The average annual amount of capitalized expenditures that the business requires to fully maintain its long-term competitive position and its unit volume… must be a guess — and one sometimes very difficult to make.
— Warren Buffett, Berkshire Hathaway shareholder letter, 1986
Download the Slides Deck: Capital_Longevity_and_Reinvestment
Listen to the deep-dive discussion – Why Successful Companies Have Negative Equity
I. MSCI Grew Twelve Percent While Returning Twice Its Free Cash Flow
MSCI grew its run rate by roughly twelve percent last year. In the same year it returned 3,041 million dollars to shareholders against 1,459 million of free cash flow — more than twice what it generated — and borrowed to do it. Shareholders’ equity now stands at minus 2.65 billion dollars.
Read those two facts together and something looks wrong. A business growing at twelve percent should need its money. Companies that grow consume capital; that is nearly the definition of growing. This one is growing quickly and giving away more cash than it earns, on purpose, while its balance sheet is consumed from the inside.
There is no contradiction. The two facts describe two different things that this collection has been calling by the same name for eighteen months.
II. The Growth Runway and the Reinvestment Runway Are Different Measurements
Every business has a growth runway: how long the demand it serves can keep expanding. Every business also has a reinvestment runway: how long it can keep putting its own earnings back to work at an attractive return. These are not the same measurement, and there is no reason for them to end at the same time.
Both runways exist everywhere. What changes from one company to another is who finances each of them and how fast each one advances. A snack manufacturer pays for its own growth runway every year, in advertising and shelf space, and stops growing the moment it stops paying. A consolidator buys its growth runway outright, one acquisition at a time. A railroad finances both runways out of its own cash flow. The interesting cases are the ones where the two are financed by different parties and move at different speeds — and those cases are easier to see at the extreme.
Both are durations, but they measure different objects: the growth runway measures the wave, the reinvestment runway measures the capacity to absorb.
We have been treating both as “quantity,” and that produced a mistaken picture. If a light toll cannot absorb much capital, the reasoning went, then it must be near the end of something — the runway must be short. That does not follow at all.
A capital-light toll does not arrive at the crossroads quickly. It is at the crossroads from the beginning, and it stays there while its growth runway keeps extending. The overflow starts early and never stops, and the volume beneath it keeps rising for years. Both statements are true simultaneously, which is exactly what MSCI’s numbers show.
The short runway is the reinvestment runway, not the growth runway. Everything else in this essay follows from keeping those two apart.
III. How Much Capacity One Dollar of Additional Business Requires — Almost None at Visa, a Full Dollar at Union Pacific
The relationship between the two runways can be measured, and it is the single most useful number in this essay. For any business, ask what an additional dollar of revenue requires in additional capacity. The answer is a coefficient, and it runs the whole range.
The distinction is not that some businesses absorb capital and others do not. Visa absorbs capital. Its capital expenditure went from 1,059 million dollars to 1,257 and then 1,482 over three years — it is rising, steadily, and it lands on the balance sheet like anyone else’s.
What matters is not how fast that spending grows but how little of the business it represents. Visa’s entire capital budget is 3.7 percent of net revenue and 6.4 percent of operating cash flow. Union Pacific’s is 15.5 percent of revenue and 40.8 percent of cash flow — four times the burden on the first measure, six times on the second. Both companies are spending. One of them has to spend in proportion to what it carries, and the other does not.
The growth rates can mislead here, and it is worth saying why. Visa’s capital spending rose about forty percent over three years while its revenue rose about twenty-two, which looks like the wrong direction until the amounts are attached: forty percent of a very small number is still a very small number. A company spending 1.5 billion dollars against twenty-three billion of operating cash flow can double its capital budget without changing what it is. The level is the elasticity; the growth rate is noise around it.
The default mental model assumes the two move together. To sell twice as much you need roughly twice the capacity. At Union Pacific that is close to true: ten percent more freight requires something like ten percent more capacity, so the denominator follows the numerator and the return earned on it comes down. At Visa, ten percent more payment volume requires perhaps two or three percent more capacity. The rate holds while the base expands.
So the property is not binary and it is not the absence of absorption. It is an elasticity: how much additional capacity one dollar of additional business requires. It is continuous, it differs from company to company, and it can be read from published statements. A business at the low end of that range has a capital problem precisely because it does not have a capacity problem.
IV. The Capital That Makes the Fee Possible Sits on Other Companies’ Balance Sheets
An earlier essay in this collection went looking for hidden capital in the expense lines — the research, the marketing, the software development that accounting may write off in the year it is spent. It found much less than expected. Across six companies, what produced lasting capacity had largely been capitalized.
The capital was somewhere else, and the answer is more interesting than the question.
Consider what has to exist for Visa to earn a fee. A bank must hold the account and issue the card. A merchant must install a terminal and connect it. A consumer must carry a phone able to run a wallet. A processor must build and maintain the software that routes the message. A telecommunications network must move it. Every one of those things is real productive capital, and every one of them is on somebody else’s balance sheet.
There is a second consequence, and it is the harder one to accept. What drives the growth is not for sale. Visa cannot buy the migration away from cash; it is produced by decisions made in billions of households and thousands of institutions. MSCI cannot buy the shift toward passive vehicles. A company in this position could sincerely want to invest in its own growth and find that there is nothing to purchase — not because it lacks imagination, but because the thing that produces the growth is not something anyone can sell it.
It is not hidden by an accounting convention. It is not in Visa’s expenses either. It is distributed across the balance sheets of thousands of third parties who never sent Visa a bill for building it. The company’s balance sheet conceals an economic base that expands for free — which is the precise opposite of the suspicion this line of inquiry began with.
V. Visa Paid for Its Network. Its Customers Pay for Its Growth.
Buffett’s 1983 observation was that when a business earns far more on its tangible assets than those assets could normally produce, the excess return is evidence of something the balance sheet does not contain, and the capitalized value of that excess is economic Goodwill.
It is worth being precise about what that something is, because the obvious answer is wrong. It is not the trend. The migration away from cash belongs to the world and would happen if Visa did not exist. No company owns a secular shift in behaviour, and no accounting of goodwill should credit one with it.
What it owns is a position between two ends of an infrastructure it does not have to extend. Visa built the middle — the processing network, the data centres, the protocols — and it paid for that, and still spends 894 million dollars a year running it. What it did not pay for is what sits at either end. The terminal belongs to the merchant. The card and the account belong to the bank. The phone belongs to the consumer. Each was bought by someone who wanted it for their own reasons, and each one enlarges what Visa can charge for.
That is the difference from See’s Candies, and it is worth stating exactly. See’s also paid for its base — the kitchens, the shops, the trucks — and Buffett could see how little of it stood behind the earnings. But See’s also paid for its growth: selling in one more city meant opening one more shop, financed by See’s. Visa paid for its network once. Its growth is paid for, in large part, by the people who benefit from it.
Same trace, read off the same ratio, applied to an object Buffett did not have in front of him in 1983.
VI. Five Companies, Five Different Pairs of Runways
Once the two runways are separated, businesses stop sorting into two camps and spread across a range instead.
| Growth runway | Reinvestment runway | Capacity per unit of growth | |
| MSCI | very long — two independent waves | negligible | lowest in the sample |
| Visa | long | small but real | roughly a quarter |
| Linde | moderate | long, contracted years ahead | close to one |
| Union Pacific | short | perpetual, mostly replacement | close to one |
| Microsoft | very long | enormous | above one — building before collecting |
Growth runway is judged by the wave; reinvestment runway by what the company can put back to work at its own rate. Capacity per unit of growth is the elasticity described in section III.
MSCI sits at one extreme because it rides two independent waves at once — the migration into passive vehicles and the expansion of private-market benchmarking — and neither requires it to supply the assets being tracked. Microsoft sits at the other, with a growth runway nobody can see the end of and a reinvestment runway so large that it is building the base before the revenue arrives, which is why its capital spending currently exceeds three times its depreciation.
Union Pacific is the interesting middle. Its growth runway is short — freight volume grows with the economy and grew one percent last year. Its reinvestment runway, however, never closes, because a network that exists must be maintained forever. Two billion of its 3.8 billion in capital spending is classified as replacement of existing infrastructure. A perpetual reinvestment runway is not the same thing as a long one.
VII. How Long the Capital Lasts Is a Separate Question Again
There is a third property, and it does not follow from either runway. Two companies can look alike on both — a real growth runway, a large reinvestment runway, a coefficient near or above one — and differ entirely in how long the thing they just bought keeps working.
| Life of the capital it deploys, and how much of it | |
| Union Pacific | Very long — corridor laid in 1862, still carrying freight |
| Linde | Long — plants of fifteen to twenty years, under contract |
| MSCI | Short, and negligible in amount — software and systems |
| Visa | Short, and small in amount — technology, three to four years |
| Microsoft | Short, and very large in amount — servers, two to six years |
Estimated useful lives as disclosed in each company’s filings.
Linde and Microsoft are the pair that makes the point. Both have somewhere to put their money, both are building assets that did not exist before, and both have demand contracted ahead of the capacity. On the first table they occupy nearly the same position. Yet Linde’s plants will be producing in twenty years under agreements already signed, while a meaningful share of what Microsoft is installing today will have to be bought again before the decade is out. Alphabet is further along the same line, spending 4.33 dollars for every dollar of depreciation against Microsoft’s three — building faster still relative to what it consumes.
This is not a criticism of either. It is a reminder that a long reinvestment runway and a durable asset are different things, and that a company can have the first without the second. What it changes is the arithmetic underneath: three hundred thirteen billion dollars of plant with a short average life creates a replacement obligation that arrives on schedule, whether the growth continues or not.
VIII. What to Check When Capital Spending Rises, and When a Return Looks Extraordinary
The distinction is only worth having if it changes what you look at. It does, in three places — and each one belongs to a different runway.
When capital spending rises, ask whether it is following the volume or lagging it. This is a question about the reinvestment runway. Compare the growth rate of capital expenditure with the growth rate of revenue over three to five years. If they move together, the denominator is following the numerator and the return earned will come down as the base expands. If spending grows meaningfully slower than what it carries, the rate can hold while the business gets larger. Visa spent forty percent more over three years while its revenue rose twenty-two — but on a base so small that the entire capital budget stayed under 1.5 billion against operating cash flow above twenty. The ratio matters more than either number alone.
When a return looks extraordinary, ask who paid for the base it is earned on. This is a question about the growth runway, and it is not answered anywhere in the financial statements. List what has to exist for one unit of the business to occur — the account, the terminal, the phone, the network — and ask whose balance sheet each item sits on. A business whose base is built by third parties can grow for as long as those third parties keep building. One that must supply the base itself grows at the speed of its own cash flow.
When the reinvestment runway looks short, do not conclude that growth is ending. This is the error the whole essay exists to prevent, and it is the one most easily made. A company with nowhere to put its money may still be riding a wave that has decades left. The overflow tells you about the second runway. It tells you nothing about the first.
One ratio helps with the first question and takes ten seconds: capital expenditure divided by depreciation. It is not a verdict — a figure near one can mean a mature business, a mix of asset lives, or simply the timing of a project cycle. What it does is separate a base that is growing from one that is merely being replaced, which is the difference between a reinvestment runway that extends and one that is perpetual without being long. Union Pacific spends 1.54 dollars for every dollar of depreciation, and two billion of its 3.8 billion is classified as replacement. Visa spends 1.21. Microsoft spends three.
Which is also a caution against reading that ratio in isolation. Visa’s technology assets are short-lived too — roughly 4.2 billion dollars of property, equipment and technology against 1.2 billion of annual depreciation implies an average life of three to four years, no longer than a server farm. The difference is not the nature of the capital. It is the amount. Microsoft carries 313 billion dollars of plant; Visa carries four. A short asset life creates a replacement burden only when there is a great deal of it to replace, which is the same point as the elasticity above, seen from the other end.
IX. The Crossroads Is Reached by Companies That Executed Well
This reframes something the previous essay treated too much as a problem.
Reaching the point where a business cannot redeploy its own earnings at its own rate is not a sign that something has gone wrong. It is the mark of a company that has executed well enough to earn more than its own opportunity set can absorb. The business that never reaches the crossroads is the one that never earned enough to face the question. Arriving there is an achievement, and returning the surplus is the answer that carries no execution risk at all. Its tax friction is the price of that absence of risk.
The three ways out — hand the money back, buy a base, rebuild the machine — are all ways of leaving the equilibrium, and each one introduces the possibility of doing it badly. That is why the record of companies that left is so mixed, and why the ones that stayed and simply overflowed have such uneventful histories.
None of which means the light toll escapes the constraint. It escapes it partially. Visa returns roughly twenty-eight percent of gross revenue in client incentives to hold its position in flows it does not own. MSCI has consumed its own equity to hand back what it cannot use. The surplus is still unabsorbable; what the second runway buys is not freedom from the crossroads but a long stay there, with the volume rising the entire time.
The question to ask of any business is therefore not how much capital it can absorb. It is how much capacity each dollar of growth requires — and who is paying to build the rest.
Watch the video