 MMcf/d expandable to 1.5 Bcf/d, in service Q2-2028; a 10% interest in the fully-contracted Louisiana LNG terminal.
• Momentum. 6 Bcf/d of gathering capacity and 4.05 Bcf/d of take-or-pay pipeline capacity in the basin management calls “the single most important U.S. supply basin in answering the near-term call for natural gas.”

The build-multiple economics, re-examined (INTERPRETATION). The ~5x build multiple — $5 of capital generating $1 of incremental annual EBITDA, a ~20% gross EBITDA yield — remains the most important number in the growth story, and it is now better evidenced than in June because one project has actually been delivered on budget. Three caveats still temper it. First, ~20% is a gross EBITDA yield, not a return on capital: after maintenance capex, taxes and the cost of the funding, the levered equity return is lower. Second, the JV means Williams now keeps 51% of the equity economics on the current five projects (compensated by the $900 million). Third, build multiples historically rise as the best-sited projects are done first — Socrates was, by construction, the easiest one.

Funding, transformed (FACT/INTERPRETATION). This is where the year’s real change sits. FY2026 growth capex of $7.3–7.9 billion cannot be self-funded against ~$6.7 billion of operating cash flow net of working capital and $2.6 billion of dividends. In June the funding plan was a stated intention to bring in project-level equity partners. It is now executed: $5.34 billion committed at a capped 6.35% cost of equity for 49% of the five power projects, plus $900 million of additional consideration, with Williams retaining operatorship and a 2033 buyout option at the remaining partner balance. Porter says the process is repeatable and faster next time, though “it’s probable in my mind that each one of these deals could perhaps be a unique separate partnership.” Combined with 3.75x normalised leverage against a 4.0x internal ceiling, management says it has “in excess of another $2 billion of incremental capacity” for near-term power projects before needing another partner.

Quality of the growth (INTERPRETATION). Materially higher than in June. The contracted base now covers the full 11%, meaning an investor is no longer paying for optionality to be won — only for it to be built and collected. That is a real reduction in risk. Two honest deductions remain. First, the growth is being bought with a lot of capital and some of it with other people’s equity, so the per-share translation is weaker than the headline: 2026 AFFO per share of $5.15 versus ~$4.78 in 2025 is +7.7%, three points below the consolidated target, and the ~$2 billion Momentum share issuance widens that gap in 2027. Second, the 2027 capital year is the pinch point — Chad Zamarin describes 2027 as “a bit of a high watermark from a capital perspective relative to earnings before the growth really kicks in, in '28.”

Verdict: high-quality and, for the first time, substantially de-risked. The 11%+ CAGR is now the most credible growth algorithm in North American midstream because the great majority of it is contracted rather than hoped for. The remaining question is not whether the EBITDA arrives but how much of it arrives per share — and on that measure the growth is high-single-digit, not low-double-digit.

6. Financial Quality

The Adjusted-EBITDA bridge — get the right denominator (FACT). Three “EBITDA” figures must not be conflated, because the valuation multiple swings materially on which is used. On the FY2026 guidance midpoint the architecture is: Modified EBITDA $8,595M less EBITDA adjustments $(195)M = Adjusted EBITDA $8,400M. That Modified figure itself includes $990M of proportional Modified EBITDA from equity-method investees — cash that arrives as distributions, not as 100%-controlled cash flow. On a trailing basis, ROIC.ai computes GAAP EBITDA of $6,993M against trailing Adjusted EBITDA of $8,128M (FY2025 $7,750M plus first-half 2026 $4,175M less first-half 2025 $3,797M). The ~16% gap between the two is real and is mostly the JV proportional consolidation. For valuation, the Adjusted denominator is defensible and is what the market uses — but an investor should know that roughly a billion dollars of it is not consolidated cash, and that the truth about the multiple sits between 14.9x and 17.7x.

Earnings quality (FACT). GAAP earnings for Williams are genuinely noisy, and this quarter was a textbook case. Q2 GAAP net income to common was $827M / $0.68 per diluted share, +51% year over year; Adjusted net income was $614M / $0.50, +8%. The GAAP figure was flattered by a $126 million gain on the June-2026 sale of the Brazos Permian II equity-method investment and a +$106 million favourable swing in unrealized commodity-derivative marks; the year-to-date GAAP figure additionally carries a $194 million gain on the January-2026 South Mansfield upstream sale. All three are excluded from Adjusted EBITDA and Adjusted EPS. Normalise to the adjusted figures and the quarter reads: Adjusted EBITDA +6% and an Adjusted EPS of $0.50 against a $0.52 consensus — a miss, and a deceleration from Q1’s +13%. The full-year picture is better (+10% year to date) and management explains the quarterly pattern as seasonality in the marketing book, but the honest one-line summary of Q2 is that the operating result decelerated while the reported result accelerated.

Free-cash-flow reality (FACT). Williams is mid-build and this is the number to be clear-eyed about. First-half 2026: operating cash flow $2,979M; capital expenditures $3,193M; common dividends $1,284M; distributions to noncontrolling interests $140M. Free cash flow after capex and dividends was roughly −$1,638 million in the half, funded by $1,671 million of net new long-term debt. Capital investments excluding acquisitions nearly doubled year over year, from $1,709M to $3,284M. Full-year growth capex is guided to $7.3–7.9 billion. In plain terms: the dividend and the growth program together consume more than the business generates, and the difference is being financed.

But is the dividend covered? Yes, on the metric that fits the business (FACT). On AFFO — Williams’s distributable-cash analogue, stated after preferred dividends and net noncontrolling-interest distributions — first-half AFFO was $3,220M against $1,284M of dividends, 2.51x coverage (2.26x in the second quarter alone; 2.47x guided for the full year against a 2.36–2.45x guide entering the year). The funding gap is discretionary, contracted growth capex, not maintenance spending: Williams could stop growing and cover the dividend comfortably. That is the correct way to read a midstream balance sheet. The honest risk is that it only holds if the contracted backlog performs, and there is no slack for a large project to disappoint.

A quality-of-earnings flag that is new this period (INTERPRETATION). Second-quarter operating cash flow fell $74 million year over year, to $1,376 million, while Adjusted EBITDA rose 6%. The company attributes this to “unfavorable net changes in working capital driven by the payment of Transco’s rate refunds in April 2026.” That explanation is credible and the item is genuinely one-off. But it is worth recording that a regulated rate proceeding produced a real cash outflow in the same quarter that the company raised its long-term growth target — a useful reminder that the regulatory bargain has a cost side as well as a protection side, and that cash and Adjusted EBITDA can diverge for perfectly legitimate reasons.

Margin and return trajectory (FACT). The multi-year trend supports “economics improve with scale.” EBITDA margin on GAAP revenue rose from 42.1% (2021) to 54.8% (2025); operating margin from 24.8% to 36.9%. Return on equity reads 20.8% (2025) and return on capital 8.3%. The cleaner economic measure, return on invested capital, was 7.44% in FY2025 and has ranged 4.2%–8.5% over 2020–2025. The gap between the levered headline and the unlevered return is the tell, and it has not closed: this is a business whose reported profitability looks strong because of leverage and because of what the return-on-capital denominator excludes.

Balance sheet and leverage — better than expected (FACT). At 2026-06-30: commercial paper $475M, current long-term debt $2,197M, long-term debt $28,121M — total debt $30,793M; cash $203M; noncontrolling interests $2,178M; preferred a trivial $35M; total stockholders’ equity $13,190M against a retained deficit of −$11,834M (improving from −$12,237M), which is a legacy artefact of decades-old write-downs and the reason the price-to-book percentile reads so extreme. Company-defined leverage was 3.67x at quarter end, down from 3.80x a year earlier, against a 3.5–4.0x target band and an internal 4.0x ceiling. Guidance is ~3.75x on a normalised full-year Momentum basis (~3.9x at year end on ~3 months of Momentum). Ratings were investment grade entering the period (S&P BBB+ stable, Moody’s Baa2 positive, Fitch BBB positive), and debt is ~99% fixed-rate and well-laddered — which insulates the cash flow from rate moves while leaving the equity multiple exposed to them.

Verdict: high-quality cash generation; “do the economics improve with scale?” answered yes — with two honest caveats. Adjusted EBITDA +10% year to date, AFFO +17%, expanding margins, a dividend covered 2.5 times, leverage moving down not up, and a fixed-rate investment-grade balance sheet. The caveats: this remains a debt-and-partner-funded growth machine with free cash flow after growth capex structurally negative in the build phase and ~$1 billion of headline EBITDA sitting in unconsolidated joint ventures; and the unlevered economic return on the whole capital base is still only ~7.4%.

7. Capital Allocation

The defining decision of the period: how the power capex got funded (FACT/INTERPRETATION). The Blackstone-led joint venture is the most consequential capital-allocation act Williams has taken in years, and it deserves to be assessed on its merits rather than cheered. The terms: $5.34 billion of committed capital, comprising $4.4 billion for 49% of expected total growth capital expenditure on the five current Power Innovation projects, plus $900 million of additional consideration paid to Williams, at a capped 6.35% cost of equity, with Williams retaining operatorship, key decision-making authority and upside participation, and holding a buyout option at the remaining partner investment balance beginning in 2033. Reuters independently confirms the $5.34 billion and the 49% non-controlling stake. Management’s claim is that the ratio of total Williams cash flow to total invested capital across the five projects improves ~56% over the primary contract term — an unaudited management calculation, treated here as hypothesis rather than evidence.

The favourable read is straightforward and largely correct: a 6.35% capped cost of equity is cheap relative to Williams’s own cost of common equity, the $900 million is genuine value transfer toward Williams, control is retained, and the buyout option preserves the long-dated upside. The skeptical read is equally straightforward: 49% of the economics of the platform that justifies the stock’s premium multiple now belongs to somebody else, permanently unless bought back in 2033 or later; the $900 million is best understood as the price of that; and the fact that a competitive process attracted Blackstone, Apollo and KKR simultaneously is evidence that private capital regards these assets as attractively priced. Porter’s own framing — “we would not have done it at all if not for the expectations we have in driving substantial additional growth” — is candid: this is a funding vehicle to enable more projects, not a value-realisation event.

Was equity issued? Yes — but not the kind the prior report warned about (FACT/INTERPRETATION). The June report named “an equity raise to fund the capex program” as the single development that would most damage the bull thesis. What happened: the capex was funded with partner equity at 6.35%, while ~$2 billion of Williams common stock was issued as acquisition consideration for Momentum — roughly 26–27 million shares at ~$75, about 2.2% dilution on 1,223 million shares outstanding. Guidance already carries 1,237 million weighted diluted shares for 2026 (up ~1% from 1,225 million in 2025), with the Momentum shares landing largely in 2027. Rob Wingo describes the mechanics: “instead of a traditional lockup, we’ll be releasing the shares over a 180-day period. And then once those shares are released, we’ll have a trading restriction that basically will limit their trading to a small percentage of our average daily trading volume.” The honest score: the named bull-killer was not triggered — this is acquisition currency for an 8.5x asset, in a transaction that reduces leverage — but shareholders were diluted, and the report should say so rather than wave it away.

M&A scorecard, updated (FACT/INTERPRETATION, Marathon lens). Williams runs a steady cadence of fee-based, basin-consolidating or Transco-adjacent bolt-ons. The record: MountainWest (2023, ~$1.5B), DJ Basin Cureton and Rimrock (2023, ~$1.27B), Gulf Coast Storage / Hartree (2024, ~$1.95B at roughly 12x trailing), Discovery and Crowheart (2024), Cogentrix power (2024, ~$1.9B — subsequently divested), the 2025 “wellhead-to-water” restructuring that monetised Haynesville upstream into JVs with Woodside and JERA while retaining the midstream fee stream, and now Momentum at ~8.5x projected 2027 EBITDA (~9x net of the NG3 noncontrolling interest). Momentum is the cheapest headline multiple of the set and the most strategically coherent — it consolidates the supply basin feeding the corridor Williams already owns, and it comes with two immediately-sanctioned expansion projects. Two disciplines to note: management explicitly declined to quantify synergies (“we’re not going to quantify yet”), and declined to give Momentum’s standalone CAGR, saying only that it is “additive” and would not be dilutive to the 11% target. Both are appropriately conservative; neither is verifiable.

Disposals (FACT). Williams continues to recycle. In 2026: the South Mansfield upstream interests sold in January ($194M gain), the Brazos Permian II equity-method interest sold in June ($126M gain), $48M of proceeds from a business sale, and $345M of net property dispositions in the half. Combined with the Cogentrix divestiture, the pattern is consistent — sell non-core and upstream, buy fee-based midstream adjacent to the core.

Dividend and buybacks (FACT/INTERPRETATION). The dividend was held at $0.525 per quarter ($2.10 annualised), declared 2026-07-28 and payable 2026-09-28, consistent with an annual step-up cadence in the first quarter. Coverage is guided at 2.47x on AFFO. There are still no buybacks — the authorisation remains dormant. This is defensible while capital is being deployed at ~5x build multiples against a stock at ~14.6x pro-forma EBITDA; it is also, as noted in June, a structural absence of an opportunistic lever. When the stock fell 11% to $70.13 in late July, the company had no repurchase programme running and no insider stepped in to buy.

Compensation and incentive alignment (FACT, DEF 14A 2026-03-18; unchanged this period). The annual incentive is 85% Adjusted EBITDA plus three safety and methane gates at 5% each. The long-term incentive is 50% Cash Return on Invested Capital plus 50% AFFO per share, with a ±25% relative-TSR modifier. Those two long-term metrics are, as it happens, precisely the right ones for the risk the JV structure introduces: CROIC penalises EBITDA bought with capital, and AFFO-per-share penalises EBITDA bought with partner equity or share issuance. The 85%-Adjusted-EBITDA annual bonus does neither — and the headline “11%+ Adjusted EBITDA CAGR” target rewards exactly the scale that spending can manufacture. The entire returns discipline rests on the long-term half of the package.

Board and leadership (FACT). Two independent directors were appointed effective 2026-07-01: Lloyd W. (“Billy”) Helms, Jr., most recently President of EOG Resources, and Robb E. Turner, who joins the Audit and Governance & Sustainability committees. Adding a former EOG president at the moment Williams is doubling its Haynesville gathering exposure is a sensible, specific hire rather than a generic board refresh. Chad Zamarin completed his first year as President and CEO on 2026-07-01, with Alan Armstrong remaining Executive Board Chair.

Insider read (FACT, SEC sweep 2026-06-15 to 2026-08-14). Eleven Form 4s and two Form 3s. Activity is entirely routine: annual director equity grants (code A) of 2,785 shares each to Helms and Turner at $71.81; tax withholding (code F) of 8,226 shares at $71.76 (Wingo) and 1,176 shares at $77.62 (Porter); an option exercise (code M) of 1,899 shares at $24.98 (Porter); and two 2,000-share sales (code S) by Terrance Lane Wilson at $74.16 and $70.65. There are no code-P open-market purchases — no insider bought with conviction, including at the July low. Neither is there alarming discretionary selling. A neutral signal, unchanged from the prior report, and a notable contrast with Enterprise Products, where a five-year Form 4 review turns up not one open-market sale across 121 filings.

Verdict: a strong period, and the best capital-allocation quarter in recent memory — with the questions that matter deferred rather than answered. Management said it would fund the power program with project-level partner equity rather than common stock, and it did, at a capped 6.35% cost while retaining control. It bought the right asset in the right basin at the cheapest multiple of the cycle. It deleveraged into a capex peak. Those are real. The deferred questions: whether the second JV prices as well as the first; whether Momentum’s synergies are as good as implied when they are eventually quantified; and whether 49%-sold power economics plus 2.2% dilution leave enough of the 11% for shareholders. On the record so far, management has earned the benefit of the doubt — which is different from having it proven.

8. Changes and Headwinds — Last Two Years

Strategic and portfolio changes (FACT).

• The pivot to power, now executed. From an announced concept eighteen months ago to five commercialised projects, a ~6 GW backlog, a segment renamed “Transmission, Power & Gulf,” a 49%-sold JV with $5.34 billion of committed third-party capital, and — as of late July 2026 — a 200 MW plant actually running. Phase 2 of Socrates is due Q4-2026; Neo (682 MW, 12.5 years, ~$2.3 billion) in H2-2028.
• The Haynesville consolidation. The 2025 “wellhead-to-water” restructuring (Haynesville upstream sold into JVs with Woodside and JERA, midstream fee stream retained), Louisiana Energy Gateway in service Q3-2025, and now Momentum at ~$5.5 billion, plus Shelby Trough Connector and Delta Access. Williams has, in two years, converted a contested G&P position into the dominant Haynesville-to-Gulf-Coast platform.
• The financing architecture. The Blackstone/Apollo/KKR JV (July 2026) is a genuinely new capability — a repeatable, low-cost, non-dilutive channel to fund a capex program that operating cash flow cannot cover.
• Governance and leadership. Zamarin succeeded Armstrong as CEO on 2026-07-01; Helms and Turner joined the board on 2026-07-01.
• Portfolio pruning. Cogentrix divested (2026); South Mansfield upstream sold (January 2026); Brazos Permian II sold (June 2026).
• Target escalation. February 2026 Analyst Day codified 10%+; May raised the contracted base to ~9%; August raised the target to 11%+ with the contracted base at 11%.

Headwinds and open pressures (FACT/INTERPRETATION).

• Valuation. The most persistent headwind, and it has worsened. Composite own-history percentile 91.3, price-to-book 99.1, price-to-sales 99.3. Sell-side and commentary scrutiny continues — GuruFocus published a GF Value of $60.81 against a $75.08 price in July 2026, and Seeking Alpha coverage has flagged a “very expensive” forward multiple. Those are third-party opinions, not evidence, but they mark where the debate sits.
• Capital intensity at a cycle peak. First-half capital investment nearly doubled; growth capex guided to $7.3–7.9 billion; 2027 described by the CEO as “a bit of a high watermark from a capital perspective relative to earnings.”
• Weak gas prices through the summer. Porter cited “pretty weak gas prices through the summer months” and hurricane season as the reasons for holding the top of the guidance range in reserve — a reminder that the G&P and marketing lines still carry commodity beta.
• Regulatory cash cost. Transco’s April 2026 rate refunds were the specific driver of the year-over-year fall in Q2 operating cash flow.
• Undisclosed hyperscaler counterparties. Still undisclosed. Zamarin now says the backlog “is supported by multiple customers beyond just our first primary customer,” which reduces but does not eliminate concentration risk, and remains unverifiable externally.
• Dilution and NCI. ~2.2% share issuance for Momentum, and 49% of five power projects sold — both of which widen the gap between consolidated EBITDA growth and per-share cash growth.

Verdict: the changes strengthen the thesis materially more than they weaken it — and the stock has already been paid for that. Two years ago Williams was a well-positioned pipeline company with an interesting idea about power. Today it is a demonstrably capable builder of dedicated generation, the dominant Haynesville-to-Gulf-Coast platform, with the majority of a 11% five-year growth algorithm under contract and a proven low-cost financing channel. That is a genuinely better company. It is also a company whose equity now sits in the 91st percentile of its own valuation history, with a 2.79% yield and negative free cash flow. The business improved; the margin of safety did not.

9. Risk Analysis

The dominant risk for Williams is valuation and rate sensitivity, not credit loss or execution failure. This is a defensive, fee-based, investment-grade business that has just demonstrated it can build what it says it will build. The probability of catastrophic or total loss is low. The real risk is paying a 91st-percentile multiple for utility-grade returns and watching the premium normalise.

Reading the matrix. The risk profile has shifted in a specific and important way since June. Execution risk fell — Socrates was delivered, the financing was secured, the contracted base rose to 11%, leverage improved. Valuation and rate risk rose — every own-history percentile moved higher, and the factor evidence shows the stock is now held more as dividend-yield carry and less as momentum, which makes it more sensitive to the long bond and less to the AI narrative. The permanent-capital-impairment risk runs almost entirely through overpaying at a peak multiple, not through the business failing.

The asymmetry of the dominant risk (INTERPRETATION). When a low-volatility, defensive business trades at the 91st percentile of its own history, the distribution of outcomes is skewed: the upside requires the premium to persist or expand, which is historically rare for a capital-intensive, regulated business; the downside only requires it to normalise, which is common. A de-rate from ~14.9x to a ~12.5x mean is ~16% of multiple compression before any change in fundamentals — and midstream multiples have compressed far more than that in past risk-off episodes, including a −23% drawdown in this very stock in 2022 with essentially no change to its contracted cash flows. The fixed-rate, laddered balance sheet protects the cash flow from a rate shock but not the equity multiple: a vehicle yielding 2.79% re-rates downward when long rates rise, independent of operating performance.

The capital-cycle risk deserves separate weight (INTERPRETATION). The single most uncomfortable fact in this report is not about Williams. It is that Kinder Morgan, ONEOK, Enterprise Products, MPLX, Targa and Cheniere are all, simultaneously, near the top of their own valuation ranges while simultaneously accelerating capital into the same LNG-and-AI-power demand thesis. Williams’s certificated corridors are protected from the consequences; the ~$7.5 billion a year it is now spending is not. Behind-the-meter power in particular — where the current barrier is turbine access and construction skill rather than a federal licence — is exactly the kind of ~20%-EBITDA-yield business that capital finds.

10. Valuation Discussion (Embedded Expectations)

No price target; this is an embedded-expectations analysis.

Where the multiple sits (FACT). At the 2026-08-14 close of $75.20 on 1,223 million shares outstanding, market capitalisation is $91.97 billion. Adding total debt of $30,793 million, noncontrolling interests of $2,178 million and $35 million of preferred, less $203 million of cash, gives an enterprise value of ~$124.8 billion. Against that:

Against its own history the stock has never been more expensive on sales (FACT). The AZI own-history percentile ranks at 2026-08-14: composite 91.3rd; price-to-earnings 75.6th (32.4x); price-to-book 99.1st (7.08x); price-to-sales 99.3rd (7.71x). Every one of those is higher than at the June report (88.7 / 70.9 / 97.4 / 97.8). Two caveats on interpretation: the price-to-earnings percentile is the least reliable because GAAP EPS carries derivative marks and asset-sale gains; and the price-to-book reading is distorted upward by the −$11.8 billion retained-earnings deficit, which artificially suppresses book equity. Weight EV/EBITDA and price-to-AFFO most heavily. But the direction is unambiguous and is corroborated by the EV/EBITDA series itself, which has moved from ~11x in 2023 to ~15x today.

Peer comparison (FACT/INTERPRETATION). The table below uses ROIC.ai’s Q2-2026 enterprise values over trailing GAAP EBITDA — directional for ranking, not adjusted-basis. Williams’s adjusted denominator is ~16% above its GAAP one, so on a like-for-like adjusted basis the gap narrows; Williams remains at the top of the set on any basis.

Williams is the second-most-expensive name in the group (behind only the far smaller DT Midstream) and the second-lowest-yielding (behind only Targa, which barely pays). It has been repriced out of the income bucket and into a growth bucket, and its 11%+ contracted CAGR is genuinely the best in the set — the premium is internally logical. The question is whether it is sufficient.

Reverse-DCF / embedded expectations (INTERPRETATION). Take the pro-forma base: FY2026 run-rate Adjusted EBITDA of roughly $8.9 billion (the $8.4 billion guide plus the ~9 months of Momentum not in it, at the ~$0.65 billion implied by 8.5x on $5.5 billion), against a pro-forma enterprise value of roughly $130.3 billion. Delivering the full contracted 11% CAGR to 2030 produces ~$13.5 billion of FY2030 Adjusted EBITDA — meaning today’s enterprise value is only ~9.6x FY2030 EBITDA. In other words, the market is permitting roughly five turns of multiple compression over four years while still expecting a positive return.

Turn that into an equity return. ASSUMPTION: ~$42 billion of 2030 net debt (the capex program peaks in 2027 and is partly partner-funded), ~$5 billion of noncontrolling interests as the JV scales, and ~1,255 million shares after the Momentum issuance. On those assumptions:

The conclusion is robust to reasonable variation in the debt and NCI assumptions: to earn a low-double-digit return from $75.20, an investor must underwrite both full delivery of the contracted 11% CAGR and the premium multiple holding at 13x for four years. Deliver the plan perfectly and exit at a 12x multiple — still above where Williams traded as recently as 2023 — and the return is about 8% a year. That is not a bad outcome; it is a fully-priced one.

On a price-to-AFFO basis (FACT/INTERPRETATION). The cleanest equity-level yardstick for a midstream business is price-to-AFFO. FY2026E AFFO of $5.15 per share puts the stock at 14.6x, a 6.8% AFFO yield, against a 2.79% dividend yield — the 2.47x coverage gap being reinvested in growth. For a business growing AFFO per share ~8%, 14.6x is not egregious in isolation, but it sits at the rich end of where pipeline cash-flow streams have historically traded, and the dividend yield is roughly half what KMI, OKE, ET and MPLX offer. Williams is asking income investors to accept half the current yield in exchange for faster growth. That trade now rests on much better evidence than it did in June — but the arithmetic above shows it still requires the multiple’s cooperation.

Scenarios (multiples and 2030 EBITDA only; no price target):

• Bear. The AI-capex cycle slows or a hyperscaler restructures; the contracted 11% fades toward 7–8%; the second power JV prices well above 6.35%; the multiple de-rates toward ~11x. Compression caps total return near or below the dividend yield, and the stock can trade meaningfully lower — as it did, by 23%, in 2022 with contracted cash flows intact.
• Base. The contracted 11% substantially delivers, with Momentum accretive and Delta Access and Shelby on schedule; FY2030 Adjusted EBITDA ~$13–13.5 billion; the multiple normalises toward ~12–12.5x. Result: a high-single-digit total return, growth plus yield less modest compression.
• Bull. The 11%+ is exceeded (management explicitly says it is “well positioned to exceed”), additional power projects are commercialised at ~5x build multiples with repeatable low-cost JV funding, and the premium holds at 13–14x. Low-double-digit to low-teens total return — requiring a midstream premium to persist through a full cycle, which is historically uncommon.

Verdict: premium-priced on every lens, and the premium is now better-earned than it was in June — but it is still fully paid. Williams is not absurdly valued: the contracted growth genuinely is the best in the sector and the execution risk genuinely has fallen. But at the top of its own ten-year range, with the second-lowest yield in its peer group and free cash flow after growth capex structurally negative, the embedded expectations require both flawless delivery and multiple persistence. The margin of safety is thin.

11. Variant Perception

Consensus (FACT/INTERPRETATION). The Street view is that Williams is the premier pure play on natural-gas-fired electricity — a defensive, fee-based, secular compounder with the best growth runway in midstream, a now-proven power-construction capability, and a repeatable low-cost financing channel, deserving of a premium multiple. After this quarter that view is not merely plausible; it is substantially demonstrated. The disagreement is about price and positioning, not about the quality of the business or the credibility of the plan.

The strongest bull case, strengthened. High-90s-percent fee-based, take-or-pay cash flow; the single best-positioned asset (Transco) for the converging LNG-export and data-center demand wave; a contracted 11%+ five-year EBITDA CAGR that excludes any further commercialisation; ~5x build multiples implying a ~20% EBITDA yield on growth capital, now validated by a completed 200 MW plant on time and on budget; the dominant Haynesville-to-Gulf-Coast platform after Momentum at 8.5x; $5.34 billion of committed partner equity at a capped 6.35%; leverage falling to 3.67x into a capex peak; an investment-grade, fixed-rate, laddered balance sheet; a dividend covered 2.47x; and a beta of 0.49. On this reading Williams is utility-grade growth and the market is correctly paying up for the best house in a structurally favoured neighbourhood.

The strongest bear case, narrowed but sharpened. Strip out the narrative and Williams is a regulated- and contracted-return business earning 7.4% on invested capital — the textbook definition of utility-grade — priced at ~32x adjusted earnings and the 99th percentile of its own price-to-sales history. Free cash flow after growth capex was −$1.6 billion in the half; growth capex is $7.3–7.9 billion against ~$6.7 billion of operating cash flow. The headline growth metric is becoming less complete: 49% of the flagship power economics belong to Blackstone, Apollo and KKR, ~2.2% of the company was issued for Momentum, and guided AFFO per share grows 7.7% against an 11%+ consolidated target. The quarter itself decelerated (+6% adjusted EBITDA, an EPS miss, operating cash flow down year over year) behind a GAAP headline flattered by $126 million of disposal gains. And the entire sector is racing capital at the same demand at the same time — the classic precondition for return compression.

The factor read — is consensus crowded, and in what? (FACT, FactorsToday). This is the most useful piece of variant evidence, and it has changed in a way that matters. In June the profile was dividend-yield-plus-momentum. Today, in the Base + Sector model (R² 0.476): DividendYield +0.952 (up from +0.775), Sector: Energy +0.828, Liquidity −0.540, Market +0.507, Sector: Utilities +0.465, LowVolatility +0.460, OilPrice +0.452, Value −0.434, SmallSize −0.341, Quality −0.271, and Momentum only +0.200 (down from +0.315). The leaderboard: y1 +35.3% (Sharpe 1.44), y3 +34.1% annualised (Sharpe 1.41, max drawdown just −12.4%), y5 +30.8% (Sharpe 1.21) — but m3 −9.8% annualised (Sharpe −0.46), roughly −2.5% raw over the quarter, with the stock 4.6% below its all-time high.

The interpretation, and the variant view. The momentum leg has drained out while the yield leg has strengthened. Williams is no longer principally held as an AI-power momentum trade; it is held as low-beta energy-and-dividend carry. Three consequences follow, and together they are the variant perception:

1. The de-rate risk has changed shape. In June the natural bear catalyst was “the AI story breaks.” Today the natural bear catalyst is “long rates back up and a 2.79%-yielding, utility-grade cash flow gets repriced against them” — a risk that is entirely independent of whether Williams executes. The +0.95 DividendYield loading is the single largest in the model; that is what the marginal holder owns.
2. The negative Quality loading remains, and it aligns with the fundamentals. A statistical model with no view on the narrative still classifies Williams as not a high-quality name — consistent with 7.4% ROIC, negative free cash flow and a debt-and-partner-funded growth model. The market is paying a quality-growth multiple for a factor profile that is neither cheap nor high-quality, only high-yield.
3. The excellent trailing track record is the trap, not the evidence. A three-year Sharpe of 1.41 with a maximum drawdown of only 12.4% makes this name feel safe at precisely the moment its valuation makes it least safe. That trailing record is a backward-looking artefact of the 2024–26 re-rate. The 2022 experience — a −23% drawdown while contracted cash flows barely moved — is the more instructive precedent for what a multiple reset looks like here.

Weighing the two sides (INTERPRETATION). The honest synthesis has shifted since June, and the shift should be stated plainly. In June the variant view was that consensus was right about the business and wrong about the price because the growth was not yet signed. That second clause is gone: the growth is signed. The variant view now is narrower and, I think, more durable: consensus is right about the business, right about the plan, and is underwriting the persistence of a 91st-percentile multiple on a 7.4%-ROIC asset whose marginal holder is a yield buyer. The disagreement is no longer about whether Williams can deliver; it is about what a delivered plan is worth, and about who is left to buy the stock at 2.79%.

The 3–5 assumptions that matter most:

1. The contracted 11% CAGR converts to per-share AFFO growth better than the 7.7% guided for 2026.
2. The second and subsequent power JVs price at or near the 6.35% capped cost of equity achieved with Blackstone.
3. Hyperscaler counterparties honour 12.5-year behind-the-meter contracts, and the customer base genuinely broadens beyond the first anchor.
4. Momentum’s synergies and growth arrive as implied, compressing the 8.5x entry multiple.
5. The premium multiple persists through a rate cycle and a market drawdown.

Falsification. The bull case is falsified by: a second JV priced materially above 6.35%; a Socrates-class project missing schedule or budget; a hyperscaler restructuring a long-dated contract; the 2027 AFFO-per-share bridge showing per-share growth stuck below ~8% while consolidated EBITDA compounds at 11%; or leverage breaching the 4.0x internal ceiling. The bear case is falsified by: per-share AFFO growth converging on the consolidated CAGR; additional power FIDs at ~5x build multiples funded without further common issuance; Momentum synergies quantified and material; and the multiple holding through an actual market or rate drawdown rather than an idiosyncratic 11% dip. The factor evidence — a strengthening dividend-yield loading, a fading momentum loading, persistent negative value and quality loadings, and a flat-to-negative three-month return at a 91st-percentile valuation — tilts the weight of evidence toward the view that consensus is offsides on the durability of the premium, even while it is now demonstrably right about the execution of the plan.

12. Fact vs. Interpretation Table

13. Open Questions

1. What does the 2027 AFFO-per-share bridge look like? With ~2.2% dilution from Momentum, a full year of 49%-sold power NCI, and the 2027 capex “high watermark,” per-share growth could compress further before it inflects in 2028. Management has not provided this bridge.
2. At what cost of equity does the next power JV price? The 6.35% cap is the single most favourable data point in the financing story. If the second deal prices materially higher, the first was a top-of-market print and the funding advantage is narrower than it appears.
3. Who are the hyperscaler counterparties, and how concentrated is the backlog really? Zamarin asserts “multiple customers beyond just our first primary customer.” That is unverifiable from the outside and is the largest single undisclosed risk in the story.
4. What are the Momentum synergies worth? Management explicitly declined to quantify them, and declined to give Momentum’s standalone CAGR. The 8.5x entry multiple is attractive only if it compresses as claimed.
5. How much of the ~6 GW power backlog converts, at what build multiple, and how quickly does that multiple drift? Socrates was the easiest project by construction. Build multiples historically rise as the best sites are used.
6. Does behind-the-meter generation have a durable barrier? The current barrier is turbine access and construction capability. Independent power producers, utilities and the hyperscalers themselves can build too. What stops a ~20% EBITDA yield from being competed down?
7. How does the equity behave in a genuine rate back-up? The DividendYield loading of +0.95 has never been tested at a 2.79% yield. The 2022 −23% drawdown is the closest available precedent, and it happened for a different reason.
8. What is the long-run steady-state ROIC once the power platform is at scale? If ~5x build multiples hold across the backlog, consolidated ROIC should rise from 7.4%. Management has not framed a target, and the CROIC PSU metric would be the natural place to disclose one.

14. What Must Be True

For the bull case to work (accepting the current price as reasonable), all of the following must hold:

1. The contracted 11%+ Adjusted-EBITDA CAGR delivers through 2030 — and management’s stated confidence in exceeding it proves out through additional commercialisation.
2. Per-share economics keep pace better than the 2026 guide implies: AFFO per share must compound closer to 9–10% than to 7.7% once the 2027 capex year passes.
3. The JV financing channel remains repeatable at a cost of equity near 6.35%, so that the remaining backlog is funded without common issuance.
4. Behind-the-meter build multiples hold near ~5x across the backlog, and hyperscaler counterparties honour long-dated contracts.
5. The market continues to award a 13x-plus forward EV/EBITDA multiple through a rate cycle.

Falsification test for the bull case: a second Power Innovation JV priced materially above a 6.35% cost of equity; or any Socrates-class project missing schedule or budget; or a hyperscaler restructuring a behind-the-meter contract; or a 2027 guide showing AFFO-per-share growth stuck below ~8% while consolidated EBITDA compounds at 11%; or leverage breaching the 4.0x internal ceiling. Any one of these would show that the premium is resting on terms that were available once rather than durably.

For the bear case to work (the premium normalises and the return disappoints), all of the following must hold:

1. The multiple compresses from ~14.9x forward toward the ~12x historical mean over the next several years — most plausibly via a back-up in long rates repricing a 2.79% yield, not via an operating failure.
2. Per-share growth continues to trail consolidated growth by roughly three points as noncontrolling interests and share issuance accumulate.
3. The sector-wide capital race compresses returns on the new build across the industry, so that ~5x power build multiples drift toward 6–7x.
4. No new source of value — a re-rating of the power platform as a separate business, or a step-change in disclosed ROIC — arrives to justify the premium.

Falsification test for the bear case: AFFO per share compounding at or above 10% in 2027–2028 while leverage falls below 3.5x; or two or more additional large power FIDs at ~5x build multiples funded entirely with partner equity at or below 6.35%; or Momentum synergies quantified at a level that visibly compresses the 8.5x entry multiple; or the stock holding a 14x-plus forward multiple through a genuine market or rate drawdown. Any one of these would show that the premium is an earned re-rating rather than a cyclical one.

The single most important thing to watch over the next two quarters: the 2027 guidance bridge — specifically the AFFO-per-share line. That one number arbitrates the entire debate. It reveals simultaneously how much of the 11% consolidated CAGR survives noncontrolling interests, dilution and interest expense; whether the 2027 capex peak is as manageable as management claims; and whether the financing architecture that makes this quarter look so good is an advantage for shareholders or a transfer of the best economics to the private funds who supplied the capital.

15. Source Appendix

See Appendix B — Source Appendix below for the full, dated source list.

APPENDIX A — Standard Diligence Questionnaire

The Williams Companies, Inc. (NYSE: WMB) · 2026-08-15

Supplemental to the analysis above. Answers are labelled Fact / Interpretation / Assumption / Open Question where the distinction matters.

General

What thoughtful questions have other investors asked about this company? The Q2-2026 call gives an unusually clean read on where sophisticated investors are focused, and the questions are notably not about whether the business works. Praneeth Satish (Wells Fargo) pressed on whether the 11% target embeds conservatism given that Momentum and Delta Access alone appear to add ~200bp — Porter conceded the company is “well positioned to exceed” it. Jeremy Tonet (JPMorgan) asked the concentration question directly: does doing business with one major hyperscaler preclude signing another? (Answer: no.) John MacKay (Goldman Sachs) asked how the next set of behind-the-meter projects gets funded and whether the Blackstone JV can be expanded — the single most important structural question about the company. Jason Gabelman (TD Cowen) asked whether the Momentum multiple should be assessed net of the LEG connector and Delta Access, i.e. whether the headline 8.5x is flattering. Robert Catellier (CIBC) asked whether Momentum is dilutive to the long-term growth rate and how the segment mix looks by 2030. Julien Dumoulin-Smith (Jefferies) asked about the lock-up on the shares issued to the seller. Manav Gupta (UBS) asked what drives the top versus the bottom of the guidance range.

Interpretation: the buy-side has moved past “can they build it” to “how much of it accrues per share, and how repeatable is the financing.” That is exactly the right question set, and it mirrors this note’s conclusion.

Cyclicality & Earnings Nature

Are earnings at a cyclical high or low? Neither, in the conventional sense — but closer to a structural inflection than a cyclical peak. Fact: ~High-90s-percent of cash flow is fee-based or rate-regulated, so reported earnings do not swing with the commodity the way an E&P’s do. The genuinely cyclical slivers are the Gas & NGL Marketing Services segment (~5% of Adjusted EBITDA) and the keep-whole exposure in the West. Fact: Porter noted “pretty weak gas prices through the summer months” as a reason for holding the top of the guidance range in reserve — so if anything the commodity-sensitive slivers are at a low, not a high. Interpretation: the multiple is at a cyclical high even though earnings are not.

Driven by the external environment or internal actions? Overwhelmingly internal at this point. The FY2026 EBITDA increase is attributable to projects placed in service (Louisiana Energy Gateway, Gulf expansions, storage), contracted rate increases at Transco, acquisitions (Rimrock, Saber, now Momentum) and the new power platform — all decisions rather than prices. The external environment supplies the demand (LNG, data centers) but Williams’s earnings growth is a function of capital deployed against contracts already signed.

How stable are revenues? Reported revenue is unstable and largely uninformative: $10.6B (2021) → $11.0B (2022) → $10.9B (2023) → $10.5B (2024) → $12.0B (2025), because it includes commodity pass-throughs and mark-to-market derivative lines. First-half 2026 total revenue of $6,083M includes a −$265M net loss from commodity derivatives. The correct stability metric is Adjusted EBITDA, which has risen every year: ~$5.0B → ~$6.4B → ~$6.8B → ~$7.1B → $7.75B → $8.3–8.5B guided. Service revenues — the fee line — rose from $4,044M to $4,358M in the half.

Outlook for products/services; how big is the market? Fact (management): US LNG export capacity is “about 18 Bcf a day” today with forecasts “above 40 [Bcf/d] in [the] next 10 years”; the earnings release cites Gulf Coast LNG demand rising ~20 Bcf/d over ten years; the Haynesville “is expected to grow by over 10 Bcf per day.” Interpretation: the LNG leg is contractually durable (15–20-year offtake); the data-center leg is the extrapolative one. Both are domestic-supply-driven and international-demand-driven, so the market is both.

Business Quality & Competitive Moat

Is the industry getting more or less competitive? Less competitive in interstate transmission and more competitive in the new power tier. Fact: new greenfield interstate pipe is effectively unbuildable — Mountain Valley required six-plus years and an act of Congress; Constitution and Atlantic Coast were cancelled after large sunk costs. That entrenches incumbents holding certificated corridors. Interpretation: behind-the-meter generation has no equivalent barrier; the current constraint is turbine supply and construction capability, both of which are replicable by independent power producers, utilities and the hyperscalers themselves.

How profitable is the business (ROIC, ROE)? Fact: FY2025 return on invested capital 7.44%; return on capital 8.28%; return on equity 20.8%. The 2020–2025 ROIC range is 4.2%–8.5%. The gap between the ~21% ROE and the ~7.4% ROIC is leverage. Interpretation: this is a utility-grade unlevered return, roughly at or slightly above a reasonable cost of capital — mid-pack among the midstream names I have recently covered (KMI ~5.8%, OKE ~8%, EPD ~10.8%).

How profitable is the industry — how many competitors, what barriers? Bifurcated. Interstate transmission earns a regulated, protected, capped return with essentially no share volatility. Gathering and processing is contested basin-by-basin with lower barriers, which is why the Momentum deal — establishing dominance in one basin rather than presence in many — is strategically the right shape. Roughly seven scale competitors matter: KMI, ET, OKE, ENB, MPLX, TRGP and DTM.

Can the business be easily understood? Yes, at the operating level: pipes, fees, contracts. No, at the reported level — the interaction of Modified EBITDA, Adjusted EBITDA, proportional joint-venture EBITDA, AFFO and GAAP net income requires care, and the addition of a 49%-sold power joint venture makes the consolidated figures a progressively less complete description of shareholder economics. This is the single most important accounting caveat in the whole analysis.

Can it be undermined by foreign low-cost labor? No. The assets are physical, domestic, and regulated. Foreign competition is irrelevant. The relevant cost exposure is to steel and steel-pipe tariffs and to skilled construction labour, both named in the company’s own risk factors.

Do brands matter? What is the nature of competition? Brands are irrelevant. Competition is over routes, rights-of-way, basin dedications and — newly — turbine delivery slots and construction capability. Zamarin’s framing of the Momentum rationale is instructive: the competition that matters is over position, not price.

Customers’ switching costs? High and physical on Transco — a captive delivery point has no alternative route. Moderate in gathering and processing, backed by minimum-volume commitments and acreage dedications (Momentum brings over one million dedicated acres). High but time-limited in power: the contracts are long (Neo is 12.5 years) but they do have an end date, unlike a pipeline’s structural position. Weak in marketing.

Financial Condition & Balance Sheet

Assets not fully recognised on the balance sheet? Yes, materially. Transco’s rights-of-way and its FERC certificate are the most valuable assets Williams owns and are carried at historical cost or not at all. The −$11,834M retained deficit is a legacy of decades-old write-downs and depresses book equity to $13,190M — which is precisely why the price-to-book percentile reads 99.1 and should be discounted as a valuation signal. Equity-method investments (Blue Racer, Appalachia Midstream, Discovery, Louisiana LNG) generate roughly $990M of proportional Modified EBITDA in the 2026 guide but appear on the balance sheet at carrying value.

Off-balance-sheet liabilities? The principal ones are the capital commitments implicit in the sanctioned project backlog — Neo (~$2.3B), Delta Access ($1.5B), Shelby Trough Connector, the Transco expansion slate — and the joint-venture obligations under the Blackstone structure. The 2033 buyout option on the partners’ 49% interest is a contingent future call on capital, at the then-remaining partner investment balance. Open question: this analysis does not have visibility into whether that option is expected to be exercised or funded.

How conservative is the accounting? Mixed, and worth stating plainly. Conservative: the Adjusted figures strip out the disposal gains and derivative marks that flatter GAAP — this quarter Adjusted EPS ($0.50) was below GAAP EPS ($0.68), which is the honest direction. AFFO is stated net of preferred dividends and net noncontrolling-interest distributions. Less conservative: Adjusted EBITDA includes ~$990M of proportional joint-venture EBITDA that arrives as distributions rather than controlled cash, making the headline denominator ~16% larger than GAAP EBITDA; and the headline growth target is a consolidated EBITDA CAGR at a moment when a rising share of that EBITDA is owned by third parties.

How CapEx-hungry is the business? Very. FY2026 growth capex guided $7.3–7.9 billion against ~$6.7 billion of operating cash flow net of working capital. First-half capital investments excluding acquisitions were $3,284M versus $1,709M a year earlier (+92%). Free cash flow after capex and dividends was roughly −$1.6 billion in the half, funded with $1,671M of net new long-term debt. This is the defining financial characteristic of the business in its current phase, and 2027 is described by the CEO as “a bit of a high watermark from a capital perspective.”

Capital Allocation & Management

How much FCF does the business generate, how is it used, what is the philosophy? On a conventional definition, free cash flow after growth capex is negative in the build phase. The sector-appropriate analogue is AFFO: FY2026E $6,375M ($5.15 per share), of which $2,585M goes to the common dividend (2.47x coverage) and the remainder is reinvested. The philosophy is explicit and consistent: dividend-first with a modest annual step-up, growth funded from retained AFFO plus debt plus — new this year — project-level partner equity, and no buybacks.

Significant acquisitions recently? Yes: Momentum Midstream / M6, announced 2026-08-03, up to $5.5 billion (~$3.5B cash and debt, ~$2B Williams equity), at ~8.5x projected 2027 EBITDA consolidated (~9x net of the NG3 noncontrolling interest). It adds >4,000 miles of pipe, >1 million dedicated acres, 6 Bcf/d of gathering capacity and 4.05 Bcf/d of take-or-pay pipeline capacity, and is subject to HSR clearance. Prior cycle: MountainWest (2023, ~$1.5B), DJ Basin Cureton and Rimrock (2023, ~$1.27B), Gulf Coast Storage / Hartree (2024, ~$1.95B at ~12x trailing), Discovery and Crowheart (2024), Cogentrix (2024, ~$1.9B — since divested), Saber (2025). Momentum is the cheapest headline multiple of the cycle and the most strategically coherent.

Buying back shares? No. A repurchase authorisation exists and remains dormant. Interpretation: rational while capital compounds at ~5x build multiples versus a stock at ~14.6x pro-forma EBITDA — but it means there is no opportunistic lever. When the stock fell 11% to $70.13 in late July, nothing was bought, by the company or by any insider.

Issuing large amounts of new shares to insiders? No. Share-based compensation amortisation was $39 million in the half — immaterial against a $92 billion market capitalisation. The meaningful issuance this period is external: ~$2 billion of stock as Momentum consideration, roughly 26–27 million shares, about 2.2% dilution, released to the seller over 180 days with a volume-based trading restriction rather than a hard lock-up.

Compensation policy of directors and management? Fact (DEF 14A, 2026-03-18): the annual incentive is 85% Adjusted EBITDA plus three safety and methane gates at 5% each (capped at 200%). Long-term performance share units are 50% Cash Return on Invested Capital + 50% AFFO per share, with a ±25% relative-TSR modifier (CROIC replaced ROCE in 2025). Interpretation: the long-term half is well-designed for precisely the risk this structure introduces — CROIC penalises EBITDA bought with capital, AFFO-per-share penalises EBITDA bought with partner equity or share issuance. The annual half does neither, and the headline public target (“11%+ Adjusted EBITDA CAGR”) rewards scale that spending can manufacture. Directors receive annual equity grants (the two new directors each received 2,785 shares at $71.81 on 2026-08-05).

Motivations of management? Growth-oriented, execution-focused, and — on this quarter’s evidence — willing to give away economics to accelerate. Zamarin was the architect of the growth and deal strategy before becoming CEO on 2026-07-01, so the strategic continuity is deliberate. The board was strengthened on the same date with Billy Helms (former President of EOG Resources), a specific and sensible hire given the doubling of Haynesville exposure. Insider behaviour is neutral: across the eleven Form 4s filed between 2026-06-15 and 2026-08-14 there are no code-P open-market purchases, only routine grants, tax withholding, an option exercise and two 2,000-share sales. Nobody is buying with conviction; nobody is selling with alarm.

Valuation & Market Data

Is the stock an ADR, MLP, or K-1 issuer? None of these. Williams is a US-domiciled Delaware C-corporation listed on the NYSE, issuing a standard 1099. It rolled up its former MLP (Williams Partners) in 2018. This is a genuine structural advantage over ET and MPLX for tax-sensitive and institutional holders, and part of why it commands a premium within its comp set.

Dividend policy? $0.525 per quarter, $2.10 annualised, declared 2026-07-28 and payable 2026-09-28 to holders of record 2026-09-11 — a 2.79% yield at $75.20. The cadence is a single annual step-up in the first quarter (~5%/year recently: $1.90 in 2024, $2.00 in 2025, $2.10 in 2026). Coverage is guided at 2.47x on AFFO. Williams has paid some dividend for over five decades, but the payout was cut 69% in 2016 during the post-Energy-Transfer-merger deleveraging, so “decades of dividends” should not be read as decades of uninterrupted growth.

How profitable is the business? See above: FY2025 net income to common $2,615M on $11,950M of revenue; FY2026E adjusted income $2,905M and Adjusted EPS $2.35. EBITDA margin 54.8%, operating margin 36.9%. ROIC 7.44%, ROE 20.8%.

Is net income diverging from cash from operations? Yes, in both directions, and it matters. First-half 2026 GAAP net income was $1,788M against $2,979M of operating cash flow — a normal, healthy relationship for a capital-intensive business with $1,176M of depreciation. But second-quarter operating cash flow fell $74 million year over year while Adjusted EBITDA rose 6%, driven by “unfavorable net changes in working capital driven by the payment of Transco’s rate refunds in April 2026.” That divergence is legitimate and one-off, but it is a real cash cost of the regulatory bargain and deserves to be recorded rather than smoothed over. Separately, GAAP net income in the half is inflated by $194M (South Mansfield) plus $126M (Brazos Permian II) of disposal gains that produce no recurring cash.

Risks & Downside

What factors would cause the stock to decline? In order of assessed likelihood-times-impact: (1) multiple normalisation from the 91st percentile of its own history, most plausibly triggered by a back-up in long rates repricing a 2.79% yield — note the DividendYield factor loading is +0.95, the largest in the model; (2) the per-share wedge widening — consolidated EBITDA compounding at 11% while AFFO per share compounds at 7–8% as noncontrolling interests and dilution accumulate; (3) a second power JV pricing materially above the 6.35% capped cost of equity, which would mark the first as a top-of-market print; (4) a hyperscaler counterparty restructuring a long-dated behind-the-meter contract; (5) sector-wide capital-cycle return compression as every peer builds into the same demand simultaneously; (6) project slippage on the 2027–2029 slate (SESE, NESE, Shelby, Neo, Delta Access).

Risk of a catastrophic loss? Low. The balance sheet is investment grade, ~99% fixed-rate and well-laddered; leverage is 3.67x against a 4.0x internal ceiling and falling; the cash flow is high-90s-percent fee-based; and the dividend is covered 2.47x. The plausible catastrophic scenarios are physical (a major pipeline incident with mass-casualty consequences and the attendant regulatory response) or geopolitical, not financial.

Chance of a total loss? Negligible on any reasonable horizon. The realistic bad outcome for an investor at $75.20 is not loss of capital but mediocre return — a high-single-digit or lower annualised total return over four to five years while the business performs as promised and the multiple normalises. That is the risk this memo is actually about.

Recent News & Events

Has the business environment changed recently? Yes, and materially, in the nine weeks since the prior report. Williams delivered its first utility-scale behind-the-meter power plant (Socrates Phase 1, 200 MW, late July 2026, on time and within budget); closed a $5.34 billion Blackstone-led joint venture funding 49% of five power projects at a capped 6.35% cost of equity with $900 million of additional consideration; agreed to acquire Momentum Midstream for up to $5.5 billion; sanctioned Delta Access ($1.5B, 2.25 Bcf/d, Q1-2029) and the Shelby Trough Connector (750 MMcf/d, Q2-2028); signed Transco’s Leidy Access and Garden Connector; upsized Power Express to 800 MMcf/d; raised FY2026 guidance by $200 million at the midpoint; and raised the long-term growth target to 11%+ with the contracted book supporting the full 11% (up from ~9% in May and ~8% in February).

Significant acquisitions? Momentum Midstream (M6) — see above. Also note the disposals: the South Mansfield upstream interests (January 2026, $194M gain) and the Brazos Permian II equity-method investment (June 2026, $126M gain).

Change in accounting policies? No change in accounting policy, but a change in segment presentation: the reportable segment formerly “Transmission & Gulf of Mexico” is now “Transmission, Power & Gulf.” Power is a named reporting pillar for the first time. Prior-period figures are restated on the new basis in the Q2 release, so the comparison is clean.

Recent changes — new markets, facilities, management? New markets: dedicated behind-the-meter power generation for hyperscalers, now proven in operation; and, via Momentum, the Shelby Trough / Western Haynesville. New facilities: Socrates Phase 1 in service; Louisiana Energy Gateway in service since Q3-2025; Southeast Supply Enhancement under construction with possible early in-service at the start of 2027; NESE compression beginning late 2026. Management and board: Chad Zamarin completed his first year as President and CEO on 2026-07-01, with Alan Armstrong as Executive Board Chair; Lloyd W. (“Billy”) Helms, Jr. — former President of EOG Resources — and Robb E. Turner joined the board as independent directors effective 2026-07-01, Turner serving on the Audit and Governance & Sustainability committees.

APPENDIX B — Source Appendix

The Williams Companies, Inc. (NYSE: WMB) · Report date 2026-08-15 · All sources accessed 2026-08-15

Sources are grouped by tier, primary first. Every non-obvious fact above traces to an entry below.

1. Primary — SEC filings and company disclosure

2. Primary — management commentary (transcripts)

Key quoted passages: Socrates Phase 1 in service “in under 18 months since commercialization” (Zamarin); JV terms — “$5.34 billion of committed capital, including $4.4 billion for 49% … plus $900 million of additional consideration … at an attractive capped 6.35% cost of equity” and the 2033 buyout option (Porter); the growth-target bridge 8% → 9% → “11% plus” and “centered on our existing contracted book of business” (Porter); Momentum funded by “$3.5 billion in cash and debt and $2 billion of equity” at “approximately 8.5x,” “approximately a 9x multiple” net of the NG3 noncontrolling interest (Zamarin); the share-release mechanics in lieu of a lock-up (Wingo); “supported by multiple customers beyond just our first primary customer” (Zamarin); the 2027 capex “high watermark” (Zamarin); “it’s not only just the turbines” (Larsen); 2027 project status on SESE and NESE (Larsen); “pretty weak gas prices through the summer months” (Porter).

3. Primary/secondary — transaction counterparties and wire services

4. Quantitative data sources

Authority note: ROIC.ai, AZI and FactorsToday are third-party aggregated or statistically-estimated data, not primary. Every material figure driving a verdict has been reconciled to the Form 10-Q or the earnings release, and where a discrepancy exists the filing governs. Enterprise value in the valuation section is recomputed from the current price and the Q2-2026 balance sheet rather than taken from any feed — see the correction to the prior report’s stated multiple in the “Changes Since” section.

5. Analytical frameworks applied

• Competition Demystified (Greenwald & Kahn) — barriers to entry as the dominant question; the moat named in the taxonomy (government licence + irreplaceable right-of-way + economies of scale + customer captivity); the market-share-stability test applied to Transco (pass) versus gathering and processing (weaker); the ROIC test applied and failed at the “high-return” threshold (the competitive-position section).
• Capital Returns (Marathon Asset Management) — supply-side capital-cycle analysis applied twice: to the permitting-constrained interstate tier (favourable phase, capital cannot enter) and to the sector-wide race into LNG-and-AI-power capital deployment, including behind-the-meter generation where the barrier is turbine access rather than regulation (see below).

6. Notes on evidence quality and limitations

1. Management commentary is hypothesis, not evidence. The ~56% improvement in the multiple on invested capital from the JV, the Momentum accretion, and the 11%+ contracted CAGR are all management calculations that are unaudited and not independently verifiable. They are labelled as such above.
2. A management-versus-evidence contradiction is flagged in the financial-quality section. Management characterised the quarter as “firing on all cylinders.” The adjusted results decelerated (Adjusted EBITDA +6% versus +13% in Q1), Adjusted EPS missed consensus, and operating cash flow fell year over year, while the GAAP headline (+51% EPS) was flattered by two disposal gains and a derivative swing excluded from the adjusted figures. Both framings are defensible; this note reports the deceleration explicitly.
3. Secondary sources are unreliable on the Momentum multiple. Several outlets cite the transaction as supporting “>11%” growth without distinguishing the ~8.5x consolidated multiple from the ~9x figure net of the NG3 noncontrolling interest. The memo uses management’s own words for both.
4. Peer EV/EBITDA in the valuation section is on a trailing GAAP basis and is directional for ranking only; each company’s adjusted denominator differs. Williams’s adjusted denominator is ~16% above its GAAP one because of proportional joint-venture EBITDA, so the like-for-like gap to peers is narrower than the table implies — but Williams remains at the top of the set on any basis.
5. The 2030 scenario table in the valuation section rests on stated assumptions for 2030 net debt (~$42B), noncontrolling interests (~$5B) and share count (~1,255M) that are estimates, not company guidance. They are labelled as assumptions, and the conclusion is robust to reasonable variation in them.
6. The 2025-07-01 single-day decline of 6.51% could not be attributed to a company-specific event with confidence and was deliberately left out of the event map rather than guessed at.