# Cheap or Expensive? How Wall Street Actually Values a Stock
> We tested both ways Wall Street values a stock against the Walmart lost decade, the Nifty Fifty, and the Intel value trap — earnings up 164%, buyers underwater.
- Author: Barebone Research, Barebone AI
- Published: 2026-04-16
- Canonical: https://barebone.ai/resources/how-to-tell-if-a-stock-is-actually-cheap
- Publisher: Barebone AI (https://barebone.ai)
---
## The Fifty-for-One Illusion
On Tuesday, June 25, 2024, one share of Chipotle cost **$3,283.04**. The next morning it cost about **$65.66**.
Nothing happened to the company overnight. No earnings report, no scandal, no burrito recall. Chipotle had executed the first stock split in its three-decade history — a 50-for-1, one of the largest in New York Stock Exchange history. Every shareholder woke up with fifty shares where they'd had one, each worth a fiftieth as much. The pizza didn't get bigger; it got cut into more slices.
Everyone nods along at that example — then goes back to calling a $10 stock "cheap" and a $200 stock "expensive."
So this lesson takes the question where the obvious answer is wrong by default — how do you tell whether a stock is cheap? — and tests Wall Street's two methods, a quick one and a real one, against the historical record. The receipts: a retailer that grew earnings **every single year for a decade, +164% in total**, while its buyers lost money. A camera company priced at **94.8 times earnings**. And a chipmaker you could buy for **roughly ten times earnings** right before it lost **$18.6 billion** in nine months.
By the end, you'll know how the pros read a price — and exactly where each of their tools lies to them.
## Price Is a Sticker. Value Is a Number.
A stock's price is what the market charges for one slice of the company — and it says nothing about how big the slice is. A $10 stock with ten billion shares outstanding is a $100 billion claim; a $3,283 stock can be a mid-cap burrito chain. Cheap and expensive only mean something relative to one number: what the business itself is actually worth. Wall Street calls that **intrinsic value** — the value of all the cash a company will generate for its owners, counted in today's dollars.
Warren Buffett compressed the whole idea into one line in his 2008 letter to Berkshire Hathaway shareholders:
> "Long ago, Ben Graham taught me that 'Price is what you pay; value is what you get.'"
Price below intrinsic value: cheap. Price above it: expensive. Everything else — the dollar figure on the screen, the 52-week low, "it's down 40% so it must be a deal" — is sticker-reading. The craft is pinning down that one number, and there are two ways to do it.
## The Quick Way: What a Dollar of Earnings Costs
The fast method is comparing — lining a company up against similar ones using **multiples**, ratios that standardize price against something fundamental. The workhorse is the **P/E ratio**: price divided by earnings per share. It converts every stock, whatever its sticker, into the same unit — *how many dollars you pay for one dollar of the company's annual profit*.
We used Barebone to pull the trailing P/E for the five biggest names in American tech at their latest fiscal quarter-ends:
Read it in plain English. A dollar of Nvidia's profit costs about **$38**. A dollar of Microsoft's costs about **$22**. The S&P 500's average dollar of earnings goes for just under **$30** — meaning three of the five biggest tech companies in the world were trading *below* the market's average multiple.
What is a multiple, really? A compressed forecast. Paying 38 times current earnings only makes sense if you expect those earnings to grow into the price; paying 22 times implies humbler expectations. High P/E, growth priced in. Low P/E, stagnation expected — or decline.
Two rules keep the tool honest. Compare like with like — a bank at 12x against a software firm at 30x tells you about the economics of banking versus software, not about which is the bargain. And remember what comparing is: triage. It's how an analyst scans a thousand stocks before lunch and flags the handful priced out of line with their peers. It never tells you what a business is *worth* — if the whole neighborhood is overpriced, the cheapest house on the street is still a bad deal.
That limitation isn't theoretical. Some of the most expensive mistakes in market history were made by investors who knew exactly what they were paying per dollar of earnings — and paid it anyway.
## What Overpaying Actually Costs
In 1999, Walmart was arguably the best retailer on Earth, and the market knew it. The shares traded between **$38.70 and $70.30** that year, at an average of **39.1 times earnings**; at the December top, the price worked out to about **55 times** the earnings Walmart reported that fiscal year.
Here's what the business did next. Not the stock — the business:
| Year | EPS | Dividend per share | Average P/E |
|---|---|---|---|
| 1999 | $1.28 | $0.19 | 39.1 |
| 2000 | $1.40 | $0.23 | 38.0 |
| 2001 | $1.50 | $0.27 | 34.9 |
| 2002 | $1.81 | $0.30 | 30.3 |
| 2003 | $2.03 | $0.35 | 26.9 |
| 2004 | $2.41 | $0.48 | 22.8 |
| 2005 | $2.63 | $0.58 | 18.3 |
| 2006 | $2.92 | $0.65 | 16.0 |
| 2007 | $3.16 | $0.83 | 14.5 |
| 2008 | $3.38 | $0.93 | 16.5 |
Earnings up **every single year — +164%** over the stretch. Dividends raised every single year. And a decade after 1999, the typical buyer was **still underwater**, despite collecting **$4.81** per share in dividends along the way.
The chart is the entire mechanism on one screen. A stock price is two numbers multiplied together — earnings, and the multiple paid per dollar of them. Walmart's earnings nearly tripled while its multiple collapsed from 39.1 to 16.5, and the product of the two went sideways for ten years. The 1999 buyer wasn't wrong about the company — they were wrong about the price, and the price ate the decade. Earnings kept compounding regardless, reaching **$4.54** by fiscal 2012, up more than 250% from 1999, which is what eventually bailed the stock out.
Microsoft ran the same experiment with a harder edge: it peaked in December 1999, fell roughly 60% in the dot-com crash, and didn't trade above that 1999 price again until **October 2016** — nearly seventeen years, during which it never stopped being one of the most profitable enterprises on Earth.
This pattern has a famous ancestor. In December 1972, the fifty most beloved growth stocks in America — the "Nifty Fifty," the one-decision stocks you supposedly bought and never sold — traded at an average of **41.9 times earnings** against **18.9** for the S&P 500. Polaroid topped the list at **94.8 times**. When Wharton's Jeremy Siegel reran the numbers a generation later, he found Polaroid's 1972 price had been almost **eight times** what its actual future earnings justified.
But Siegel's study carries a twist: as a *group*, the Nifty Fifty returned **12.2% a year** from December 1972 through August 1998, against **12.7%** for the S&P 500. Bought at the very top of a mania, the basket still nearly kept pace — because consumer brands like McDonald's, Coca-Cola and PepsiCo went on to justify prices even higher than 1972's.
That's what a multiple really is: a bar the business has to clear. Some companies clear 40x. Most don't. The P/E told you exactly how high the bar was set — it couldn't tell you whether the jump would be made. For that, you need the second method.
## The Real Way: Buying Every Dollar It Will Ever Make
Buy a share and you own a slice of every dollar of cash that business will ever produce, from today until it shuts its doors. The honest way to value it is to count those dollars. That's a **DCF — a discounted cash flow** — the machine underneath every serious valuation on Wall Street.
The logic takes three steps: project the cash the business will generate, year by year; discount each year back to what it's worth today; add it up. That sum is the intrinsic value.
The middle step is the one that confuses people: why mark future money down? Because a dollar today can be invested and grow, which makes it strictly better than a dollar tomorrow. If money in your hands compounds at 8% a year, then $100 arriving a year from now is worth about $92.59 today. Stretch the wait and the markdown gets brutal:
At an 8% discount rate, $100 arriving in year ten is worth about **$46** today. In year thirty, about **$10**. Distant promises trade at deep discounts — which is also why companies whose profits live far in the future swing hardest when rates move.
Run a stock price against that sum and the verdict is binary: below intrinsic value, cheap; above it, expensive. This is the real way because it prices the business itself rather than the neighbors — comps can tell you Microsoft is cheaper than Nvidia per dollar of earnings; only a DCF asks whether either is worth owning at today's price at all.
It's also where the trouble starts.
## Where Both Tools Lie to You
**The DCF's problem: the machine is honest, the inputs aren't.** A DCF demands forecasts — how fast cash flows grow, for how long, at what discount rate — and tiny, defensible disagreements produce wildly different answers. Take a business that will hand its owners $1 billion of cash next year, growing 3% forever, discounted at 8%: the perpetuity math says **$20 billion**. Nudge to 7% and 4% — choices nobody could call crazy — and it says **$33 billion**. Go 9% and 2%, and **$14 billion**:
| Scenario | Discount rate | Growth forever | Value of the same business |
|---|---|---|---|
| Sunny | 7% | 4% | **$33.3B** |
| Base | 8% | 3% | **$20.0B** |
| Gray | 9% | 2% | **$14.3B** |
Same company. One percentage point of mood, applied twice, and the "intrinsic value" more than doubles. This is why the most famous practitioners treat the model with suspicion. At Berkshire Hathaway's 1996 annual meeting, Charlie Munger said of his partner:
> "Warren often talks about these discounted cash flows, but I've never seen him do one."
Buffett's reply: "It's true. If the value of a company doesn't just scream out at you, it's too close." The professional defense against your own spreadsheet is the **margin of safety** — acting only when price sits so far below any reasonable estimate that forecasting errors are survivable.
**The multiple's problem: the E is the past, and the past can vanish.** In 2021, Intel earned **$19.9 billion** — and for a stretch of 2022 you could buy the whole company for roughly ten times those earnings, the cheapest-looking sticker in big tech. The multiple said bargain. The market said decay, and the market was right: Intel lost **$18.6 billion** in the first nine months of 2024, and the stock fell **60.1%** that year — the worst in its five-plus decades as a public company — before being dropped from the Dow Jones Industrial Average in favor of Nvidia. A **value trap**: the P/E was low because the E was about to disappear, and the sellers pricing it down to "cheap" knew that first.
Hold both failures in your head and the symmetry appears. A high multiple is a forecast of growth that may not come — Polaroid. A low multiple is a forecast of decline that may well come — Intel. Neither is a discount sticker. Both are predictions, quoted in dollars per dollar of profit, and the only edge is knowing something true about whether the prediction is wrong.
## Does Valuation Even Predict Anything?
Fair question. In 2012, Vanguard's researchers tested the popular forecasting signals — dividend yields, GDP growth, past returns, simple and cyclically adjusted P/Es — against roughly a century of U.S. market history. Almost nothing worked. The one signal with real power was starting valuation: the cyclically adjusted P/E explained about **40% of the variation** in real returns over the *following decade* — and essentially none over the following year.
That's the shape of the tool. Valuation says almost nothing about the next twelve months — cheap gets cheaper, expensive keeps climbing, sometimes for years. Stretch to a decade and the entry price becomes one of the few things that measurably mattered. Valuation is not a timing device. It is the dial that sets what the long run can pay you.
## What This Means
Four things worth keeping:
**Never judge a stock by its sticker.** A 50-for-1 split changes the price 98% and the value 0%. A $5 stock can be the most expensive thing you own; a $3,283 one was a perfectly ordinary claim on a burrito chain. The only prices that mean anything are ratios: dollars paid per dollar of earnings, price against discounted cash.
**Ask what the multiple is assuming.** At 38x you're underwriting years of rapid growth; at 10x you're betting the current earnings survive. The only question that pays is whether the embedded forecast is wrong — and at Walmart in 1999 and Intel in 2022, it wasn't wrong in the buyer's favor.
**Compare to triage, calculate to decide.** Multiples are for scanning the field fast; a DCF — even a rough one on a napkin — is for forcing yourself to write down what actually has to be true for today's price to make sense. If the value doesn't scream at you with conservative inputs, Buffett's rule applies: too close, move on.
**The price you pay sets your decade.** Walmart's shareholders were right about the company for ten straight years and had nothing to show for it. The Vanguard data says that's not an anecdote, it's the rule: starting valuation is the rare signal that reliably shapes long-run returns — and only long-run ones.
The market quotes you a price every second of the trading day. Whether it's a bargain depends on a number it never shows you. The work — quick way or real way — is producing that number yourself before you pay the sticker.
---
*Data: Barebone | Sources: Chipotle investor relations (June 2024), The Rational Walk Wal-Mart ten-year case study (2010), Walmart fiscal 2012 results, Jeremy Siegel, "Valuing Growth Stocks: Revisiting the Nifty Fifty" (AAII Journal, 1998), Vanguard, "Forecasting Stock Returns" (Davis, Aliaga-Díaz & Thomas, 2012), Berkshire Hathaway 1996 annual meeting and 2008 shareholder letter, Intel SEC filings (2021–2024) | Data as of April 16, 2026*