The Owner Earnings Formula: Buffett's Cash-Earnings Method

In the appendix of his 1986 Berkshire shareholder letter, Warren Buffett published a nine-line correction to GAAP net income. He called the result owner earnings. It has been quoted in every value-investing book published since. The owner earnings formula is the closest thing in fundamental analysis to a canonical "real profit" calculation. It is the number a long-term owner of the business could actually extract.

It is also one of the most-quoted, least-computed formulas in investing. Most analysts who reference it have never sat down and run the math. The reason is the hardest input — maintenance capital expenditure — which is rarely disclosed and always argued.

This essay is the long version of the owner earnings formula. Where it came from. How to compute each term. And a worked example using Microsoft's fiscal-year 2025 10-K — a filing where the formula's central ambiguity, growth versus maintenance capex, decides the entire answer.

The result, applied to Microsoft, is uncomfortable. Run two reasonable versions of the owner earnings formula on the same inputs, and the calculated intrinsic value differs by 40 percent. The formula is not broken. It is doing exactly what it was designed to do — making you state your assumption out loud.

What the owner earnings formula actually says

Buffett's 1986 definition, in his words:¹

"These represent (a) reported earnings plus (b) depreciation, depletion, amortization, and certain other non-cash charges ... less (c) the average annual amount of capitalized expenditures for plant and equipment, etc. that the business requires to fully maintain its long-term competitive position and its unit volume. (If the business requires additional working capital to maintain its competitive position and unit volume, the increment also should be included in (c)...)"

Reduced to a single equation, the owner earnings formula is:

Owner Earnings = Net Income
               + Depreciation, Depletion, Amortization (D&A)
               + Other non-cash charges
               − Maintenance capital expenditure
               − Working-capital additions required for maintenance

The point of the correction is component (c). GAAP cash flow does not distinguish between capex that keeps the existing business running and capex that builds new business. Consider a company that spends $10B to maintain its current factories and another $40B building a new product line. The cash flow statement reports the entire $50B as a single line. Operating cash flow minus that $50B understates what a long-term owner could actually withdraw — $40B of it was a discretionary investment.

The owner earnings formula tries to back out only the maintenance portion. That sounds clean. In practice, every analyst who uses it has to estimate one number that the company never tells them.

The owner earnings formula step by step

Five components, in calculation order.

1. Net income (start here). The bottom line of the GAAP income statement. This is the number GAAP delivers, before any of Buffett's corrections. It includes non-cash charges (good — we will add them back) and excludes growth capex (good — we will not adjust for that).

2. Add D&A. Depreciation, depletion, and amortization are non-cash expenses. They reduce reported net income but do not consume cash. Add them back to get closer to cash generation. The number lives in the cash flow statement's "adjustments to reconcile" block.

3. Add other non-cash charges. The 1986 letter says "certain other non-cash charges". In modern practice the two material items are stock-based compensation and deferred income tax expense. Buffett's original framing predated SFAS 123 and the universal use of equity awards. Most practitioners today add SBC back here, then immediately subtract it again under (c). SBC is real economic cost even though it does not consume cash. We will treat SBC explicitly in the worked example.

4. Subtract maintenance capex. This is the variable that does the work. Maintenance capex is the dollar amount the company would have to spend to keep its existing operations producing at current scale and competitiveness. It is almost never disclosed. Three workable approximations:

• D&A as a proxy. For a steady-state business not growing its physical footprint, D&A roughly equals maintenance capex. Quick, defensible for mature businesses.
• Percentage of total capex. Estimate maintenance as 60-85% of total capex for mature businesses; lower for high-growth.
• Disclosed split. A small number of filers (REITs, some utilities, occasionally industrials) break growth and maintenance capex out in MD&A. When they do, use it.

5. Subtract working-capital additions for maintenance. Required only if the business has to grow its working capital to maintain current operations — typical for distributors and some retailers. For most businesses this term is zero in the long run. Skip it on a first pass; add it back if the inventory line is growing faster than revenue.

Worked with placeholder numbers:

Line	Value
Net income	$100B
+ D&A	+$20B
+ Other non-cash (SBC, deferred tax)	+$10B
− Maintenance capex	−$25B
− ΔWC for maintenance	$0
= Owner earnings	$105B

This is the answer the owner earnings formula was designed to produce. The trouble starts when you try to fill in line 4 for a real company.

How to estimate maintenance capex

The hardest variable in the owner earnings formula is the one the company will never give you. Three methods, in order of how much they assume:

Method 1 — D&A as a proxy. If the business is steady state, the physical and intangible assets being depreciated each year roughly equal the capex required to replace them. So maintenance capex ≈ D&A. This is the default for mature, slow-growth businesses (consumer staples, mature industrials, utilities). It fails when the asset base is growing rapidly or shrinking rapidly.

Method 2 — Heuristic percentage of total capex. Estimate maintenance as 60-85% of total reported capex for mature businesses; 30-50% for high-growth businesses currently building out new capacity. The wider that range, the more honest you are being about your uncertainty. Defensible when the company is plainly between steady state and active expansion.

Method 3 — Disclosed split. Some 10-Ks break growth capex and maintenance capex out explicitly in MD&A. REITs often do this (recurring vs. discretionary capex). When the disclosure exists, use it. It is the only one of the three methods that is not a guess.

The thing to remember: maintenance capex is the variable where most owner-earnings calculations go wrong. Two analysts using the same 10-K and the same formula can land on owner-earnings figures that differ by 30 to 50 percent purely because of this assumption. Always state which method you used. Never hide the assumption behind a single calculated number.

A worked owner earnings formula on Microsoft FY2025

Microsoft's fiscal-year 2025 ended June 30, 2025. The 10-K was filed with the SEC on July 30, 2025.² Here are the inputs:

Line item (MSFT FY2025)	Value
Net revenue	$281.72B
Net income	$101.83B
Depreciation (D&A reported as depreciation)	$22.00B
Stock-based compensation	$11.97B
Cash from operating activities	$136.16B
Capital expenditures	$64.55B
Dividends paid	$24.08B
Common stock repurchased	$18.42B
Common shares outstanding (Apr 2026)	7.428B

This is a classic case where the owner earnings formula's central ambiguity matters. Microsoft's $64.55B of FY2025 capex is dominated by AI data-center build-out — not maintenance. Treating it as steady-state maintenance crushes the result. Treating it as zero ignores reality. Both numbers are defensible. Neither is right.

Method A — Conservative free-cash-flow proxy. OCF minus all capex, no maintenance adjustment:

OCF                         $136.16B
− Capex (all)               −$64.55B
= Free cash flow            ≈$71.61B

Per share: $71.61B ÷ 7.428B = $9.64. Apply Gordon growth with r = 10%, g = 4%:

IV per share = 9.64 × 1.04 / 0.06 ≈ $167.09

Method B — D&A as the maintenance-capex proxy. Buffett's owner earnings using D&A in line 4:

Net income                  $101.83B
+ D&A                        +$22.00B
+ SBC (non-cash)            +$11.97B
− Maintenance capex (≈ D&A)  −$22.00B
− SBC (real economic cost)  −$11.97B
= Owner earnings            ≈$101.83B

Per share: $101.83B ÷ 7.428B = $13.71. With the same r = 10%, g = 4%:

IV per share = 13.71 × 1.04 / 0.06 ≈ $237.66

The margin of safety against Microsoft's $450.24 share price:

Method	IV/share	Price	MoS
A — Conservative FCF	$167.09	$450.24	−169%
B — Buffett's owner earnings, D&A as maint capex	$237.66	$450.24	−89%

The two methods, both legitimately the "owner earnings formula", produce intrinsic values that differ by 42 percent. Neither is correct in isolation. The honest answer is that Microsoft today is being valued for AI growth optionality that neither version of the formula can capture. The difference between Method A and Method B is the analyst's implicit guess. The guess is how much of Microsoft's $64.55B capex is building future profit.

If you want a framework for what the gap between intrinsic value and market price actually means, see our walk-through of the margin of safety formula.

## Where the owner earnings formula breaks

The formula is sturdy for steady-state businesses with predictable capex. It struggles, sometimes badly, in five common situations.

Heavy growth capex disguised as maintenance. Microsoft, Meta, and Alphabet are all in this category right now. Their reported capex line includes tens of billions of dollars of discretionary expansion. Owner earnings using the D&A proxy materially overstates available cash; owner earnings using total capex understates it.

Heavy R&D businesses. Pharma, semiconductors, and parts of biotech expense R&D entirely in the period incurred. Economically, R&D is closer to capex — it produces future product. The owner earnings formula does not adjust for this. A literal application understates the cash a long-term owner of a research-heavy business could withdraw if R&D were paused.

Lumpy capex cycles. Cyclicals and capital-heavy industrials run multi-year capex cycles. A single year's owner-earnings calculation can land at the peak or trough of the cycle. Use a 5-to-7 year average of capex and D&A, not a single year.

Capital-light businesses. Asset managers, software-only businesses, and franchised consumer brands have D&A and capex that are tiny relative to net income. The owner-earnings formula technically applies but adds little — net income already approximates cash earnings. For these, focus on share count creep and SBC instead.

Insurance and banking. Different framework entirely. Net income, OCF, and capex do not map cleanly to economic earnings for financial institutions. Buffett's own analytical method for these businesses (book value per share, plus an explicit valuation of the float for insurers) is a separate framework.

How to use the owner earnings formula without being misled

Three habits that separate a defensible owner-earnings calculation from a number you should not trust.

Show the maintenance-capex assumption explicitly. Never present an owner-earnings figure without disclosing which method you used (D&A proxy, percentage of capex, disclosed split). The number alone is meaningless without the assumption behind it.

Run it two ways and show the spread. Compute owner earnings using Method 1 (D&A) and Method 2 (a conservative percentage of total capex). The spread between the two is a measure of your uncertainty. If one version says "buy" and the other says "pass", you do not have a thesis. You have a guess.

Cross-check with the 5-year average. Pull D&A, capex, and OCF for the last five fiscal years and average them. Lumpy cycles get smoothed; one-time write-downs get absorbed. If your one-year owner-earnings number is materially different from the 5-year average, find out why before relying on either. For the technique of pulling that data efficiently from a 10-K, see our working investor's guide to the 10-K.

What the owner earnings formula is actually for

The owner earnings formula was Buffett's 1986 correction to a GAAP earnings figure he thought systematically misled long-term owners. It still does that job. It does not, and was never designed to, produce a single defensible number for any business. It produces a method, and the method makes you state your maintenance-capex assumption out loud so a critic can argue with it.

That is the formula's contribution. It moves the argument from "what is Microsoft worth?" to "how much of Microsoft's $64B capex is building future profit?". The second question is harder. It is also the only one whose answer matters.

For more long-form essays on valuation methodology, see the [rest of the Hub](/hub).