If something cannot go on forever, it will stop.
— Herbert Stein, former Chairman, U.S. Council of Economic Advisers
Download the Slides Deck: The_Compounder_Crossroads
Listen to the deep-dive discussion – The internal physics of business compounding
I. The Same Label on Different Businesses
Coca-Cola and PepsiCo sell drinks to the same world, ride the same wave, and spend almost exactly the same share of revenue on capital — 4.4 percent for one, 4.3 percent for the other. Over a decade Coca-Cola’s revenue grew fourteen percent while its return on invested capital doubled, from roughly nine percent to nearly nineteen. PepsiCo’s revenue grew fifty-four percent while its return on invested capital did not move at all: fifteen and a half percent at the start, fifteen and a half percent today. One improved what it earned on a base it had stopped enlarging. The other enlarged its base without improving what it earned on it. Nothing in those numbers says which is the better business, and this essay will not say either — but the two now face entirely different futures, and that difference is the subject.
Now take one company rather than two. Coca-Cola in 1970 was a growth story: a wave the world was still discovering, and a bottling system that could swallow capital as fast as it could be deployed. Coca-Cola today grows volume at roughly two percent, spends fewer dollars on capital equipment than it did ten years ago, and hands two-thirds of its earnings to shareholders. It is the same brand, the same moat, and an entirely different machine.
Both observations resist the vocabulary normally used for them, because the same word is doing too much work. Coca-Cola and PepsiCo are both called compounders. So are Visa and Union Pacific, which have almost nothing in common beyond the label. The word describes what happened to the earnings and says nothing about the machine that produced them — and the machine is what determines where a business can go next, which of its options destroy value, and which of them only look as though they do.
II. Rate, Reinvestment, and Duration
Three measurements describe any business. The rate: what it earns on the capital it employs, measured as return on invested capital rather than return on equity, since leverage inflates the second without enlarging the base it is earned on. The quantity: how much capital it can actually put back to work at that rate. The duration: how long it can keep doing so before the opportunity or the protection runs out. They multiply, so all three must be present in some degree for anything to compound at all — and no business holds all three at their maximum, because growth that must be bought lands on the balance sheet and growth that arrives free does not.
None of that is a discovery of this collection. Miller and Modigliani established in 1961 that what an investor capitalises is the opportunity a firm has to make further investments yielding more than the normal market rate. The identity that growth equals return on capital multiplied by the share of earnings reinvested is ordinary corporate finance. Michael Mauboussin has spent a career on the third measurement under the name competitive advantage period. Bruce Greenwald established that growth is worth something inside a franchise and nothing outside one. Michael Jensen, in 1986, named what becomes of a company holding more cash than it can profitably invest. And Chuck Akre’s three-legged stool sets the reinvestment runway beside the business and its management — the leg he has said became the most important of the three as he grew older.
One further reading tells you whether a position is healthy, and it is the most useful single question available. Write the business as two parts: A is the toll, the position that collects; B is the growth, the wave it rides and the capital that growth consumes. In a business of quality, A absorbs B — the toll generates more than the growth consumes, and what cannot be consumed is surplus. In a deteriorating one the flow reverses: B absorbs A, growth devours the toll, and the business drifts toward the barrier from which nothing returns.
Three questions establish the direction, and their order matters. Is capital actually going back into the business? Is it financed from within, out of operating cash flow, rather than raised from markets? And does the return durably exceed the cost of capital? The second is the sharpest, because quantity is the only component that can be bought on credit: no company can borrow a higher rate of return and none can borrow duration, but any company can borrow size. Duke Energy spends fourteen billion dollars against five billion of net income, runs negative free cash flow, issues debt and pays its dividend anyway — the base grows with someone else’s money, which is intermediation rather than absorption. The third is the floor beneath everything: if capital were free, more quantity would always be better and there would be nothing to trade off. Waste Connections is the purest illustration in this collection because nothing about it is hidden — real cash flow, a wide moat, disciplined management, and a return sitting almost exactly at its cost of capital. The spread is zero, and quantity multiplies nothing.
One trap remains. An average return can be carried entirely by capital deployed long ago: Canadian National earns a real spread on a network laid a century ago while its recent capital produces almost nothing. The average conceals the increment, and it is the increment a buyer acquires today.
III. Compounding Requires Somewhere to Put the Money
Used loosely, the word covers anything whose value rises. Used strictly, it means one thing: a business that redeploys its own earnings, inside itself, at a return above what the capital costs — so that the base grows out of what the base produced. That is the only arrangement in which the three components multiply rather than merely coexist, and it is far scarcer than the vocabulary suggests.
Linde shows what it takes, and it shows it through a stretch that looked like failure. It paid for Praxair in a merger of near-equals and spent six years digesting it — six years in which the combined return on capital sat below what either company had earned alone, and every quarterly report showed a business that had bought size and lost quality. It emerged with four points of spread intact on a base twice as large, funded throughout from its own operations rather than from markets. Union Pacific took the harder route still, improving a base it could not have bought at any price: the corridor was laid in 1862, the rights of way no longer exist to be acquired, and no amount of capital would reproduce it. Its money went into what the corridor could carry rather than into more corridor, and the rate rose as the base enlarged — from 13.4 percent to 16.2, with net income up sixty-nine percent on capital expenditure up eight.
Now count the businesses that fail the same test, because the count is the argument. Duke Energy raises the capital rather than retaining it — fourteen billion of spending against five billion of net income, negative free cash flow, and a dividend paid from borrowings. Canadian Pacific bought its base too dear, paying thirty-one billion in stock for Kansas City Southern and watching its return fall from 16.6 percent to 6.7, where it has stayed for five years. Waste Connections has every visible attribute of quality and a spread of essentially nothing, its return sitting almost exactly at its cost of capital. Canadian National earns a real spread that its recent capital did not produce, inherited from a network laid a century ago while today’s spending returns almost nothing on its own. Four ways to fail, each with an excellent business underneath, against two that pass.
And here is the part that makes the crossroads unavoidable rather than merely common. The businesses least able to meet the strict definition are the ones that earn the most. Visa carries a new fintech’s transactions across rails already laid and already paid for, so almost nothing lands on the balance sheet and the numerator rises while the denominator stands still. The rate is extraordinary for exactly the reason there is nothing to reinvest in. Such a business creates enormous value — intrinsic value rises without capital being consumed, which is the cleanest form of value creation there is — but it does not compound in the strict sense, because nothing accumulates inside it. Everything it earns beyond what it can use must leave.
So compounding sets its own limit. A high return generates surplus faster than a small base can absorb it, and the gap widens with success rather than narrowing. Even the businesses that do meet the strict definition meet it temporarily, and the reason sits in a single relationship. Internal growth is the product of two things: the return a business earns on its capital, and the fraction of its earnings it can put back into that capital. A young company might reinvest four fifths of what it earns at twenty percent, and grow at sixteen percent a year. The same company at maturity might earn thirty percent — a better rate than it ever managed — and be able to redeploy only a few hundredths of its profits, so that growth from reinvestment falls below two percent. The rate did not deteriorate. It improved. What collapsed was the fraction that could be reinvested. The engine does not lose power; it runs out of anything to turn. A base compounding at fifteen percent inside an economy growing at four doubles its share of that economy every eleven years or so. Extended far enough, the business becomes the economy — which is another way of saying that the runway was always finite, and that the better the business, the sooner it runs out. Nothing has to go wrong for this to happen. The law of large numbers arrives on schedule, and it arrives fastest for the companies that did everything right. And it is worth being exact about what stops. The business does not stop; it goes on growing at whatever rate the world around it grows. What stops is the compounding — the part that came from money the business put back into itself. That is the machine that runs out of work, and its owner is left holding the proceeds. Every compounder therefore arrives, sooner or later, at the same place — holding cash it cannot put back to work at the rate that produced it. What it does next is the subject of everything that follows.
IV. Every Business Runs Out of Places to Reinvest
A business does not stand in one place for its life. It occupies a different position at every stage, and the stages run in a sequence that is remarkably consistent.
At the beginning, B absorbs A. The toll does not yet exist, or does not yet cover what growth consumes, and the business burns more than it collects. Most end here. Then the direction inverts: the toll generates more than the growth consumes, and the surplus is large enough to fund the extension of the base. This is the phase the market calls a growth stock, and the pricing is not irrational — rate, quantity and duration all appear present at once. It is also the phase that cannot last, because the wave that made the growth free is finite and the base that absorbed capital eventually fills. Then maturity, and the sorting: some businesses still find capital to absorb above its cost, and go on compounding from within; others cannot, and must do something else with what they earn.
Coca-Cola ran the whole sequence, which is why it is the most useful single case in this collection. For decades the wave was free in the strongest sense and the base could take everything the company could deploy. Today the volume grows at roughly two percent on a two-year average by the chief executive’s own reckoning, the growth comes from price and mix, and the base has been deliberately sold. The A is as magnificent as it ever was. The B is no longer free.
Which introduces the distinction that decides everything at maturity, because it is not obvious from outside. A mature business that says it is investing for growth may be doing one of two different things. Enlarging the base means deploying capital that becomes an asset and earns a return on itself — new track, new plants, new capacity. It lands on the balance sheet, and the return on it can be measured. Buying share means spending that runs through the income statement to capture a larger portion of a wave that already exists — advertising, promotion, incentives paid to the parties who route the volume. It creates no asset, earns no measurable return of its own, and decays the moment it stops. The first is absorption. The second is the cost of the wave: the price of holding a position rather than the price of building one, invisible on a capital-intensity screen because only one of the two appears as capital.
This is not peculiar to consumer brands. Even the purest tolls pay for their share: Visa’s client incentives — the payments made to the institutions that issue its cards and route its volume — consume close to twenty-nine percent of gross revenue and have been growing faster than revenue itself. The market grows for free. The share of it is purchased, and the purchase recurs every year.
V. Return It, Buy a Base, or Rebuild
The sorting at maturity leaves one question, and nearly everything else in this collection is an answer to it: what does a business do with a surplus it cannot reinvest? Three answers cover most of what companies actually do.
The first is to return the surplus, which is the position accepted. Visa distributes and repurchases. MSCI repurchases with such persistence that shareholders’ equity has been consumed below zero. Coca-Cola pays out roughly two-thirds of earnings and has not reduced its share count in a decade. The business acknowledges that it cannot redeploy at its own rate and hands the problem to its owners. Few companies choose one answer purely: Coca-Cola also buys brands — Costa, BodyArmor, fairlife — so a fraction of its surplus takes the second route while the bulk takes the first. What matters is the proportion, and where the proportion sits tells you which answer the company has really chosen.
And returning the surplus does not end the compounding — it relocates it. A business that can no longer reinvest still compounds its earnings: the toll raises its price, the mix improves, operating leverage does its work, and intrinsic value rises without capital being consumed at all. What it no longer does is accumulate. Inside a business that can still absorb, the compounding happens internally and automatically, untaxed until sale, whether the owner acts or not. In one that cannot, it has been externalised: the cash arrives, taxed on arrival, and compounds only if the owner redeploys it at a rate the business itself could not achieve. Same phenomenon, different location, and the difference belongs entirely to the owner.
The second is to buy a base, which is the position denied. S&P Global paid forty-four billion dollars for IHS Markit and watched its return on invested capital fall from fifty-seven percent to ten. Canadian Pacific paid thirty-one billion in stock for Kansas City Southern, diluted its owners by twenty-three percent, and saw return on invested capital fall from 16.6 percent to 6.7, where it has stayed for five years. Neither bought a bad asset; both bought an excellent one. The arithmetic is what recurs: a purchased base enters the balance sheet at the price paid, premium and goodwill included, not at the value that produced the seller’s return, so the buyer’s return on new capital begins at the seller’s earnings divided by the buyer’s price. Pay a full multiple for an irreplaceable asset and the incremental return starts low precisely because the asset was irreplaceable — irreplaceability is what makes the premium expensive.
The third is to rebuild the machine itself, which takes a decade rather than a quarter. Between 2015 and 2018 Coca-Cola sold its bottling operations: revenue fell from 41.9 billion dollars to 31.9, a quarter of the company shed on purpose. A collapse in revenue normally crushes margins; this one did the opposite. Operating margin rose from 22.4 percent to 31.1, capital spending fell toward four percent of revenue, and return on invested capital roughly doubled. Coca-Cola sold its capacity to absorb in exchange for a higher rate on what remained — a deliberate move from a business that could reinvest to one that no longer needed to. PepsiCo made the opposite choice in 2010, buying its bottlers back and keeping the direct-delivery network that puts its products on the shelf. Its rate has not improved — fifteen and a half percent then, fifteen and a half percent now — but it sits comfortably above the cost of the capital it employs, and it has been earned on a base that grew by more than half. What PepsiCo kept is the ability to put money back to work inside itself, which is precisely what Coca-Cola gave up. Measured against the strict definition set out earlier, the company whose return doubled is the one that stopped compounding, and the company whose return stood still is the one that did not. Neither was wrong. They were choosing between different things. And Microsoft and Alphabet are now travelling in the other direction entirely, building physical bases of a size no railway ever contemplated, acquiring the capacity to absorb that a software business never had.
A business can accept where it stands, deny it, or move. What none of them can do is keep the rate and acquire the quantity at the same time, which is the constraint the whole essay has been describing.
VI. Buying a Base and Building One Look the Same
Denying a position and moving are almost indistinguishable from outside, and this is where the distinction has to be made deliberately. In both cases a high-return business commits enormous sums to enlarge its base; in both cases the reported return on capital falls; in both cases the market is told the company is investing in growth. Buying IHS Markit and building datacentres produce the same headline and the same compression of rate. The market has treated them alike, and for a defensible reason: at the moment the spending begins there is very little to tell them apart. The distinction has to be made in the accounts, and it has to be made before the returns arrive.
Three tests separate them.
Is the spending financed from within? Microsoft generated roughly one hundred seventy billion dollars of operating cash flow against ninety-seven billion of capital expenditure, issued no equity, left its share count unmoved and reduced long-term debt while building. Alphabet has crossed the line: free cash flow negative, tens of billions raised in equity and notes.
Does the new base bill third parties? A base that serves only its owner is a cost centre wearing the costume of a toll — its return arrives, if at all, as an improvement in something else, indirect and unattributable. Microsoft’s new capacity is contracted before it exists, in commercial obligations booked ahead of delivery. Meta’s infrastructure is consumed by Meta, bills no one, and is partly held off its own balance sheet.
Is it a programme or a transaction? Rebuilding the machine takes a decade, is paid for in size, and is announced as a retreat — Coca-Cola shrank by a quarter and said so. Buying a base takes a quarter, is paid for in premium, and is announced as growth. The first admits which position it is leaving. The second pretends the terrain is flat.
The honest qualification belongs here. Passing the three tests is not proof that new capital will earn its cost; it is evidence that the attempt is the kind that can. Microsoft’s current return remains an average containing decades of accumulated software, and the return on the new base is demonstrated by nothing yet. The tests tell you which journey you are watching. They do not tell you where it ends.
VII. What the Owner Ends Up Holding
Hold a business long enough and your return becomes its return. What passes to you is therefore not a number but a machine: a rate, a fraction that can be reinvested, and a duration — caught at whatever stage of its life the business happens to occupy when you buy it. Which is why the structure has to be identified rather than assumed, and why two businesses earning the same return on capital hand their owners different outcomes if only one of them can keep the money.
And the investor occupies a position of his own, which also moves. Consider the one most associated with Coca-Cola. He bought it in 1988, when it stood in the growth phase described above, and has held it since. What he has bought and built in the decades after is of the opposite kind: a transcontinental railroad, a regulated energy platform — capital-heavy bases that swallow enormous sums at returns that are respectable rather than extraordinary. Same investor, opposite ends of the same problem.
The explanation is not a change of taste but a change in his binding constraint. An investor deploying a few hundred million dollars is constrained by rate: capital can go anywhere, so it should go where the return is highest. An investor deploying hundreds of billions is constrained by quantity: most of the highest-returning businesses on earth cannot absorb what he has, and a modest return on a very large base beats a magnificent return on a base too small to matter. The scarce component chose the road. Nothing about the businesses changed at all. And the symmetry is worth stating from the other side, because it is the one advantage a small owner holds and rarely notices: the positions closed to the largest investor in the world remain open to almost everyone else. A constraint of size is not a constraint of skill, and it does not travel downward.
Which is why the crossroads matters more than the label. Knowing that a business compounds tells you very little. Knowing whether it has reached the point where it can no longer redeploy its own earnings — and which of the three roads it has taken since — tells you almost everything: what will arrive in your hands, what will stay inside, what is being risked, and whether the spending you are watching is construction or purchase. That judgment has to be made before the evidence arrives, which is the only moment at which it is worth anything.
What an owner buys is never a rate of return. It is a position, at a moment, already travelling.
Watch the video