Costco: A Wonderful Business at a Price That Already Spent the Decade
Some companies fail analysis because they are hard to understand. Costco fails for the more painful reason: it is easy to understand, easy to admire, and simply not for sale at a sensible price. The gates below are passed in order, three out of four, and the fourth is not close.
Gate 1: Circle of Competence
Costco is a membership warehouse retailer, and the structure of the business is a deliberate inversion of normal retail. Merchandise is sold at a markup intentionally capped at roughly 11–14%, leaving a net margin of 2.9%. The real profit engine is the annual membership fee. In FY2025, paid members numbered 68.3 million, generating roughly $5.3 billion in fee income, up 10% year over year — money with almost no variable cost attached, which falls essentially in full to operating income. Sell at cost, charge for admission. The merchandise is the retention device; the fee is the business. Charlie Munger sat on the board for 25 years and called it his favorite company in public.
Is it knowable? On both levels, yes. The information is unusually complete: a 10-K for fiscal 2025 (ended 2025-08-31), quarterly releases, monthly sales disclosures — a rarity among retailers — and full visibility into renewal rates and member counts. And the predictability holds: consumers will still want groceries at the lowest possible price in ten years. A renewal rate in the 92% range (US and Canada) sustained for close to 30 years is not a probability bet; it is a subject for analysis. Fifteen years of Amazon threat narratives have not stopped membership from growing. Describing Costco in ten years — more warehouses (25–35 new openings per year), more members, periodic fee increases — is analysis, not speculation.
Gate 1: passed.
Gate 2: Moat
The moat’s foundation is cost advantage — the kind explicitly enshrined in our constitution as a textbook example — reinforced by intangibles: the Kirkland brand and the consumer trust that “Costco means lowest price,” plus the gentle switching cost that membership creates. The engine is the flywheel: scale buys purchasing power, purchasing power buys lowest prices, lowest prices buy members, members buy scale.
The structure itself is the hardest part to copy. The self-imposed markup ceiling means a competitor matching it must destroy their own margins to do so.
The pricing power evidence is clean. In September 2024 the annual fee rose from $60 to $65 (executive tier, $120 to $130), and afterward the US/Canada renewal rate stood at 92.2% (most recent FY2026 quarter), 89.7% worldwide. Raising price without a prayer service — a textbook pass, of the same species as See’s Candies’ Valentine’s increases. Note the elegant twist: this pricing power resides in the fee, and is deliberately not exercised on merchandise. The restraint is the moat.
The direction is widening. Paid members grew from 58.8 million (FY2023) to 63.7 million (FY2024) to 68.3 million (FY2025); 27 new warehouses opened in FY2025 with up to 35 planned for FY2026. Management has proactively flagged that renewal rates may drift slightly lower as online sign-ups grow, but rates remain in the 92% range — no structural damage signal.
On the AI-era question: neutral to strengthening. Physical cost advantage and member trust are not the information-asymmetry moats AI erodes; AI and automation may deepen the cost edge through logistics and inventory optimization. AI-driven instant delivery is a convenience offensive, but it is the Amazon threat’s fifteenth variation, and it does not collide with Costco’s value proposition of lowest unit cost.
Gate 2: passed.
Gate 3: Management
The capital allocation record is exemplary. Dividends have risen for 22 consecutive years (to $1.47 per share as of April 2026). When retained cash exceeds reinvestment opportunities, it goes back as special dividends: 2012, 2015, 2017 ($7), 2020 ($10), 2023 ($15), January 2026 ($12, roughly $5.3 billion in total). No empire-building acquisitions; surplus money returned. Share buybacks, at roughly $903 million in the most recent fiscal year, are conservative — declining to buy heavily at today’s price is itself evidence of discipline. Growth capex ($5.5 billion in FY2025, $6.0–6.5 billion planned for FY2026) flows into new warehouses, a proven high-return reinvestment.
The retained-earnings test passes: market capitalization has multiplied over five years against retained profits — though a substantial portion is multiple expansion, a deduction reserved for Gate 4. On candor: pre-announcing possible renewal rate softness before it appears is the opposite of quarterly gamesmanship. Industry-leading wages produce low turnover, which reinvests into the long-term cost advantage. No accounting massage on record.
Gate 3: passed.
Gate 4: Price
Owner earnings: the screening figure is $5.03 billion, but that treats all capex as maintenance. FY2025 net income was $8.10 billion, operating cash flow $13.3 billion, capex $5.5 billion — simple FCF around $7.8 billion. Since much of that capex builds new warehouses, a conservative owner earnings estimate sits at $8–9 billion, near net income. Taking $8.5 billion generously against a $406 billion market cap yields an owner earnings yield of roughly 2.1% — less than half the ~4.3% on the 10-year Treasury. The absence of a margin of safety is not debatable; it is arithmetic.
Valuation range: at 15x, $8.5 billion gives about $128 billion; stretched to an indulgent 20x for the subscription-quality fee business, about $170 billion — roughly $290–385 per share. The market price of $406 billion (PER 46.3, ~$915 per share) is more than 2.4 times even the top of that range. A 30% discount target lands near $90–120 billion ($200–270 per share), a price that realistically arrives only with a market-wide panic.
The See’s lesson applies with precision: See’s in 1972 was “three times book,” not “48 times owner earnings.” Buying Costco today is buying a wonderful company at an un-wonderful price, and paying most of the next decade’s return up front. With the multiple at the top of its historical forward range (40–45x), normalization alone could offset a year’s earnings growth. The gate fails on price, not quality.
Gate 4: failed.
Verdict
Watch. Gates 1 through 3 pass as cleanly as any business we examine — five-year average ROE of 28.2% with a standard deviation of just 2.2%, a constitutional-example moat, and 22 years of dividend increases plus six special dividends. The failure point is price alone, and the correct action is observation, not purchase. The file reopens if the PER enters the 25–28 range (market cap around $220 billion).
What would change the verdict: a structural collapse in the subscription premise — US/Canada renewal rates below 90% triggering a moat review, below 88% discarding the thesis — or member growth stalling while fee income growth drops under 10% without a fee increase cycle. Conversely, if fee headroom proves far larger than estimated, today’s price could yet be vindicated; and the merchandise gross margin drifting above its historical band would signal self-harm to the moat itself.
This analysis is AI-generated, educational, and not investment advice. Figures may contain errors or be delayed. Disclaimer