Williams

IThe business

Williams is a Tulsa company that moves natural gas and charges a fee for doing it. It owns an interest in and operates over 32,000 miles of pipeline in 24 states and the Gulf of America, 35 gas processing plants, 9 plants that split natural gas liquids, 423 billion cubic feet of gas storage and about 800 customers, and its 5,987 employees do the work.⁷ The company says its system delivers one third of the nation’s natural gas.³ In a toll-road picture, the interstate pipelines are the highway, the gathering systems are the on-ramps in the shale fields, and the new power plants are a private road built for one customer.

All figures are in US dollars and follow US accounting. TTM means the twelve months to 30 September 2026, which I build from the 2025 10-K and the nine months reported on 2 November.⁷⁵ The third-quarter report is the latest one out.

Over those twelve months revenue was $12.5B, adjusted EBITDA $8.25B, AFFO $6.46B and net income to common holders $3.13B C (calculated).⁷⁵³ Net income includes about $0.4B of pre-tax gains on asset sales in the first nine months, which adjusted figures leave out.³ Adjusted EBITDA includes the company’s share of the EBITDA of the pipelines and plants it owns jointly with others, about $1.0B in 2026, while the cash from those arrives as distributions, so I treat AFFO as the cash measure.³

The toll road itself: where the money comes from

The company reports four segments plus an Other. Almost three quarters of adjusted EBITDA comes from the first two.

Transmission, Power and Gulf

This segment earned $3.71B in 2025, 48% of the total C (calculated).⁴ Its core is three regulated interstate pipelines. Transco runs about 9,600 miles from Texas and the Gulf through the southeast and mid-Atlantic to the New York area, with design capacity of about 20.6 million dekatherms a day, 62 compressor stations and about 188 billion cubic feet of storage available to its customers. Northwest Pipeline (NWP) runs 3,900 miles from the San Juan basin to the Canadian border. MountainWest, bought in 2023 for $1.0B, adds 2,200 miles in the Rockies and 64 billion cubic feet of storage.⁷¹³ It also holds half of the 745-mile Gulfstream line to Florida, six Gulf Coast storage fields with about 120 billion cubic feet, and Gulf of America gathering lines, platforms and crude pipelines.⁷

These pipelines are paid for capacity that customers reserve, whether or not they use it. In 2025 Transco’s customers reserved about 20.8 million dekatherms a day on average and flowed about 15.0 million, and the bill is on the reserved number.³ The regulator, the Federal Energy Regulatory Commission (FERC), sets the rates in a rate case; the 10-Q says changes in commodity prices and volumes have a limited near-term effect on these revenues because most of the cost of service is recovered through the reservation charge.⁵ In 2025 these regulated transportation, storage and other revenues were $3.65B R (reported).³ The top ten customers made up 44% of them, and Transco’s three largest 22% of its revenue.⁷ Transco’s last rate case ran from a filing in August 2024 to a settlement the FERC approved on 30 December 2025, effective 1 March 2026; refunds for the gap were paid in April 2026.⁵

The newest piece is power. Williams builds gas-fired power generation next to the data centres and plants that need it, and sells the output under long contracts. It has five projects in a joint venture it calls Pantheon: Socrates (556 megawatts in Ohio, with the first half in service in late July 2026 and the second due this quarter), Apollo (490 megawatts, Ohio), Aquila (520 megawatts, Utah), Socrates the Younger (340 megawatts, Ohio) and Neo (682 megawatts, Ohio), which add up to about 2,590 megawatts.⁵ The contracts run 10 to 12.5 years, are primarily fixed price, and give the customer an option to extend. Williams named no customers in the filings I read, and I name none.⁵ The company says it has more than 6 gigawatts of backlog, and it expects to announce more power projects soon.¹⁴³ In July it sold 49% of the five projects to funds led by Blackstone, with Apollo and KKR alongside, for $5.34B of committed capital.¹⁴ I come back to what that deal means for owners below.

The one in-service example gives a first look at the economics. Socrates South is under an operating lease with fixed payments of about $205M a year for 2027 through 2031, on $1.0B of gross assets, and it earned $35M of fixed lease income in the third quarter, its first.⁵ That is a revenue yield of about 20% on cost before any operating cost, and I use 17% for EBITDA in my base, which is my estimate E (estimate).

Northeast gathering and processing

Northeast G&P earned $2.03B in 2025, 26% of the total C (calculated).⁴ It gathers and processes gas from the Marcellus and Utica shales in Pennsylvania, New York, Ohio and West Virginia, through consolidated ventures in which Williams owns 65% (the Northeast JV) and 66% (Cardinal), plus a 50% stake in Blue Racer.⁵ It gathered about 4.1 billion cubic feet a day on the consolidated systems in the first nine months and processed about 2.0 billion, and EBITDA grew 6% over nine months.³ Fees are the product; the volumes depend on producers drilling.

West

West earned $1.45B in 2025, 19% of the total C (calculated). It spans the Haynesville, Eagle Ford, Barnett, DJ, Piceance and Wamsutter areas, plus NGL storage and a half interest in the Conway fractionator.⁵⁴ On 3 September 2026 Williams closed on Momentum Midstream for about $5.1B, paid in cash, 26.9 million of its own shares, and $869M of debt it left in place. Momentum has 6 billion cubic feet a day of gathering capacity and 4 billion of pipeline capacity in the Haynesville, and it contributed $113M of revenue and $40M of segment EBITDA in its first 28 days.⁵¹⁵ The weighted-average time until its customer contracts come up for renewal is about nine years.⁵ The segment also carries a cost: minimum volume commitments in the Eagle Ford are stepping down in 2026.⁵

Marketing, and the commodity exposure that is left

Gas and NGL marketing earned $193M of adjusted EBITDA in 2025, 2.5% of the total C (calculated). Its derivatives swing reported earnings: a $116M unfavourable change in the first nine months of 2026.⁷³ About 93% of NGL production was under fee-based contracts in 2025 and 7% under contracts that pay in barrels of NGLs.⁷ The upstream interests are the one place with direct price risk, now smaller after the South Mansfield sale.⁵

How much of it is contracted

At 30 September 2026 Williams had $32.5B of remaining performance obligations, the revenue it has contracted for and will earn if customers keep their contracts; $21.0B of that is Transco and $4.0B NWP.⁵ For the interstate pipelines, the figure assumes today’s rates, which the FERC can change in a future rate case. It leaves out renewals and customers who have not signed. In 2027 alone it is $4.72B, which is 56% of the company’s 2026 EBITDA guide C (calculated); the figure is revenue, not profit, so I read it as the size of the base and not as a profit forecast.⁵³

Where the cash goes

The company expects $7.3B to $7.9B of growth capital in 2026, up from $6.1B to $6.7B in its February guide, plus $850M to $950M of maintenance capital. The plan names the power projects, Transco expansions, Haynesville growth and the Northeast as the main uses.⁴³⁵ Larger projects include the Southeast Supply Enhancement (about 1.6 billion cubic feet a day, FERC-approved January 2026, in service as early as the third quarter of 2027), the Northeast Supply Enhancement (400 million a day), Power Express (800 million a day, 2030) and Line 200, the 3.1 billion a day pipeline to the Louisiana LNG plant, in which Williams owns 80% of the pipe and 10% of the plant.⁵⁷ Those projects are the whole growth case, and the reason owner cash falls so far short of AFFO.

IIThe moat

A durable business keeps others from taking its profit. For a pipeline the protection is physical and legal: nobody lays a competing line beside a full one without years of permits, and the toll comes from a regulator or a contract. The ten-year record shows what that did for owners, and it is mixed.

AFFO per share went from $2.95 in 2018 to $4.78 in 2025, which is 7.1% a year, and the dividend from $1.36 to $2.00, which is 5.7% a year C (calculated).¹¹⁴⁶ Adjusted EBITDA grew from $4.44B in 2016 to $7.75B in 2025, 6.4% a year C (calculated).⁹⁴ The share count went from 751M diluted weighted in 2016 to 1,225M in 2025, because Williams sold $2.1B of stock in 2017 and in 2018 bought in the public units of its subsidiary Williams Partners (the WPZ merger) with shares.⁶¹¹ The first line to read in the story is the 2016 dividend: it was cut from $0.64 to $0.20 a quarter that year, and raised to $0.30 in 2017.¹⁰ That was a company with too much debt and too little cash, and it is why the leverage chart below starts at 5.3 times.

A holder at the end of 2016 had a price of $31.14 and now has $67.86, plus $14.71 a share of dividends declared from 2017 to 2025 C (calculated).¹⁶ The price fell to $20.05 at the end of 2020 and was $26.04 at the end of 2021. The big move came after: $34.83 at the end of 2023, $54.12 a year later and $60.11 at the end of 2025.¹ I read that as two things at once, a better business and a higher multiple on it, and the next chart separates them.

EBITDA rose 55% from 2019 to 2025 while the multiple paid for it went from 10.7 to 13.6 times, and it is 14.9 now C (calculated).⁶¹ A pipeline whose protection was priced at 10 times in 2019 and at 15 times today has become more valuable on paper than in cash flow, and a new buyer pays for the protection, not for a bargain.

What does protect it

• Reservation contracts. Customers pay for capacity whether they use it or not, which is why volume is not the main driver of pipeline revenue. Transco’s contracted revenue at 30 September was $21.0B, with $10.5B beyond 2030.⁵
• A regulator that lets them earn on what they spend. New Transco expansions are built against signed customer commitments; the 10-Q says the Transco expansions in the 2026 plan are fully contracted with firm transportation agreements.⁵ The Northeast Supply Enhancement is fully subscribed under precedent agreements signed in August 2025.⁵
• Physical position. Transco links the Gulf Coast and Marcellus supply to the Southeast and the New York area, where there is little spare room; the company says the southeast and Gulf Coast are the fastest growing gas demand regions.⁷
• Fee-based gathering with minimum volumes. About 93% of NGL production was under fee-based contracts, and some contracts carry minimum volume commitments that pay a shortfall.⁷
• Long contracts on the new power. 10 to 12.5 years, primarily fixed price, with $2.0B of future lease payments already scheduled for Socrates South.⁵

Where it is thin

Four things. First, the regulator can reset the rate: the 2025 Transco settlement, effective in March 2026, is the current rate, and Transco will be back at the FERC.⁵ Second, permits can be taken back at the edge. On 8 September 2026 the Third Circuit Court of Appeals vacated the New Jersey Water Quality Certification for the Northeast Supply Enhancement and sent it back to the state, which puts a project meant to be in service as early as the fourth quarter of 2027 in question.⁵ Third, the gathering business depends on someone else drilling; the Eagle Ford minimum volumes are stepping down this year.⁵ Fourth, the power contracts end. They run 10 to 12.5 years from 2026 to 2028 and the customer holds the extension options, so the earliest, Socrates, comes up around 2036, which is my terminal year.⁵

Capital intensity is the quiet part. Maintenance capital is about $0.9B a year, 11% of EBITDA, against $2.3B of depreciation in 2025, and I can’t tell from the filings how much of the $7.6B of growth spending is really holding the base where it is.⁴⁶ I deal with that in Valuation by testing how little the answer moves when I change the growth budget.

IIIThe competitive advantage

The question is what Williams can do that others can’t. On the evidence the answer is position, delivery and the ability to fund it.

On delivery, it raised its 2025 EBITDA guidance three times, by $350M in all, and ended at the $7.75B record, and the company says five-year adjusted EBITDA growth was 9% a year and EPS growth 14% a year to 2025.⁸⁴ In 2026 it brought Socrates South into service in late July, with the second half due by mid-November, closed Momentum ahead of schedule in the company’s words, and placed four pipeline projects in service, adding 413 million cubic feet a day.³⁵ I can verify the guidance steps and the quarter; I can’t test the 14% a year of EPS beyond what the company says, and it starts from the pandemic year.

On position, the link with demand is the point. The gas leaving the shale fields has to get to buyers that are growing: LNG export docks on the Gulf Coast, gas-fired power plants in the Southeast and Ohio, and industrial users. The Energy Information Administration (EIA) expects US LNG exports of 17.6 billion cubic feet a day in 2026 and 18.6 in 2027.¹⁶ Williams’ own proxy cites Wood Mackenzie for US gas demand growth of nearly 35% over the next decade after 73% since 2010.⁸ Williams has tied itself to the LNG chain on both ends: it owns 10% of Louisiana LNG, builds the pipe that feeds it, manages the gas supply and buys about 10% of the LNG produced.⁷

On funding, the company is rated BBB+ (stable) by S&P, Baa2 (positive) by Moody’s and BBB (positive) by Fitch, and borrowed $2.75B in September 2026 at 5.0% to 6.4%.⁵¹⁷ It raised $3.74B in July by selling a minority share in the power projects and, in 2026, $5.5B of long-term debt; both are real advantages and both are why a holder reads the share count.⁵

Against its closest rivals

For a pipeline, rivals are other lines to the same markets, and for gathering, other gatherers in the same basin. I have not built a peer table from other companies’ filings, so I print no rival numbers I haven’t checked. The comparison I can make is against the closest page on this site, the pipeline and utility company Enbridge.

New supply and new demand

The risk to the advantage isn’t a competing pipe. It is demand that does not show up or arrives late. The EIA’s October outlook shows the share of US electricity from natural gas at 40% in 2025 and 2026 and 39% in 2027, with Henry Hub gas at $3.16 in 2027, 9% lower than in 2026, so the data-centre story is not visible in the national gas-share number yet.¹⁶ Williams’ power projects are contracted at fixed prices on 10 to 12.5 years, which protects it from that. The pipeline expansions lean on utilities and LNG buyers who have signed commitments. What is not protected is the next set: the company says its backlog is more than 6 gigawatts against 2.6 announced, and the filings I read name no customer or contract for the difference.¹⁴

IVManagement

Chad Zamarin became president and chief executive on 1 July 2025 and joined the board, after Alan Armstrong ran the company for 14 years as chief executive and moved to executive chair. Stephen Bergstrom is the independent lead director.⁸ John Porter is chief financial officer and Larry Larsen chief operating officer since 2025.⁷ The proxy calls this a multi-year planned transition, and the proxy says Zamarin, the newest director, facilitated multiple of the company’s acquisitions.⁸ In July 2026 the board added two directors, going from ten seats to twelve, and in August Michael Creel said he will leave the board at the 2027 annual meeting.¹⁹²⁰ Armstrong is still on the payroll and the board, which means the person who built the plan and the person who is paid to carry it out are in the same room.

How they are paid, and what that pays them to do

Zamarin’s 2025 pay was $8.68M: $0.83M of salary, $6.0M of stock awards, a $1.49M annual bonus, $0.33M of pension value and $0.03M of other items. It was $4.92M in 2024 and $5.02M in 2023 C (calculated). The board says 90% of the chief executive’s target pay is at risk, and the pay ratio to the median employee is 61 to 1.⁸ Armstrong’s 2025 pay was $17.9M, of which $13.5M was stock awards.⁸

The annual bonus for 2025 was based on adjusted EBITDA and environmental and safety measures. The stock units that vest after three years pay out on two measures: AFFO per share and, since the 2025 awards, cash return on invested capital (CROIC), with relative total return against peers able to raise or lower the result by 25%. Before 2025 the second measure was return on capital employed.⁸ The 2022 awards, which vested in February 2025, paid 191.7% of target, because AFFO per share beat the stretch goal and the company’s annualised return of 30.9% put it in the top third of peers.⁸

I read this plan two ways. First, adjusted EBITDA and AFFO per share both count cash before growth capital, so they reward getting cash flow up and don’t ask what the capital cost. That is the same incentive I would expect to sit behind a growth company that borrows to pay its dividend. Second, CROIC is a return-on-capital measure, so in principle a project that earns less than the company’s average pulls it down. That is a test the Enbridge page found missing in that company’s plan, and it carries half the weight in Williams’s performance units. The test I can’t make is whether the target is hard: the proxy states the 2022 target for return on capital employed, 9.65%, and not the new CROIC targets I could compare.⁸

How much of their own money is in

A good deal. At 24 February 2026 Armstrong owned 2.89M shares including units about to vest, worth about $196M at today’s price C (calculated); Zamarin 587,470, about $40M; Porter 251,794, about $17M. The 25 directors and executive officers together owned 4.46M shares, 0.4% of the company.⁸ The rules require the chief executive to keep shares worth six times base pay and other named executives three times, with 50% of after-tax vested stock kept until they get there, and directors five times the amount the proxy sets for them.⁸ That is a real stake and a smaller share than at a founder-led company. In the quarter to 30 September no director or officer adopted or ended a trading plan, and the general counsel filed a notice of a proposed sale on 2 November; I took no view from it.⁵²¹

What they did with capital over a full cycle

On the whole this is a team that repaired the balance sheet in 2016 and 2017 and has built steadily since, paying a rising dividend and buying at fair prices. The record also shows the cost: net debt went from $23.3B at the end of 2016 to $32.2B, the diluted share count from 751M to 1,255M including awards, and the 2026 plan spends more than operating cash flow can cover. That is not a mistake; it is what building looks like when new capital earns somewhat more than it costs. I give credit for a clear plan and for the cut when it was needed, and I count the stock issued and the investor’s claim on the power cash as costs.

VFinancials

Reading Williams’s statements takes care, because five things in them do not mean what they seem to.

• Revenue is partly pass-through. It was $12.5B over the last twelve months, but $2.74B of nine-month revenue was product sales of gas and NGLs bought and resold, and derivative gains and losses sit in the line. The segment fees are the business.⁶³
• Reported earnings swing on one-offs. Net income in the nine months was $2.39B against adjusted net income of $2.17B, because of gains on asset sales and changes in the fair value of derivatives.³
• AFFO is before all capital spending. It starts from operating cash flow, takes out working capital, preferred dividends and minority distributions, and stops. It does not take off the $0.9B of maintenance capital or the $6B or more of growth capital, which is why the 2.4 times dividend cover the company reports is not the whole picture.³
• Adjusted EBITDA includes a share of partners’ businesses. About $1.0B of the 2026 figure is Williams’s proportional share of EBITDA from the pipelines and plants it owns jointly, whose cash arrives as distributions after their own costs.³
• Debt measures differ. The company’s ratio is 3.80 times at 30 September, after subtracting $857M of cash it spent on long-lead power equipment that the customer would reimburse if an order were cancelled. Mine, which counts all debt and subtracts only cash, is 3.90.³⁵

The latest quarter

The third-quarter report came out after the close on 2 November. Adjusted EBITDA was $2.039B against $1.920B a year before, up 6%; AFFO was $1.589B, up 10%; adjusted earnings per share were $0.54 against $0.49; net income to common holders was $699M; and cash from operations was $1.505B, up 5%. Dividend cover on AFFO was 2.42 times.³ The company said it is tracking toward the upper half of its $8.3B to $8.5B EBITDA range, kept growth capital at $7.3B to $7.9B and put year-end leverage near 3.75 times by its own measure.³

The share price moved a lot around the report. The stock fell 4.2% on 2 November, from $74.35 to $71.19, before the release; rose 1.8% to $72.47 on 3 November, the day of the call; and fell 6.4% to $67.86 on 4 November on 20.5M shares, against 6.9M to 9.0M a day on 29 October, 30 October and 2 November.¹² I do not know why. I found no company filing on 4 November. The 2 November filings include a prospectus that lets the former Momentum owners resell up to 26.9M shares, 2.1% of the company, with a cap of 537,496 shares a day among those under lock-up through 3 March 2027, which is about 4% of the roughly 12M extra shares traded on 4 November, so it can only be a small part of the story.¹⁵ I treat it as supply the market has to absorb, and it is already in my share count.

What an owner can really take out

Here is the sequence that matters. Over the twelve months to September, AFFO was $6.46B. Maintenance capital took about $0.9B, which is my estimate from the guide, and growth capital and equity investments about $7.2B, since capital expenditures were $7.64B and equity-method investments $0.47B C (calculated).³⁶ That leaves about −$1.7B after stock pay of roughly $0.1B. Using the cash flow statement instead, the figure after every dollar of capital and the minority owners is −$2.33B C (calculated). Common dividends paid were $2.55B. So the business did not pay its dividend out of owner cash in the last twelve months, and in 2025 it covered only 10% of it C (calculated).

The gap was paid for with $5.5B of new long-term debt in nine months, less $1.1B of repayments and $0.7B of commercial paper, and with $3.74B from the power investor, $0.5B of asset sales and a rise in cash from $63M to $1.76B.⁵ I show the investor’s money separately because it is partner capital: it buys 49% of the cash from five projects, and for a holder of the common stock it is a claim ahead of them. Every pipeline company that grows runs this way. What it means for a holder is that the dividend measures what the business can pay if it keeps growing, and not what it can pay out of cash left over today. The easy way out is to spend less on growth, since most of the $7.6B is discretionary until it is built, and that is the main reason I think the dividend is safe at 40% of AFFO. The cost is the growth.

The ten-year view is better: cash after all capital was positive every year from 2019 to 2024, between $1.0B and $3.1B, and covered the dividend in 2021, 2022 and 2023 C (calculated). 2025 and the last twelve months are when the build started to run ahead of it. The company harvested its base for several years and has chosen to spend again.

Stock pay is a small matter. Equity-settled awards were $93M in 2025 and $65M in nine months of 2026, about 1.6% of AFFO, and the shares come from new issues, so I deduct about $0.1B a year in the model.⁶⁵ Buybacks were $139M in two years, so they don’t offset it.

Cash taxes are the question I can’t settle from the filings. Williams booked an income tax provision of $857M in 2025, a 23.7% rate, and paid $162M in cash, $68M in 2024 and $31M in 2023; $744M of the 2025 provision was deferred, and deferred tax liabilities were $5.17B at year-end.⁷⁶ The 2025 tax law restored full bonus depreciation for good, which keeps the deferral going as long as the company keeps spending; in the nine months of 2026, $716M of a $725M provision was deferred.⁷³ If growth spending falls to the level of depreciation, the deferral stops growing and cash taxes move toward the book charge. In my base the 2036 numbers are EBITDA of $16.0B, depreciation of about $4.0B on my estimate and interest of $3.2B, which leaves about $8.8B of taxable income and $2.1B of tax at 23.7%, or 13% of EBITDA against about 2% in 2025 E (estimate). That is why my model lets cash conversion fall from 96.0% of EBITDA in 2027 to 85.2% in 2036, and the choice is worth about $12 a share against keeping it at 96%.

Balance sheet

At 30 September 2026 Williams had $2.20B of debt due within a year and $31.73B of longer debt, $33.93B in all, against $1.76B of cash, so net debt was $32.18B C (calculated). There was no commercial paper outstanding, and liquidity was $6.5B: the cash, a $3.75B credit facility to 2031 and a $1.0B 364-day facility.⁵ The covenant allows debt to EBITDA of 5.0 times, or 5.5 times after an acquisition, and the three ratings agencies rate it BBB+, Baa2 and BBB, two with positive outlooks.⁵ Noncontrolling interests were $5.76B at book, up from $2.19B in December, mostly the first $3.14B from the power investor.⁵ Interest expense was $1.13B in the nine months, and the 2026 guide is $1.54B.³ The September notes carry 5.0% to 6.4%, and the 2026 interest guide is about 4.9% on average net debt of roughly $31B by my estimate, so new debt pushes the average up, and my base case uses 5.5% from 2027 on, against about 4.9% implied by the 2026 guide.¹⁷³

My 3.90 times is a snapshot taken a month after a $5.1B purchase: Momentum’s EBITDA was in the twelve months for 28 days only. If I also count the $3.74B the investor had paid in by then as debt, the ratio is 4.4. The company’s 3.5 to 4.0 target, my 3.9 held through the model, and that figure are three different things.¹⁴³ One note in the 10-Q is worth the space: the company says certain assets bought in 2012 to 2014 may carry a value above fair value, though it judges them recoverable; an impairment would not change cash.⁵

VIValuation

First, how this goes badly wrong

Before the numbers, I asked what would most likely hurt an owner here, and worked back. Three things stand out.

• The growth costs more than it earns, and the balance sheet funds the difference. In the last twelve months the business spent $8.1B on capital against $6.5B of AFFO. If debt costs one point more on the whole stock and I require 10% instead of 9.0%, my base falls to $53; if growth capital costs 12 times the EBITDA it adds and not 8, it falls to $52 E (estimate). Stock was sold in 2017, shares paid for the Williams Partners roll-up in 2018 and for Momentum in 2026, and the share count is 2.3% higher than in February.⁶¹¹⁵
• The power projects disappoint. The company has put $5.34B of someone else’s money into five projects with 10 to 12.5 year contracts, and that investor takes 49% of the cash until it is paid back with a target return of 6.35%. If the five projects earn half what I assume, my base falls to $62; if they earned nothing and $1.5B of capital was lost, it falls to $55 E (estimate). Neither is likely; the first costs about a tenth of the price and the second about a fifth.¹⁴²²
• The multiple goes back. The stock sells for 14.9 times trailing EBITDA, against 9.7 to 10.7 times at the year-ends of 2019 to 2023. At 11 times on today’s EBITDA the stock would be about $42, and at the 10.2 times of 2022 and 2023 about $37 C (calculated). That is the way a good business becomes a poor purchase without a single bad quarter.⁶¹

None of these needs anything exotic. They set my bear. They also explain why I start from the cash that reaches an owner and not from AFFO, and why I test how much the answer depends on the growth budget and the cash tax date.

How I built it

I value the cash that could be paid to common holders after the company funds its own growth. I start from adjusted EBITDA for 2026 of $8.45B, in the upper half of the company’s $8.3B to $8.5B range, of which I put $0.17B in the power projects.³ After that, the core business’s EBITDA each year changes by an organic rate on existing assets, plus the prior two years of core growth capital divided by the capital it takes to earn a dollar of EBITDA, plus $0.35B in 2027 for the rest of a year of Momentum. I set that cost at 8 times in my base, which is 12.5 cents of EBITDA per dollar. The company does not publish a return on its backlog, so I tested it. From 2020 to 2025 Williams spent $21.0B on capital expenditures, acquisitions and equity investments, and adjusted EBITDA rose $2.73B from 2019 to 2025, which is 7.7 times on everything and 5.8 times if maintenance capital was near $0.85B a year C (calculated).⁶¹¹⁴ I use 8, which is worse than both. Part of that growth came from rate increases and a low 2020 base, and the next expansions are in corridors that are harder to permit, as the New Jersey ruling shows. Each turn on this input moves my base by about $5 to $6.

The five power projects are separate. Their total growth capital is $9.0B, which is the $4.4B the investor funds at 49% divided by 0.49.¹⁴ I assume they earn 17 cents of EBITDA on each dollar, with 40% of the run rate in 2027, 78% in 2028 and all of it in 2029. The anchor is Socrates South: $205M a year of fixed lease payments on $1.0B of gross assets, 20 cents before operating costs I can’t see.⁵ The investor takes 49% of the cash from those projects, which I take as 49% of 97% of project EBITDA, until the money it put in, $5.34B, is repaid at 6.35%, and in the base case that ends around 2038, with $1.2B still owed at the end of 2036, which I deduct from the value E (estimate).²² After the explicit years I keep 85% of the power EBITDA, since the contracts end between 2036 and 2041 and the customer holds the extension option.⁵

From EBITDA I take 96.0% in 2027 falling to 85.2% in 2036 for cash taxes, the cash I count from equity investments and other items, which is how the 2026 guide is built (AFFO $6.375B, EBITDA $8.4B at the midpoint) with tax drifting up as deferred tax comes due E (estimate).³⁷ Then I take off interest at 5.5% on net debt, preferred dividends, minority distributions of $0.33B growing 4% a year, stock pay of $0.10B and the investor’s share. That gives AFFO, which in my model is after stock pay and so a little below the company’s own measure. I then take off maintenance capital, $1.0B in 2027 and growing 3.5%, and growth capital, which starts at $4.6B for the core plus $3.3B for the power projects in 2027 and fades to $3.0B for the core, and add the new debt that holds net debt at 3.9 times EBITDA after the investor’s share. What remains is owner cash. I discount it at 9.0% with end-of-year timing, add the December 2026 dividend of $0.525, and divide by 1,255M shares. Net debt at the end of 2026 is $32.8B E (estimate): the $32.2B at 30 September, plus about $0.6B of the fourth quarter’s dividend, growth and maintenance capital over AFFO and the investor’s October $1.0B.⁵³

Two things in that setup matter. The value depends only a little on the size of the growth budget: with $2.0B of core growth capital every year my base is $64, and with $6.0B every year $72 C (calculated). And owner cash is about $2.11 a share in 2027 and $2.75 in 2028, level with the $2.10 dividend only because the last $0.6B arrives from the investor, and rises to $4.47 in 2029, when the power building ends. I don’t assume the company stops spending; I assume the spending earns a fair return, which comes to the same value E (estimate).

What the price implies

At $67.86 Williams has a market value of $85B. With net debt, minority interests and preferred shares the enterprise value is about $123B, or 14.9 times trailing adjusted EBITDA, 14.6 times my 2026 figure and 12.9 times trailing AFFO per share C (calculated).⁵¹ The multiple of EBITDA at each year-end since 2019 was between 9.7 and 10.7 through 2023, then 13.5 and 13.6; at the end of 2025 the share price was 12.6 times AFFO per share, against 6.7 to 8.7 in 2018 to 2023 C (calculated). So the price is high against the company’s own history, and what the market is paying for is the next ten years: the power projects, LNG and the Southeast build-out. The right question is what must be true, so I ask it in reverse.

I did not set my assumptions to land here. I set the growth capital cost at 8 times after the record test above, the cash tax path from the tax note, and the power yield from Socrates South, and the base landed at $67.9. It is a coincidence that this is within a few cents of the price, and the right way to read it is as a range: the grid and the table below show that a single turn of growth-capital cost, or a point of required return, moves the value $5 to $14. The price sits inside the error of my own estimate.

What I can’t judge

Four inputs decide this page and I can’t know any of them. I don’t know what the next $4B to $5B a year of growth capital will earn, because the company does not say. I don’t know when cash tax arrives, because it depends on how long the company keeps building. I don’t know whether the power customers extend after 10 to 12.5 years, or whether the rest of the 6 gigawatts of backlog is signed on the same terms. And I don’t know what return owners need to ask for. So here is the grid. Read my $68 as one cell in it.

The discount rate

I use 9.0%. The US ten-year Treasury yielded 4.90% on 4 November and the thirty-year 5.28%.²³ With Damodaran’s 4.41% implied equity risk premium on 1 November and a five-year beta of 0.63 for Williams against the S&P 500 on month-end closes, the capital asset pricing model gives 7.7% C (calculated).²⁴¹²⁵ With a beta of one it gives 9.3%. I choose between them. A beta measured over five calm years doesn’t say what happens to a company with $33B of debt and a build programme when long rates move, and the thirty-year is above 5%. I could not find Transco’s allowed return in the documents I opened, so I can’t use it as a floor. The same 9.0% is on the Enbridge page, where the risk-free rate is 1.1 points lower, so my standard here is a little less demanding, and the table below shows other rates.

A cross-check against Williams’s own history. Enterprise value over adjusted EBITDA was between 9.7 and 13.6 at the seven year-ends from 2019 to 2025 C (calculated).⁶¹ At my base value the same measure on trailing EBITDA is 14.9, at my bear 9.8 and at my bull 21.3. So my base sits at the top of what the market paid before 2024 and above the 13.5 and 13.6 of the last two year-ends, on a business whose EBITDA is growing faster. My bear is below every year-end in the record and my bull is far above it. The multiple in my base at the end of 2036 is 11.4 times, so my base lets today’s premium fade.

Other uses of the same dollar

The ten-year US Treasury pays 4.90% with no risk to the money R (reported).²³ Williams’s dividend alone yields 3.1%, before tax. My base returns 9.0% a year at this price, my bull 15.5% and my bear 0.2%, so even my base, which assumes the growth works, returns a little less than the 4.90% plus the 4.41% premium I use for risk. The closest comparison on this site is Enbridge: a price 6% below its base, 9.9% a year on its base cash flows, 2.3% on its bear and a 5.6% dividend yield.¹⁸ On my numbers Enbridge’s price leaves more room than Williams’s, and it pays almost twice the yield; Williams has the stronger growth and Enbridge has more debt. What I can’t say is that the same dollar does better in one than the other, since the bears and bulls on both pages are my guesses about inputs I said I can’t judge.

The margin of safety

• Price$67.86Close, Q4 2026 [1]
• Base value$68My estimate E (estimate)
• Margin of safety below$45set from my bear case C (calculated)

I set the margin of safety from the bear case, because with four inputs I can’t judge a discount to base is no protection. Below $45, 34% under my base value, my $34 bear value is at least three quarters of the price paid. The stock last closed below $45 on 26 September 2024 ($44.74).¹ At $67.86 it is above that line, about 51% above, so there is no margin of safety by this rule.

Two different questions

How good is the business, and how good is the price? On the first, this is one of the best-placed gas infrastructure companies in the United States: regulated pipes into the East Coast, gathering in the fastest-growing shale area, a record of meeting guidance, an investment-grade balance sheet and a visible slate of projects. Its weaknesses are that it spends more than it earns after capital, that a dividend at 40% of AFFO is easy to pay only because growth is optional, and that per-share growth of 7% a year since 2018 was helped by cheap tax deferral. On the second, at $67.86 the price is at my base, my bear is 50% under it, and at 14.9 times EBITDA it is above anything in the record before 2024. That looks like a very good business at a fair price, not a price that protects against error, and for a company borrowing to build, the second matters more.

Where this could be in 5 and 10 years

These are my values in the page’s own model, in US dollars of the year named. Value at a date is the present value, at that date and at 9.0%, of the owner cash that comes after it, less what is still owed to the power investor, divided by today’s 1,255M shares. Owner cash paid per share is the December 2026 dividend plus all owner cash after capital from 2027 to 2036. The return is the yearly rate at which today’s price grows to the table’s cash plus the end-2036 value. I assume no new share issues.⁵

For the base to stand at five years, AFFO has to reach about $6.61 a share by 2031 from $5.25 now, 4.7% a year, and adjusted EBITDA $12.7B by 2030, 10.4% a year from 2025 against the company’s 11% or more. At ten years it needs EBITDA of $16B and net debt of $60B at about 3.9 times. The most likely way my base is wrong is the one the grid shows: growth capital at 10 times gives $59, and faster cash tax pulls it toward $63.

These are estimates of value, not forecasts of the share price, and the market price can sit above or below value for years.

VIIThe case for the price

What most people miss

Three things cut in Williams’s favour. The power investor’s target return is 6.35%, with a 10% promote to Williams and a right to buy it out between years 7 and 14, while Socrates South pays about a 20% revenue yield on cost.¹⁴²²⁵ Contracted revenue is $32.5B, $10.5B of it Transco’s after 2030.⁵ And cash tax was $162M in 2025 against an $857M provision, with $5.17B of deferred tax to come; that flatters today’s cash and not tomorrow’s, so I put it in my conversion path.⁷

The bull case

Gas demand is rising from power, LNG and industry, and Williams owns the pipes that connect supply to it, on contracts that pay whether or not gas flows. The company cites Wood Mackenzie for US gas demand growing nearly 35% over the next decade, and the EIA expects US LNG exports of 18.6 billion cubic feet a day in 2027, up from 17.6.⁸¹⁶ The company says it has more than 6 gigawatts of power backlog and targets 11% or more a year of adjusted EBITDA growth to 2030; Chad Zamarin repeated that on 2 November.¹⁴³ That is management, which is how I treat it. A named outside view: Blackstone’s Robert Horn and Rick Campbell said in July that they share deep conviction in this area and that Williams supplies hard assets for the AI buildout. Their fund is the investor in the five projects, so it is a buyer with money in.¹⁴ I found no analyst view from the past month that I could check against its source, so I name none.

Against reality

Today’s numbers support the first half. Adjusted EBITDA is up 9% for nine months, the first power plant is in service, and the 2026 EBITDA guide was raised in August, mainly for Momentum.³²⁶ They lag on the second: the last quarter grew 6%, the share count is 2.3% higher than in February, cash after all capital is −$2.3B, and a court vacated the New Jersey water certificate for the 400 million cubic feet a day Northeast Supply Enhancement.³⁵⁶ The 11% target is also a step up from the 6.4% a year of the last nine years C (calculated).⁹⁴ If the plan is built at about 8 times the EBITDA it adds, the power projects earn what Socrates South suggests, cash tax arrives over the decade and debt stays near 5.5%, then $67.86 is possible and my $68 follows. Today: demand yes, first plant yes, court ruling and cash after capital not yet, the rest unknown.

Multiply it out for the bull, which needs 11% or more a year to 2030, growth capital at 6.5 times, power at 20 cents with full extensions, slow cash tax and debt at 5.1%. My guesses are 35%, 30%, 40%, 40%, 35%, which multiply to 0.6%. They overlap, so perhaps 5%, and the 25% weight I give the bull in my $70 weighted value is generous.

What a 10x would take. Ten times $67.86 is $679, about $852B of market value. At today’s 12.9 times AFFO that needs AFFO per share of $52 against $5.25, about $66B in all, which is 26% a year for ten years. With AFFO at 78% of EBITDA and EBITDA at 66% of $12.5B of revenue, that is about $84B of EBITDA and $128B of revenue, ten times today’s. AFFO per share grew 7.1% a year from 2018 to 2025 C (calculated) and my bull gives 90% in ten years. Doubling the multiple with no growth would still leave a fivefold gap. The best stretch on the price chart, from the $9.25 close on 18 March 2020 to $67.86, was 7.3 times in under seven years, from a pandemic low, with the multiple rising too. Tenfold moves at this size come from new markets, not pipeline tolls, and Williams pays out 40% of what it earns. Tenfold from here is close to impossible.

Where I come down

What moved me was the investor’s 6.35% money, which cuts what Williams pays for the first $9B of power, and the $32.5B contracted base, which is why my bear is not lower. The 11% growth target did not move me, because the last quarter grew 6% and the record is 6.4% a year, and neither did the multiple, which is the debate. What would raise my values is evidence on the inputs I can’t see: each turn of growth-capital cost is worth about $5 a share, and each year of delay in cash tax is worth a few dollars. At $67.86 the price is at my base, within the error of my own estimate, and above my $45 margin-of-safety price.

VIIIRisks, and what would change my view

The three ways this goes badly wrong are at the start of Valuation. These are the rest.

• Rates and refinancing. $2.2B of debt falls due within a year and $31.7B is longer.⁵ Every point on the average rate costs about $0.33B of AFFO, 5% of the total C (calculated).
• Equity. Stock was sold in 2017, paid out for the 2018 roll-up and for Momentum in 2026, and 26.9M Momentum shares were registered for resale on 2 November.⁶¹⁵
• Regulators and courts. The FERC sets the rate on the largest pieces, and a state or a court can stop a pipeline.⁵
• Customers and producers. The top ten gathering customers are about 55% of fees and the top ten pipeline customers 44% of regulated revenue; the power customers are not named in the filings I read, and Eagle Ford minimum volumes step down in 2026.⁷⁵
• Commodity and derivatives. Marketing and upstream swing reported earnings; the nine-month swing from derivatives was $116M.³⁷
• Physical damage. Offshore assets in the Gulf of America are exposed to storms.⁵
• Impairment. Some 2012 to 2014 acquired assets may carry values above fair value; a write-down would not change cash.⁵
• Leadership. A chief executive of 16 months, with the previous one as executive chair and a board that has grown from ten seats to twelve.⁸¹⁹

Policy and outside risk, and what it means for the value

Williams has no foreign operations.⁵ Its exposures are policy and prices abroad: LNG depends on export permits and global buyers; the power projects depend on turbines and equipment subject to tariffs and long lead times, with $857M of cash already spent on long-lead equipment that the customer reimburses if an order is cancelled; and the pipelines depend on state and federal permits, as the 8 September New Jersey ruling shows.⁵ I have not found Williams’s own estimate of its tariff exposure, so the effects below are mine.

Some of this may be in the price: the stock fell 15% from May’s high close, and I can’t say how much of that was these.¹ Any single scenario costs $3 to $6 a share; several together with a higher rate are what hurt. Hurricanes and cyber attacks are also in the filings’ risk lists, and I cannot size them.⁵