Williams Companies: A Tollgate Worth Watching, Not Yet Touching
The Business, and Whether We Can Know It
Williams Companies is a midstream natural gas business that handles roughly a third of the gas in the United States — about 33 Bcf/d. Its crown jewel is Transco, an interstate pipeline running from the Gulf Coast to New York that carries around 15% of U.S. gas consumption. Revenue comes mostly from long-term fixed-fee contracts and FERC-approved cost-of-service rates. In plain terms: Williams sells the ability to move gas, not the gas itself. Commodity prices matter far less than volumes.
That is a tollgate, and tollgates are knowable. Public filings — the 10-K, 10-Q, earnings materials, FERC rate documents, and 2026 guidance (EBITDA of $8.05–8.35 billion, growth capex of $7.0–7.6 billion) — are sufficient to reconstruct the business, its finances, and its competitive position. Predictability passes too: the claim that the East Coast will still need gas for power generation, LNG exports, and data centers in ten years, and that it will move along Transco’s corridors, is analysis rather than a probability bet. Ironically, a regulatory environment that makes new interstate pipelines nearly impossible only increases the predictability of the existing asset.
One caveat: the recently enlarged growth capex program — much of it aimed at data center power — carries returns that are one notch less certain than the core pipeline business.
The Moat
The source of the advantage is efficient scale plus intangible assets: FERC certifications and established rights-of-way. Running a new interstate pipeline through densely populated corridors is, for practical purposes, impossible — blocked by cost, regulation, and local opposition. The market supports only the incumbents.
Pricing power exists, though with a ceiling. Cost-of-service rates guarantee a regulated return on invested capital and pass inflation through to rates. Transco expansion projects are pre-sold under long-term contracts before construction begins, and in oversubscribed corridors — the Southeast, the East Coast — negotiating leverage sits with Williams, not the shippers. This is not Seesaw-candy pricing power; it is regulated pricing power. But it is real.
The direction is widening. New competitor pipelines are getting harder to permit, gas demand from LNG exports and power generation is rising structurally, and expanding established corridors is far cheaper than building new lines. Growth stacks on top of the existing moat rather than diluting it.
The AI-era reassessment reinforces this. Data center power demand pulls gas-fired generation demand upward, and Williams has positioned itself as picks-and-shovels across gas supply, pipelines, and generation — including a 682MW Neo project and a $5.34 billion power joint venture with Blackstone. The catch, of course, is that the market has already paid for this insight: the stock trades at a P/E of 33x against a five-year median of 24.8x. The moat strengthening is fact; the price is the problem, and we will get there.
Management
The record has clear light and shadow. The stain: the failed 2016 Energy Transfer merger and a 69% dividend cut (from $0.64 to $0.20 quarterly), the product of excessive leverage and an overreaching deal. The recovery: a decade of restored discipline since — leverage held near the industry-norm target of ~4.0x, dividend growth restored to a 5% CAGR, a $1.5 billion buyback program launched in 2021 deliberately designed for opportunistic execution during price drops, and 2025 dividend coverage (AFFO basis) of 2.40x. Cutting the dividend rather than dodging it in 2016 was, whatever its causes, an act of financial honesty.
The watch item is new. Alan Armstrong handed over the CEO role in July 2025 after fourteen years, to Chad Zamarin. The new chief has roughly 2.7x’d growth capex — from $2.6–2.9 billion in 2025 to $7.0–7.6 billion planned for 2026 — a large bet made at what may be the top of a data center power cycle. Empire-building risk deserves monitoring. Offsetting this, the Blackstone joint venture externalizes some capital burden, which is a balanced move.
The retained-earnings test passes over the past five years: roughly 30% annual shareholder returns, with Transco expansions converting into EBITDA growth (2026 guidance implies +6%). Fair warning: a meaningful portion of that return is multiple expansion, not operating result. On candor, guidance has historically been presented conservatively and beaten. The negative signal: $5.3 million of insider selling over the past three months, and no buying.
Price
First, a data correction worth flagging. The screening value of $72M in owner earnings is a distortion — it subtracts growth capex in full. Normalized, using only maintenance capex: net income of $2.79 billion, plus D&A of ~$2.2 billion, minus maintenance capex of ~$0.9 billion (2026 guidance of $850–950 million; even conservatively rounded to $1.0 billion), yields roughly $4.0 billion. Cross-checks: 2025 AFFO of $5.86 billion (a generous measure), and operating cash flow of $5.9 billion less $0.9 billion maintenance capex, about $5.0 billion. A conservative range of $4.0–4.5 billion in owner earnings is fair.
That puts the owner-earnings yield at $4.0 billion / $91 billion, or about 4.4% — not meaningfully above the ~4.3% ten-year Treasury.
Valuation: this is a low-growth (EBITDA +5–6%) regulated-asset business carrying $30.3 billion of debt (D/E ~2.3 — the screening data’s “0.0” is an XBRL tagging error; the 19% ROE is leverage-amplified). Generous multiples are hard to justify. A low-growth 12x on $4.0 billion gives $48 billion; an optimistic 15x on $4.5 billion, assuming growth capex performs as planned, gives $68 billion. Intrinsic value: $48–68 billion, or $39–56 per share.
The market price is ~$74.5, a market cap of $91 billion — roughly 34% above even the optimistic scenario. There is no margin of safety; there is a negative one. The 33x P/E fully prices the data center narrative. A target buy price of ~$33 per share (about a $40 billion market cap) applies a 30% discount to the midpoint of the value range.
Verdict
Watch. Gates 1 through 3 pass cleanly — a BNSF-shaped tollgate business, a widening moat in a world that no longer permits new ones, and a management team with a recovered, if recently tested, capital allocation record. Gate 4 fails, and fails decisively: the market is paying, today, for a bull case that hasn’t happened yet. The proposition “AI increases gas demand” can be entirely true without making the current price sensible. Until the price approaches ~$33, this is a good business at a bad number.
What would change the verdict: sustained data center demand that pushes owner earnings toward $6 billion with growth capex earning above regulated returns — or, on the downside, evidence of discipline breakdown (uncontracted construction, leverage drifting above 4.0x, or the new CEO repeating 2016) that would make even $33 too generous.
This analysis is AI-generated, educational, and not investment advice. Figures may contain errors or be delayed. Disclaimer