Valuation in an Age of Fast Creative Destruction

In the 1840s, Britain built railways at a pace that looked like the future arriving early, because it was.

The trains ran. The freight moved. The country was permanently changed.

And a great many of the people who financed it lost most of what they put in, because they paid prices that assumed every line would earn well, forever, without competition.

Around 2000, enormous quantities of optical fiber were laid across oceans and continents. The internet did in fact need that fiber, and eventually used all of it.

The companies that laid it were restructured, merged or wound up first, and their shareholders did not wait around to be proved right.

A technology can be transformative and still be a poor investment at the price being asked.

Being correct about the future is not the same as being paid for it. What you are paid for is the gap between what you pay and what the business actually produces.

Cheap intelligence does not change that. It changes two things underneath it: which businesses will produce cash, and for how long.

Price Is What You Pay, Value Is What You Get

Price is a fact. It is on the screen, it is precise, and it is available every second of the trading day.

Value is an estimate. It is imprecise, it requires judgment, and it is never available on demand.

The entire discipline consists of holding the estimate steady while the fact moves around, and acting only when the gap between them is large enough to be worth something.

Most investors do the reverse. They let the fact of the price revise the estimate of the value, which is why a rising price feels like evidence.

A Multiple Is a Bet on Durability

Suppose you require a 9 percent annual return, and a business produces $1 of earnings per share that never grows and lasts forever.

That stream is worth $1 divided by 0.09, which is $11.10. Eleven times earnings.

Now suppose those earnings grow at 4 percent forever. The stream is worth $1.04 divided by 0.05, which is $20.80. About twenty-one times earnings.

So when a business trades at thirty times earnings, the price contains an assumption. At a 9 percent required return, thirty times implies growth of roughly 5.5 percent a year, forever.

Forever is doing a great deal of work in that sentence.

Now shorten the life

Take a business producing $60 million of free cash flow a year. At 9 percent, with no growth and no ending, it is worth about $667 million.

Now assume the same $60 million, but the moat lasts only ten years and then the cash stops.

The present value of ten years of $60 million at 9 percent is about $385 million.

Fifteen years gives about $484 million.

Read those three numbers together. The cash flow is identical in every year it exists. The only thing that changed is how long it lasts, and the value fell by 42 percent.

Durability is not a footnote to valuation. Durability is most of valuation.

What Cheap Intelligence Does to Durability

AI does not shorten every moat. It shortens some and lengthens others, and telling them apart is the work.

Moats built on cognitive labor are the ones at risk. If a business earns its returns because it employs many skilled people doing text, code, analysis or design, and a competitor can now do the same volume with a quarter of the headcount, the cost advantage was never a moat.

Moats built on physical bottlenecks tend to lengthen. Land, grid connections, generation capacity, port access, specialized fabrication: cheap intelligence increases demand for all of them and cannot manufacture any of them.

Moats built on trust, accountability and regulation also tend to lengthen, for the reason Part II gave. When output is abundant and cheap, the scarce thing is someone who will sign their name to it and carry the liability.

Moats built on switching costs and proprietary data hold up well, provided the data is truly proprietary and not simply a copy of what everyone else has.

The one-sentence test: if abundant intelligence is a substitute for what this business sells, its durability is shrinking; if abundant intelligence is a complement to what this business owns, its durability is growing.

Margin of Safety, Recalculated

Margin of safety is the oldest idea in serious investing: buy at a price far enough below your estimate of value that being wrong does not ruin you.

The size of the discount should scale with the uncertainty of the estimate, and uncertainty about durability has increased for a large share of the economy.

Work it through. If you estimate a business is worth $1,000 million and pay $900 million, your margin is 10 percent. A modest error in the growth rate erases it entirely.

If instead you pay $650 million for the same estimate, you have a 35 percent buffer, which is roughly what shortening the assumed moat from forever to fifteen years would have cost you.

That is the practical rule. For a business whose advantage depends on cognitive work, ask what it is worth if the moat lasts ten years rather than thirty, and treat the lower figure as the estimate.

For a business whose advantage is a physical or regulatory bottleneck, the usual discipline applies, and the main risk is that you pay a premium everyone else has already noticed.

Two Questions for Any Business in This Era

The first question is whether cheap intelligence is a cost saving or a price cut for this business.

If a company's costs fall 20 percent and its competitors' costs fall 20 percent, the saving is competed away into lower prices and the customer keeps the gain. That is the normal outcome in a market without a moat.

The saving is retained only where something stops price competition: a brand, a license, a switching cost, a bottleneck, a contract.

The second question is whether the company is a user of AI or a toll on it. Users get a temporary efficiency. Owners of the inputs, the infrastructure, the power and the distribution collect a fee on everyone else's efficiency.

Neither answer is automatically good news at the price being asked, which returns us to the railways.

Base Rates, Used Carefully

Three patterns recur across technology booms, and they are the closest thing to a base rate an ordinary investor can use.

First, the technology usually delivers, roughly on the expected scale, but later than the enthusiasm assumed.

Second, the economic surplus tends to flow to customers and to owners of complementary bottlenecks rather than to the technology providers competing with each other.

Third, the distribution of outcomes among the providers is brutally uneven, and it is not obvious in advance which ones survive.

The conclusion is not to stay away. It is the reason Lesson 21, Engine One: Own the Machines, argued for broad ownership first: broad exposure captures the winners without requiring you to identify them.

Concentration should be reserved for the small number of businesses you understand well enough to have an opinion about durability, which for most people is a very small number indeed.

What the Three Readers Do

Maya

Maya, 38, holds a meaningful share of her $60,000 in retirement accounts through her employer's plan and has begun receiving some equity in the software company she works for.

She applies the durability test to her own employer and reaches an uncomfortable conclusion: the campaign production her team once did is the part being automated, and the switching cost of customer data is the part that will still be there in ten years.

That does not tell her to sell anything. It tells her that her employer equity and her salary are exposed to the same single business, and that concentration is a risk she is carrying without having chosen it.

Her decision is to keep the equity she is granted, to stop adding to it voluntarily, and to direct her new surplus into broad ownership instead.

Tom

Tom, 47, values his own review service using the same arithmetic he would apply to a listed company.

It produced $21,300 of return on capital last year after he paid himself a market wage. A buyer would not pay twenty times that.

The reason is durability. Three clients produce 70 percent of the revenue, the relationships are with him personally, and the process lives in his head rather than in a documented system.

At a 9 percent required return, a buyer who believed the cash would last four years would pay roughly $69,000, which is about three and a quarter times earnings.

Tom now knows exactly what raises that number: more clients, written contracts with notice periods, a documented process another reviewer could follow, and a professional accreditation that makes his sign-off hard to replace.

Leo

Leo, 24, has been reading about a category of AI infrastructure companies and wants to put most of his $880 monthly surplus into two of them.

He does the multiple arithmetic instead. At the prices being quoted, he calculates the growth rate implied by a 9 percent required return and finds it requires better than 5 percent growth every year for decades.

He cannot say whether that is true, and the honest recognition that he cannot say is the valuable part.

His decision is to put the core of his surplus into broad ownership and to cap any single concentrated position at 10 percent of his invested capital until he can write a page on a business's durability that survives his own scrutiny.

Worksheet

1. Pick one business you are considering owning, or already own, and write its current price and its current annual earnings or free cash flow.
2. Divide price by earnings to get the multiple, and write it down before you form any opinion.
3. Using a 9 percent required return, calculate the growth rate the price implies if the business lasted forever, and write whether you believe it.
4. Recalculate the value assuming the cash flow lasts only ten years and then stops, and note the percentage difference.
5. Write one sentence naming the moat, then one sentence naming what abundant intelligence does to that moat: substitute or complement.
6. Decide the discount you require given that uncertainty, and write the maximum price you would pay.
7. Do the same exercise for your own business, paying yourself a market wage first, and write the three things that would most increase its durability.
8. Write the single observation that would tell you your durability assumption was wrong, and where you would see it.

Common Mistakes

Confusing being right about the technology with making money

The railway investors were correct about railways. The fiber investors were correct about bandwidth.

Correctness about the technology tells you nothing about the price, the competitive structure, or who captures the surplus.

Treating a high multiple as a verdict rather than an assumption

A multiple is not a description of quality. It is a compressed statement about growth and duration that you can decompose in about two minutes of arithmetic.

Assuming AI is bad for every incumbent and good for every newcomer

Incumbents with distribution, data and customer relationships often absorb a new technology and widen their advantage with it. Newcomers with a clever product and no distribution frequently do not survive contact with that.

Applying a perpetual model to a business with a shrinking moat

If the advantage rests on cognitive work that is being commoditized, a model that assumes cash forever is not optimistic. It is wrong.

Using margin of safety as a slogan

A margin of safety is a number. If you cannot state the discount you required and why, you did not apply one.

Letting a falling price do your analysis

A business whose moat is eroding gets cheaper on every screen as the earnings that support the price disappear. Cheapness measured against last year's earnings is not a margin of safety.

The RW Finance Perspective

RW Finance treats valuation as the last step, not the first. Price only becomes meaningful once you can describe the business, its quality, its financial strength and its moat.

That ordering exists precisely because of what this lesson describes. A valuation model is an amplifier of your durability assumption, and a model built on an unexamined moat produces confident nonsense.

Our Moat and Risk work asks the durability question directly: what protects these returns, and what would permanently impair this business rather than merely depress it for a year.

Valuation then compares an estimate of value against the market price, and the size of the gap is the margin of safety you are actually being offered.

The next lesson, The Twenty-Year Portfolio, stops examining businesses one at a time and assembles everything into a single allocation: a core of broad ownership, a satellite of concentrated ownership you understand, trust and data as unpriced assets, and runway as the stabilizer.

For readers who want the full treatment of statements, quality, moats, management, valuation and risk, Program 1, Learn to Think Like a Long-Term Investor, is where that work is taught properly.

Key Takeaways

• A transformative technology and a good investment are different things, and past booms delivered the first far more reliably than the second.
• Price is a precise fact and value is an imprecise estimate, and the discipline is refusing to let the first revise the second.
• At a 9 percent required return, a flat perpetual earnings stream is worth about eleven times earnings and one growing at 4 percent about twenty-one times.
• Shortening an assumed moat from forever to ten years cut the value of a $60 million cash flow from about $667 million to about $385 million.
• Moats built on cognitive labor are shortening, while moats built on physical bottlenecks, trust, accountability and regulation are generally lengthening.
• The test is whether abundant intelligence substitutes for what a business sells or complements what a business owns.
• A cost saving available to every competitor is competed away into lower prices unless something stops price competition.
• Margin of safety should scale with uncertainty, which means a larger discount for businesses whose durability now depends on assumptions you cannot verify.