Capital intensity by itself, it’s part of the equation.
— Chris Hohn
Download the Slides Deck: Heavy Tolls Capital Absorbers
Listen to the deep-dive discussion – The Dangerous Illusion of High ROE
In 2025, Duke Energy reported a return on equity of roughly twelve percent. In the same year it spent fourteen billion dollars on capital expenditure against five billion in net income, generated negative free cash flow of 1.7 billion, and issued 11.9 billion in long-term debt. It paid its dividend.
Both facts are true. The return is real, in the accounting sense. So is the cash that left the building. A reader who trusts the first figure and never examines the second will conclude that Duke earns twelve percent on the capital its shareholders have committed. A reader who examines both will conclude something else: that the capital being deployed did not come from the business.
This is not an accusation of impropriety. Duke’s disclosures are complete, and Southern Company states plainly in its own filings that dividends, capital expenditure, and debt maturities will exceed operating cash flow through 2028. The model is announced. What the announcement reveals is that a category of business which promises to absorb capital at a maintained return—the capital-intensive toll, the regulated network, the infrastructure that cannot be replicated—frequently does no such thing. It absorbs the capital of others.
The distinction matters because absorption is the whole argument for owning such businesses. A heavy toll earns its place not through the rate it returns but through the quantity it can reinvest at that rate. The rate is modest by construction; that is the trade. What must be true, for the trade to be worth making, is that the reinvestment happens, that it happens with the company’s own money, and that the money earns more than it costs.
Three conditions. Almost nothing satisfies all three.
What Return on Equity Cannot See
Return on equity is the figure most often cited and the least suited to the question. It measures net income against shareholders’ equity, and both terms can be moved without touching the underlying economics.
Capital expenditure does not appear on the income statement. A utility that spends fourteen billion dollars on new transmission capitalizes the outlay; only depreciation, a fraction of it, reaches earnings. Net income is preserved while cash departs. The regulator then authorizes a return—typically nine to ten percent—on the enlarged rate base, so earnings rise mechanically with the spending, independent of whether any cash was generated to fund it.
Leverage completes the effect. Debt finances the expenditure without diluting equity. The numerator grows; the denominator does not. A return on equity of twelve percent, at a debt-to-equity ratio of 1.5, corresponds to a return on invested capital nearer five or six percent. The figure that looks like performance has been inflated at both ends.
AerCap shows the mechanism stripped of regulatory apparatus: a return on equity of 21.1 percent against a return on invested capital of 6.03 percent, and free cash flow negative in ten of the last thirteen years. The gap is borrowing, and nothing else.
Only return on invested capital can say what the capital earned. Only free cash flow can say where it came from. Both questions must be asked, and they eliminate different companies.
Three Ways to Fail
Duke Energy — the capital is raised, not retained. The rate base expands, earnings rise, and the regulator guarantees a return on every new dollar deployed. But the dollar was borrowed. Nothing has compounded; a larger base has been assembled with someone else’s money, and the shareholder’s claim on it has been encumbered accordingly. Duke’s ratio of funds from operations to debt stands near 14.8 percent against a 15 percent threshold for its credit rating—below the line for several years, while a 103 billion dollar capital plan proceeds. A business that absorbs so heavily that its solvency erodes is not in a comfortable position. It is in a race.
Canadian Pacific Kansas City — the base was purchased, and it does not earn. CP absorbed thirty-one billion dollars of Kansas City Southern assets, paid in stock: shares outstanding rose from 731 million to 901 million, a dilution of twenty-three percent. Its return on invested capital fell from 16.64 percent in 2020 to 6.70 percent in the first quarter of 2026 and has not recovered in five years. Against a cost of capital near seven to eight percent, the base it bought earns less than the capital that bought it. The incremental return—additional earnings divided by additional capital—is approximately four percent. Price-to-book collapsed from 6.06 to 1.99, which is the market recording that invested capital doubled while earnings did not follow.
The thesis offered for CP was that its depressed return reflected a base recently ingested and not yet productive, and therefore carried room to expand. The arithmetic supports the claim; five years of data do not. Synergies of 1.2 billion have been promised for 2024, then 2025, then 2027. A return that does not rise in five years is not a return awaiting expansion.
Waste Connections — the capital is retained, and earns nothing above its cost. This is the most instructive failure of the three, because every visible signal is favorable. Free cash flow of 1.4 billion, growing at double digits. Leverage of 2.75 times. Operating margin expanded from 14.2 percent to 19.2 percent across a decade. Morningstar assigns a wide moat and low uncertainty—the only company in this survey to receive both.
Its return on invested capital is 7.86 percent. Against a cost of capital of seven to eight percent, the spread is approximately zero.
The cause is visible in the strategy. More than five hundred acquisitions have been completed, nineteen in 2025 alone, each adding goodwill to invested capital without adding proportionate earnings. And the organic growth is composed of price, not volume: core pricing rose six percent in the most recent quarter while volumes declined 1.5 percent, following four quarters of steeper declines. The landfills are irreplaceable because no new landfills are permitted. That scarcity is the moat, and scarcity does not grow. It is monetized.
A company that absorbs vast quantities of its own capital at a return equal to its cost creates, over any horizon, exactly nothing. The base swells; the shareholder receives the cost of capital and no more. Cash flow, discipline, and a wide moat are not sufficient. There must be a spread.
Three Ways to Succeed
What follows is not the result of examining every capital-intensive company in the world. It is three that pass, and they pass differently. The differences are the substance.
Canadian National — the spread is real, and inherited. Return on invested capital of 12.6 percent against a cost near seven to eight percent: a spread of roughly five points, wider than either company that follows. By the standard governing this question, CN compounds.
Yet between 2016 and 2025 its return on invested capital declined from 15.6 percent to 12.6 percent, its operating margin from 44.1 percent to 38.1 percent, and its net income from 5.48 billion to 4.72 billion. Capital expenditure rose thirty-seven percent over the same period. Earnings per share moved from 7.24 dollars to 7.57—an increase of 4.5 percent across nine years, achieved only because nineteen percent of the shares were retired. Absent the repurchases, earnings per share would have fallen.
CN’s average return remains high because the network was built a century ago and has long since been paid for. The capital committed recently has produced nothing. The average conceals the incremental, and it is the incremental that the buyer of a share today acquires.
Linde — the rate holds while the base doubles. In 2018 Linde merged with Praxair, an acquisition that enlarged its base beyond recognition and diluted its shares by sixty-three percent. Its return on invested capital collapsed from 11.11 percent to 3.64 percent. It took six years to recover, and it recovered to 11.17 percent—almost precisely where it began.
This is not the disappointment it appears to be. The opportunity curve descends: the best projects are undertaken first, and absorbing more capital ought to dilute the return earned on it. To absorb an entire competitor and emerge with the rate unchanged is not stagnation. It is an achievement, and a rare one.
The arithmetic is unambiguous. Value created equals the spread multiplied by the capital invested. Linde earns 11.2 percent against a cost of seven to eight percent: four points, on a base that doubled. Operating income rose from 2.32 billion to 9.25 billion. Earnings per share rose from 5.21 dollars to 14.61—nearly triple, despite the dilution. Capital expenditure consumes fifty-one percent of operating cash flow, and free cash flow of 5.33 billion funds all of it from within.
The rate contributed nothing. The base multiplied what the rate produced. This is the ordinary form of compounding in a capital-intensive business, and it is already scarce.
Union Pacific — the rate rises while the base grows. Union Pacific retains 54.6 percent of its earnings and reinvests 3.79 billion dollars, forty-one percent of operating cash flow, in a network of thirty thousand miles. Free cash flow of 5.86 billion, twenty-four percent of revenue, funds the absorption entirely from within.
Its return on invested capital was 13.35 percent in 2016. It was 16.23 percent in the first quarter of 2026.
Across that interval net income rose from 4.23 billion to 7.14 billion—sixty-nine percent—while capital expenditure rose eight percent. Operating margin widened from 36.5 percent to 40.2 percent, which Morningstar places in the top decile globally. The moat rating has registered 1.00, its maximum, every month since 2021.
Both terms are moving. The base grows and the rate earned upon it improves. Value created rises by multiplication rather than by addition, and this should not be possible, because the marginal project is always inferior to the one before it.
Why the Second Engine Exists
Union Pacific does not purchase a base. It improves one.
The track was laid beginning in 1862. Capital committed today does not extend the network; it doubles a segment, lengthens a train, reduces an operating ratio. The marginal dollar buys efficiency in an asset already owned, already irreplaceable, and still imperfect.
Duke must build new lines. CP had to buy Kansas City Southern. Waste Connections must acquire competitors. Linde acquired Praxair. Each purchases a base, and a base purchased at market price earns, at best, the market’s return—which is why Linde’s rate did not rise, and why CP’s collapsed.
An existing network, unrepeatable and unfinished, permits absorption at an improving return. No amount of capital expenditure can manufacture it. It must already exist, and it must not yet be finished.
The Hierarchy
Three regimes, distinguished by what happens to the rate as the base enlarges.
Where the rate falls beneath the cost of capital, absorption destroys. CP doubled its base and quartered its return on equity; Waste Connections absorbs at no spread at all. Quantity multiplies whatever the spread contains, and when the spread is zero or negative, quantity multiplies nothing.
Where the rate holds above the cost of capital, absorption compounds. Linde doubled its base and preserved four points of spread; value created doubled while the rate stood still. This is the ordinary case, and it is already rare, because the opportunity curve descends and few businesses can absorb without diluting what they earn.
Where the rate rises as the base enlarges, both engines run. Union Pacific improves an asset it cannot buy and could not rebuild. The condition is not capital, and not management. The condition is possession of something already constructed, still imperfect, and impossible to replicate.
Canadian National satisfies the test and warns against reading it too simply. Its average return is the highest of the three, and its recent capital has produced nothing. A rate inherited from a century-old network is not a rate available to the capital committed this year.
The heavy toll is not rare because good businesses are rare. It is rare because absorption—financed from within, and returned above its cost—is the one thing a company cannot purchase.
Watch the video