Reverse DCF: What a 35 P/E Actually Demands

Thirty-five times earnings sounds like an opinion. It is actually an equation with the growth rate hidden on the far side of the equals sign, and almost nobody solves for it before buying. I used to argue about whether a stock deserved its multiple by comparing it to the sector average, which settles nothing, since a whole sector can be overpriced together. A friend asked me last year to defend a position I held at 35 times earnings, and when I tried, I realized I had never put an actual number on what I was betting the business would do for the next ten years. That gap is what a reverse discounted cash flow model closes.

A normal DCF asks you to forecast cash flows for a decade, pick a discount rate, pick a terminal growth rate, and it hands back a fair value. Small changes to any of those three inputs move the answer by a wide margin, which is why two people using the same spreadsheet template can land 40 percent apart on the same company. I stopped trusting that output years ago. What I do now, especially with the expensive names I am tempted to buy, is run the model in reverse: hold the price as the known number, hold the discount rate at something defensible, and solve for the one variable actually in question, the growth rate the market needs to be right about.

A reverse DCF does not produce a valuation. It produces a checkable claim about the future, and a checkable claim is worth more than a fair-value guess nobody can verify.

Every multiple is a growth claim, not a verdict

Say a company earns $10 a share and trades at $350, a P/E of 35. Instead of debating whether that multiple is fair, reverse DCF turns it into a specific question: does this business’s earnings power plausibly compound at some rate for the next decade, and can I defend that rate with something other than a story. Cintas has come close to sustaining that kind of run for two decades, and even a business that steady needed a believable growth case to justify 40 times earnings, a case I walked through the arithmetic on earlier this year. The same method applies to any name trading at a premium.

The advantage over a normal DCF is that a growth rate is falsifiable in a way a fair-value estimate is not. Nobody can prove a stock is worth $420 instead of $380. Plenty of people can look at a 20 percent growth assumption and ask whether a company of that size, in that industry, has ever actually grown that fast for ten straight years.

The mechanics, with round numbers

Take the $10-EPS, $350-price example again. Assume a 9 percent discount rate, close to the long-run cost of equity for a large, diversified business, and a terminal growth rate of 3 percent starting in year eleven, roughly long-run nominal GDP growth. Build ten years of earnings compounding at a trial rate, discount each year back to today at 9 percent, add a discounted terminal value, and adjust the trial rate until the total matches $350.

Solved properly, the number that comes out is about 12.3 percent a year for a decade, not the round 14 percent a back-of-envelope guess tends to produce. That gap matters. Earnings compounding at 12.3 percent for ten years grow to roughly 3.19 times the starting base, or $31.90 on a $10 base. At 14 percent they grow to 3.71 times, or $37.10. That five-dollar difference in year-ten earnings is the entire disagreement between a defensible growth case and an aggressive one, and it comes from rounding a single input.

Three checks before I trust the number

Once I have the implied growth rate, I run it through three checks before deciding anything. Has any company of a similar size sustained that rate for a full decade, not two good years surrounded by average ones? Does the number sit meaningfully above the industry’s underlying unit or volume growth, since margin expansion alone rarely closes a ten-point gap by itself? And what happens if I move the discount rate by a single point in either direction, because a growth rate that only survives at 8 percent and collapses at 10 percent is not one I want to depend on?

That last check moves more than people expect. Push the discount rate in the $350 example from 9 to 10 percent and the required growth rate jumps from about 12.3 percent to roughly 14.7 percent, a swing of nearly two and a half points from a single-point change in an assumption nobody can observe directly. If a stock’s growth case only clears the bar at the lower discount rate, I treat it as a much thinner margin of safety than the headline multiple suggests.

Running the arithmetic on three real multiples

The same solve works on real companies, using each one’s actual price and trailing P/E rather than a round example. I hold the discount rate at 9 percent and the terminal growth rate at 3 percent across all three so the comparison is apples to apples, which is not how I would size a real discount rate for each business individually, but it isolates what the current multiple alone is asking for.

Company	Price / trailing P/E	Implied 10-year EPS growth
Microsoft	around $493.78 at 27.5x	about 9.2% a year
Cintas	around $197.64 at 40.3x	about 14.2% a year
Costco	around $895.31 at 45.0x	about 15.6% a year

Microsoft’s number is the one I find easiest to believe, since 9.2 percent is below what its own cloud and AI-linked revenue has been doing, a case I laid out in more detail when I looked at the stock at 27.5 times earnings. Cintas asks for more, 14.2 percent, which is a real stretch for a uniform-rental business unless steady margin expansion and buybacks keep doing the work they have done for years. Costco’s 15.6 percent is the hardest claim of the three on revenue growth alone, and I think the honest reading, one I went into when I wrote about paying up for a warehouse club, is that the market is pricing in membership-fee increases and international unit growth that have not been announced yet, not just same-store sales trends.

Where the model breaks

The model has one structural blind spot: it treats growth as smooth, compounding earnings per share, and it cannot tell you whether that growth comes from more revenue, fatter margins, or fewer shares outstanding. Those three sources are not equally durable. A company buying back 4 percent of its shares a year can hit a demanding reverse-DCF growth target on mid-single-digit revenue growth alone, so a heavy repurchaser can clear what looks like an aggressive bar without the underlying business growing anywhere near that fast.

That is my specific counter-case, and it is a real risk to the argument above. If I ran Cintas’s implied growth rate using per-share earnings instead of assuming it all comes from revenue, and layered in its actual buyback pace, the underlying revenue growth required would come in several points lower than 14.2 percent, and I would have been wrong to call the multiple as demanding as the headline growth number makes it look. The model tells you what growth the price needs. It does not tell you which lever pulls it, and that second question is the one that actually decides whether I trust the story.

The number I watch now is not the P/E by itself. It is the gap between the implied growth rate and a company’s own trailing decade of revenue growth. Under five points, margin trends or buybacks usually explain the difference on their own. Past ten points, I want a specific reason before paying that price, and without one I would not hold the position.

I keep the model deliberately simple on purpose. A discounted cash flow with a dozen moving parts feels more rigorous, but every extra input is another place to smuggle in a conclusion you already wanted, by nudging a margin assumption or a share-count forecast until the fair value matches the price you were hoping to pay. Holding the discount rate and terminal growth rate fixed and solving for one number removes that temptation, because there is nowhere left to hide the bias. The trade-off is that the single growth number is cruder than a full model, which is why the three checks above exist: they are the manual work a bigger spreadsheet would otherwise do for you, and skipping them is how a reverse DCF turns from a discipline into just another way to confirm what you already believed.

Worth saying plainly: none of this tells you when a stock will re-rate, only what has to be true for the current price to make sense over a decade. A stock can clear its implied growth rate for years and still go nowhere if the multiple compresses, or miss it for a year or two and still work out if the market is patient. The model is a filter for buying decisions, not a timing tool, and I do not use it that way.

Analysis and opinion only, not investment advice. Price and multiple data for Microsoft, Cintas and Costco reflect figures compiled from public filings as of the date below; underlying filings are on SEC EDGAR for Microsoft and Costco, and the reverse-DCF math above was checked on September 23, 2026.