The Graham Number Explained: Benjamin Graham's Formula for Finding Undervalued Stocks

Benjamin Graham - the value investor who taught Warren Buffett - built a formula in the 1940s to answer one narrow question: what's the most a genuinely conservative investor should pay for this stock? We ran it on 11 real large-cap names using live September 2026 data. Three trade below their number. Apple trades at roughly eight times its.

Benjamin Graham published "Security Analysis" in 1934 and "The Intelligent Investor" in 1949, and both books are still in print because the underlying discipline - separate what you're paying from what you're actually getting - never really goes out of date. Buried inside that discipline is a small, specific formula that gets less attention than his broader ideas but is easy to check for yourself: the Graham Number.

It isn't a prediction, a rating, or an AI-generated score. It's a ceiling - the highest price Graham believed a conservative investor should be willing to pay for a stock, based on nothing more than trailing earnings and the company's book value.

The Formula

Graham Number = √(22.5 × EPS × Book Value Per Share)

The 22.5 isn't arbitrary. Graham set two hard limits for what he called a defensive investment: a price-to-earnings ratio no higher than 15, and a price-to-book ratio no higher than 1.5. Multiply those together - 15 × 1.5 - and you get 22.5. The formula is really just those two conservative valuation ceilings, combined into a single number expressed in dollars per share instead of two separate ratios.

A worked example: a company earning $5.00 per share with $40.00 of book value per share has a Graham Number of √(22.5 × 5 × 40) = √4,500 ≈ $67.08. If the stock trades below $67.08, it passes Graham's test. Above it, it doesn't - full stop, regardless of the story.

What 11 Real Stocks Score Right Now

Using live fundamentals and September 2026 prices, here's where a mix of large, familiar names actually land against their own Graham Number:

Source: FMP fundamentals and live quotes, checked September 2026. Trailing-twelve-month EPS and book value per share.

Two names that would normally sit in this kind of table - Intel and Ford - are missing entirely, and that's itself informative: both had a net loss over the trailing twelve months when this was checked, and the Graham Number formula can't produce a result from a negative EPS. No number, not a zero, not a placeholder - the formula simply doesn't apply.

Only three of the eleven - AT&T, Citigroup, and Wells Fargo - currently trade below their own Graham Number. That's not a coincidence of which names were picked; it's a reasonably accurate snapshot of what this specific, conservative screen finds in a market where large technology and consumer-brand companies trade at big multiples of their tangible book value. Apple, at roughly 7.3 times its Graham Number, isn't overvalued by that fact alone - it just means the formula's inputs (trailing earnings and book value) capture a shrinking fraction of what the market is actually paying for.

Why So Few Stocks Pass This Test in 2026

Graham built this formula in an era when a much bigger share of a typical public company's value sat directly on the balance sheet - factories, equipment, inventory, real estate. Book value per share was a reasonable proxy for what the business was actually worth if you shut it down and sold the pieces.

That's a much weaker assumption today. A company like Apple or Microsoft derives most of its value from brand, software, recurring subscription revenue, and network effects - assets that either don't appear on the balance sheet at all or are dramatically understated relative to the cash flow they produce. Book value per share for those companies is tiny relative to their earnings power, which mechanically drags the Graham Number down and makes the stock look expensive by this one measure even when the underlying business is compounding just fine.

This is exactly the same lesson that shows up in how EBITDA and other single-metric shortcuts break down for asset-light businesses - one formula built for one kind of company will misjudge a company built differently. The Graham Number isn't wrong when it flags Apple as "expensive." It's just answering a narrower question than most investors assume: not "is this a good business," but "does this pass a specific, tangible-asset-weighted valuation ceiling written for a 1940s industrial economy."

Where the formula is genuinely most useful is on the kind of company it was designed for: mature, earnings-positive, balance-sheet-heavy businesses - banks, telecoms, industrials, insurers. That's exactly the group where three of the eleven names above (AT&T, Citigroup, Wells Fargo) actually cleared the bar.

What the Graham Number Doesn't Tell You

A stock trading below its Graham Number is not automatically a buy, and a stock trading above it is not automatically overvalued. The formula is silent on:

• Why the market is discounting it. AT&T's telecom business carries real secular pressure and a heavy debt load; Citigroup has spent years trading below book value while working through a multi-year restructuring. A cheap Graham Number can mean "overlooked" or it can mean "priced correctly for real risk" - the formula alone can't tell you which.
• Growth. The formula uses trailing EPS, so a company with earnings about to accelerate or decline sharply gets the same treatment as one with flat, stable earnings.
• Earnings quality. Graham himself spent whole chapters on financial strength and the durability of a company's earnings before trusting any valuation shortcut. A single EPS figure says nothing about accounting quality - that's exactly the gap tools like the Beneish M-Score and Altman Z-Score are built to check, and pairing the Graham Number with one of those is closer to how Graham actually worked.

Treat a low Graham Number the way you'd treat any single screening filter: a reason to look closer, not a reason to buy. The companion piece here is how to avoid value traps - cheap-looking stocks that stay cheap because the business is genuinely deteriorating, not because the market is missing something.

Using It as a Starting Filter, Not a Final Answer

The practical way to use the Graham Number today is as the first step in a funnel, not a standalone buy signal:

1. Screen for it. Filter for positive trailing EPS and a price-to-book ratio and P/E combination inside Graham's original ceilings, or approximate it directly with Stock Alarm's screener using P/E and price-to-book columns together.
2. Check earnings quality next. Run candidates through a quality check like the Piotroski F-Score or Altman Z-Score before assuming "cheap" means "safe."
3. Set an alert instead of watching the price. Once you've identified a name worth tracking, a price alert at your own calculated fair-value ceiling does the watching for you - useful specifically because Graham-Number candidates tend to be slower-moving, unglamorous stocks that don't reward checking the quote every hour.

The formula's real value isn't precision - book value and trailing earnings are both blunt instruments. It's discipline: a fixed, unemotional ceiling that doesn't move just because a stock has been going up.

Related Reading

• Altman Z-Score: How to Spot Bankruptcy Risk Before It Happens
• Piotroski F-Score: The 9-Point Checklist for Quality Value Stocks
• Beneish M-Score: How to Detect Earnings Manipulation
• How to Avoid Value Traps
• Book Value Per Share Explained
• Free Cash Flow Explained: The Metric Warren Buffett Says Matters Most
• How to Set Stock Price Alerts: Complete Setup Guide for Traders