Costco Wholesale COST

Reference table

1. What the business actually is

The 10-K describes Costco as "principally engaged in the operation of membership warehouses" in fifteen countries, operating on the concept that "offering low prices on a limited selection of nationally-branded and private-label products in a wide range of categories will produce high sales volumes and rapid inventory turnover."

The physical footprint, from the FY2025 10-K:

A structural fact that matters and is easy to miss: of the 914 warehouses, only 189 involve a lease, and "141 of the 189 leases are land-only leases, where Costco owns the building." Costco owns the building at 866 of its 914 locations. This is why its lease footprint is trivially small — the 10-K discloses only "$1,094 of lease payments for leases that have been signed but not commenced" — and it is a genuine and under-appreciated difference from a heavy-lessee retailer.

Revenue by merchandise category (seg_prod, $M) — and this table reconciles to the filing to the dollar:

The reconciliation is exact and worth stating, because it is unusual. Core merchandise (food and sundries + fresh food + non-foods) grew from $199,656M to $218,742M — an increase of $19,086M, and the 10-K states "Sales increased $19,086 or 10% in core merchandise categories." The "Other" line grew $1,201M and the 10-K states "Sales in warehouse ancillary and other businesses increased $1,201, or 2%." Both to the dollar. Unlike several names in this batch, the vendor's segment file for Costco is clean and requires no correction.

The 2.4% growth in "Other" is the gasoline line, and it explains a great deal. Warehouse ancillary includes gasoline, pharmacy, optical, food court, hearing aids and tyre installation. In FY2025 "lower gasoline prices negatively impacted net sales by $2,329, or 93 basis points, with an 8% decrease in the average price per gallon" even though "the volume of gasoline sold increased approximately 2%." In the third quarter of fiscal 2026 the sign reversed and the same line grew 29% — which is the single most important thing to understand about the most recent quarter (Section 3).

Revenue by geography (seg_geo, $M):

The 10-K's critical-audit-matter discussion states that "the Company generated $200,046 million and $36,923 million of total revenue in the United States (U.S.) and Canada, respectively, for the year ended August 31, 2025" — again an exact match to the vendor file.

Channel. "Net sales for e-commerce represented approximately 7% of total net sales in 2025", and "digitally-enabled sales ... represented approximately 10% of total net sales." Digitally-enabled comparable sales rose 21% in the third quarter of fiscal 2026 and 23% in the second — by a distance the fastest-growing thing in the company, on a base of roughly a tenth of sales.

Competition, in the filing's own words: "Walmart, Target, Kroger, and Amazon are among our significant general merchandise retail competitors in the U.S. We also compete with other warehouse clubs, including Walmart's Sam's Club and BJ's Wholesale Club."

2. Membership — the engine, the stale statistic, and the tailwind that is running out

This is the section that decides the name, and none of it is visible in the vendor payload.

2a. How the fee works

"Our annual fee for these memberships is $65 in the U.S." Executive members pay "an additional annual fee of $65" and earn "a 2% reward on qualified purchases (generally up to a maximum reward of $1,250 per year)." The 10-K reports that Executive members numbered 38.7 million at FY2025-end (against 35.4 million and 32.3 million in FY2024 and FY2023) and that "the sales penetration of Executive members represented approximately 73.6% of worldwide net sales in 2025."

Renewal rates, from the filings, quarter by quarter:

Both filings give the same explanation for the drift: "Renewal rates were negatively impacted by a higher number of memberships sold online, including through digital promotions, entering the renewal rate calculation. These memberships renew at a slightly lower rate on average." That is a mix effect, not a deterioration, and we read it as such — but it is a mix effect that will not reverse, because the online channel is the growth channel.

Deferred membership fees — the forward book, and the vendor reports it as zero. The 10-K: "Deferred membership fees at the end of 2025 and 2024 were $2,854 and $2,501." That is +14.1%, and it is contracted fee income that will be recognised over the following twelve months. The vendor's balance sheet reports deferredRevenue = 0.00 in all three years. The filing wins; we use $2,854M (Section 8).

2b. The statistic the whole market repeats, and why it is now wrong

The most-cited fact about Costco — and the highest-conviction claim in our knowledge base, at conviction 85 from 2023 — is that "~75% of Costco operating income comes from membership fees at ~90% incremental margin." We have checked it against six years of filings and it is materially stale.

Membership fees are 51.3% of operating income, not 75%. The 75% figure is roughly right if you measure fees against net income in fiscal 2022 (72.3%) — which is probably what the original claim meant — but on either measure the ratio has been falling for five straight years, by fourteen points of operating income and by seven points of net income since fiscal 2022.

This cuts both ways and we refuse to pick only the flattering half. In Costco's favour: the merchandise business is genuinely more profitable than the folklore allows, which means the company is less dependent on a single annuity than the standard bear case assumes. Against Costco: the "it's really a subscription business" argument is the main justification offered for paying a 47.7x multiple, and it is 24 points weaker than when it was formulated. A business where roughly half the operating profit is thin-margin retail should not be valued as though nearly all of it is a subscription.

2c. The tailwind that is running out — the sharpest number in this dive

The 10-K: "As previously reported, we increased our annual membership fees in the U.S. and Canada, effective September 1, 2024. We account for membership fee revenue on a deferred basis, recognized ratably over the one-year membership period. The fee income increase accounted for approximately 40% of membership income growth during 2025."

The same disclosure appears, updated, in each subsequent quarterly report:

Read that column downward: 40%, then 35%, then 25%, then nothing. The fee increase became effective 2026-09-01 for the last cohort of renewing members and is recognised ratably over their year; by the fourth quarter of fiscal 2026 there is no incremental fee benefit left in the comparison. Membership fee growth from that point is new sign-ups plus Executive upgrades alone.

Note also that membership fee growth has already stepped down from +14% (Q2) to +11% (Q3) as the contribution fell from 35% to 25%. The mechanical arithmetic implies membership fee growth in the high single digits once the increase is fully out of the base. Membership fee revenue is 51.3% of operating income; a step down of three to five points in its growth rate is roughly one and a half to two and a half points off operating income growth, with no offset unless merchandise margin improves.

The counter-case, stated fairly, is that the company will raise the fee again. The 10-K explicitly lists "membership fee changes" among the matters its forward-looking statements address, and the last increase was effective 2024-09-01. A new increase is the single cleanest upside catalyst available and no consensus estimate in the file contains one. It is also entirely at management's discretion and therefore, under our house rule, an input rather than evidence. It is not in our base case; it is in our bull case, explicitly.

3. Growth and margins — what a 12% quarter actually was

3a. The annual record

Two things to hold about this table. First, fiscal 2023 was a 53-week year, per the 10-K — so the +6.8% is flattered by an extra week and the following year's +5.0% is penalised by the same week. The underlying two-year run rate is roughly 6%. Second, operating margin has expanded 51 basis points in two years, from 3.35% to 3.77% — on a base of 3%, that is a 12.5% relative improvement in profitability and it is why earnings have grown faster than sales.

A definitional warning about that gross-margin column. The vendor's 12.8% is gross profit over total revenue, and total revenue includes membership fees, which carry essentially no cost of goods. The company's own gross margin measure is expressed as a percentage of net sales excluding membership fees, and on that basis the 10-K reports 11.03% for FY2025 excluding gasoline deflation — an increase of 11 basis points. The two numbers are not comparable and the vendor never says so. Merchandise gross margin is 11%, not 12.8%; the difference is the fee.

3b. The quarterly record, and the quarter that missed

Note: the fourth quarter is a 16-week period and is not comparable to the 12-week first three quarters. Fiscal 2026 is a 52-week year ending 2026-08-30.

On its face, the third quarter of fiscal 2026 was the best in the series. Revenue +11.6%, EPS +15.2%, operating margin back to 4.0%. It still missed. earn_cal records actual EPS of $4.93 against an estimate of $5.00 — the only miss in the recent series — on revenue of $70.527B against an estimate of $69.829B. A revenue beat and an earnings miss. The 10-Q filed 2026-06-03 explains exactly why.

Decomposing the 12% net sales growth, from the filing:

Three points follow, and together they are the analytical core of this dive.

First, 3.25 of the 12 points were petrol prices and currency — neither of which the company controls, both of which reverse. Excluding them, growth was roughly 8.8%, and core merchandise sales grew 7%.

Second, the comparable-sales mix inverted. In fiscal 2025 comparable sales of +6% came from "increases of 5% in shopping frequency and approximately 1% in average ticket" — the highest-quality composition available, because it means more members shopping more often. In the third quarter of fiscal 2026 it was 7% average ticket and 2% frequency. With petrol prices up 20%, a jump in average ticket is largely a fuel-price artefact. The number of shopping trips is growing at 2%, down from 5%.

Third, and this is what produced the miss: gasoline is close to gross-profit-free and it diluted the margin. The 10-K states the mechanic plainly: "A higher penetration of gasoline sales will generally lower our gross margin percentage." In the quarter, reported gross margin as a percentage of net sales fell 21 basis points; excluding the gasoline effect it was 11.26%, up one basis point. And underneath even that, "the gross margin in core merchandise categories, when expressed as a percentage of core merchandise sales ... decreased nine basis points. The decrease was primarily due to fresh foods and foods and sundries."

So the quarter beat on revenue because petrol got expensive and missed on earnings because petrol makes no money and the merchandise margin went slightly the wrong way. That is a completely benign explanation of the miss in business terms — and a completely unhelpful one for a stock trading at 47.7x, because it means the headline growth rate is not the growth rate you are paying for.

Two offsets we do record. SG&A excluding the gasoline effect fell two basis points to 9.14%, helped by "the absence of a charge related to a one-time expense for increased employee vacation." And the effective tax rate was 25.4% against 26.2% a year earlier, which is worth roughly $0.07 of EPS on its own — i.e. the tax rate contributed about as much as the size of the miss.

3c. The consensus path

Two notes on this table. (a) FY2026E is not really an estimate any more. Three quarters are reported ($4.50 + $4.58 + $4.93 = $14.01) and only the 16-week fourth quarter remains, with consensus at $6.57 — a sum of $20.58 against the $20.540 in the estimates file. The 46.1x "forward" multiple on FY2026E is functionally a trailing multiple. The first genuinely forward year is FY2027E and every forward multiple in this dive is computed on it. (b) The FY2029 and FY2030 rows rest on nine analysts each and nothing in this dive depends on them — but the shape they describe, deceleration to 4.8%, is the direction of travel consensus itself models.

4. Balance sheet and capital returns — three corrections, all in the company's favour

Correction one: netDebt discards short-term investments. The vendor reports netDebt of −$5.988B (i.e. net cash of $5.988B), which is total debt ($8.173B) less cash and cash equivalents only ($14.161B). It silently discards $1.123B of short-term investments. Counted properly:

> Cash and short-term investments $15.284B − vendor total debt $8.173B = net cash of $7.111B, not $5.988B.

The vendor understates the net cash position by 18.8%. This is the same failure mode found across this feed, at smaller scale.

Correction two — FILING OVERRIDES VENDOR: the vendor omits the current portion of long-term debt. The 10-K states that "As of the end of 2025, long-term debt with fixed interest rates was $5,805", and separately that the current portion is "included in other current liabilities." The vendor's longTermDebt is $5.713B. The $92M difference is the current maturity. Rebuilding total debt from the filing:

Correction three — and this is the largest: the vendor's balance sheet is eleven months out of date, and the 10-Q shows a materially stronger position. The vendor's most recent balance sheet is fiscal 2025, dated 2025-08-31. The 10-Q filed 2026-06-03 states:

> "Cash and cash equivalents and short-term investments were $19,996 and $15,284 at May 10, 2026, and August 31, 2025."

Cash and short-term investments have risen $4.712B, or 30.8%, in thirty-six weeks. Against long-term debt of approximately $5.736B (the FY2025 $5.805B less the $69M the Japanese subsidiary repaid in March 2026, per the 10-Q) plus $96M of short-term borrowings and roughly $2.46B of finance leases, current net cash is approximately $11.7 billion — 65% above the corrected FY2025 figure and 95% above the vendor's headline.

Debt composition, from the filings. "The Company's long-term debt consists primarily of Senior Notes", with "Other long-term debt ... Guaranteed Senior Notes issued by the Company's Japanese subsidiary." The Japan notes were issued in fiscal 2024, "totaling approximately $500, at fixed interest rates ranging from 1.400% to 2.120% ... maturity dates range from November 2033, to November 2043." Costco repaid "the $1,000 outstanding principal balance on its 2.750% Senior Notes" in fiscal 2024 and has repaid $103M, $77M and $75M of the Japanese notes in fiscal 2025, 2024 and 2023, plus $69M in the first thirty-six weeks of fiscal 2026. The direction is steady, unforced deleveraging. The fair value of long-term debt including the current portion was $5,259M at 2026-05-10 against $5,370M at FY2025-end.

The lease position is trivially small and it is the mirror image of a heavy lessee. Because Costco owns the building at 866 of its 914 warehouses, the only off-balance-sheet lease disclosure in the entire 10-K is "$1,094 of lease payments for leases that have been signed but not commenced" — about 0.26% of market capitalisation. For comparison, several names in this batch carry uncommenced lease commitments in the tens or hundreds of billions. We checked for the failure mode and it is genuinely absent here.

Enterprise value. km_ttm.enterpriseValueTTM reports $409.638B. Rebuilt on the FY2025 balance sheet (market cap $420.352B + total debt $8.173B − cash and short-term investments $15.284B) it is $413.241B, a −0.9% residual — immaterial, and we accept the vendor figure for scale. Rebuilt on the current 10-Q balance sheet it would be roughly $408.7B. evToEBITDATTM of 28.3x and netDebtToEBITDATTM of −0.74x are arithmetically fine.

Other balance-sheet facts, FY2025: total assets $77.099B; property, plant and equipment net $34.634B (from $31.649B); inventory $18.116B against accounts payable $19.783B — Costco's payables exceed its inventory, which is the balance-sheet expression of the near-zero cash conversion cycle; goodwill only $994M and intangibles zero; total equity $29.164B; retained earnings $22.650B. Book value per share $75.48, so the stock trades at 12.56x book. Current ratio 1.072, quick ratio 0.611. Cash conversion cycle 0.46 days (27.7 days of inventory + 4.7 days of receivables − 31.9 days of payables). Interest coverage 76.9x.

The vendor's grahamNumberTTM of $183.88 and grahamNetNetTTM of −$45.96 are arithmetically correct and analytically meaningless for an asset-light-inventory, high-return retailer. We note them only to say we are not using them.

Contingencies are immaterial. The 10-K's Note 10 states the company "has recorded an immaterial accrual with respect to some matters described below, in addition to other immaterial accruals for matters not described below" and that for unaccrued matters the possible loss "cannot, in the Company's view, be reasonably estimated." No valuation adjustment is made. Self-insurance liabilities exist but the extracted text does not carry the figure.

Risk factors: no change. Both 10-Qs state "There have been no material changes in our risk factors from those disclosed in our Annual Report on Form 10-K." The one risk factor worth quoting because it is live and dated: "Government actions in various countries relating to tariffs affect the costs of some of our merchandise ... Higher tariffs are more likely to adversely impact rather than improve our results."

Capital returns

Dividends — and the vendor is a quarter behind. The 8-K filed 2026-04-15 discloses that the board "approved a quarterly increase from $1.30 to $1.47 per share, $5.88 on an annualized basis" — a 13.1% increase, declared 2026-04-15 and paid 2026-05-15. The 8-K filed 2026-07-08 confirms the $1.47 rate was maintained for the following quarter. The vendor's lastDiv and dividendPerShareTTM are both $5.54, a trailing figure. The forward run rate is $5.88, a yield of 0.62% rather than 0.584%. We use the filing.

And the special dividend, which changes the whole shareholder-yield calculation. The 10-K: "Cash dividends declared in 2025 totaled $2,183 or $4.92 per share, as compared to $8,589 or $19.36 per share in 2024. Dividends in 2024 included a special dividend of $15 per share, resulting in a payment of approximately $6,655."

This is the honest way to state Costco's shareholder yield: about 1% a year averaged across a special-dividend cycle, not the 0.58% the vendor's trailing dividend field implies and not the 0.62% the current run rate implies. The knowledge base makes the structural point: Costco "could support meaningful leverage given cash-flow durability but keeps a conservative, self-funding balance sheet ... returning cash via special dividends." With net cash back to roughly $11.7B and rising by $4-5B a year, another special dividend is a live possibility. No consensus estimate in the file contains one.

Buybacks are token and the share count barely moves. The programme is "a $4,000 authorization approved by our Board of Directors in January 2023, which expires in January 2027." In FY2025 Costco "repurchased 943,000 ... shares of common stock, at an average price per share of $957.66 ... totaling approximately $903", leaving $1,962M available at FY2025-end and $1,359M at 2026-05-10 after $603M of first-36-week repurchases at an average $945.46. Diluted shares outstanding: 444.8M (FY2021), 444.8M (FY2022), 444.5M (FY2023), 444.8M (FY2024), 444.8M (FY2025) — flat for five years. The buyback offsets stock compensation of $860M and does nothing more. Anyone modelling share-count shrinkage in this name is modelling something that has not happened in half a decade.

5. Cash flow

Operating cash flow grew 17.6% in fiscal 2025 and is up a further 17.6% year-on-year through thirty-six weeks ($11.133B against $9.468B). The 10-Q attributes the increase "to higher cash flow provided from operating income, as well as reduced net investment in merchandise inventories ... a result of faster inventory turns and improved payment terms with suppliers." That is high-quality cash generation: the working-capital contribution comes from turning inventory faster, not from stretching payables further.

Capital expenditure is real expansion, not maintenance. The FY2025 spend was $5,498M against depreciation of $2,426M — 2.27x, and the trailing figure is 2.41x. The 10-Q states the fiscal 2026 intention is "approximately $6,500", up 18% year on year, "as we continue to invest in new warehouse openings, remodel existing locations, expand our depot network, and further develop our digitally-enabled businesses." Under our half-weight rule this is a plan, not a fact; but it is corroborated by the $4,228M already spent in thirty-six weeks.

The warehouse cadence, from the filings — and it is running slightly behind the plan. The 10-K said Costco would "open up to 35 new warehouses, including five relocations, in 2026." The 10-Q reports "We opened 16 new warehouses, including two relocations, in the first thirty-six weeks of 2026, and plan to open 13 additional new warehouses, including one relocation, in the remainder of fiscal 2026" — 29, not 35. Against 27 opened in fiscal 2025 and 30 in fiscal 2024, this is a flat-to-modestly-higher build rate, not an acceleration.

Free cash flow and the yield. Trailing free cash flow is $8.80B on the vendor's per-share figures, a free-cash-flow yield of 2.09% and a price-to-free-cash-flow ratio of 47.7x — the same multiple as earnings, which is what you would expect from a company whose depreciation and capex are both real and roughly proportionate. Free cash flow covers the dividend 3.6x at the new $5.88 rate. There is no funding question anywhere in this business.

6. Valuation — priced in or room?

At $947.85 (market cap $420.4B, 443.5M shares):

Peer context. The vendor peer set contains three genuine warehouse-club comparables — Walmart ($887.7B), BJ's Wholesale ($12.3B) and PriceSmart ($6.0B) — plus Target ($67.3B) as a broadline retailer, and then a set of consumer-staples names of a different type entirely: Coca-Cola ($372.4B), Procter & Gamble ($351.9B), Philip Morris ($291.2B), Dollar General ($28.1B), Dollar Tree ($25.1B) and Ollie's Bargain Outlet ($4.7B). No peer multiples are supplied in the file, so no peer-multiple comparison is drawn. The knowledge base does supply dated third-party observations of the multiple, and we use those in 6b.

The vendor's scoring model rates Costco B+ with an overall score of 3. The composition is instructive: return-on-equity and return-on-assets sub-scores of 5 (the maximum), debt-to-equity 3, discounted-cash-flow 3 — and price-to-earnings and price-to-book sub-scores of 1, the lowest available. The vendor's own model says exactly what this dive says: excellent business, expensive stock.

6a. What today's price assumes (the inversion)

At $947.85 — 41.8x FY2027E consensus EPS of $22.682, and 47.7x trailing — the price embeds roughly the following falsifiable claims:

• Revenue compounds 7-9% annually through the fiscal year ending August 2028, reaching $350.494B. (Consensus-derived: $301.963B FY2026E on 26 analysts, $327.272B FY2027E on 25, $350.494B FY2028E on 21.) This is the best-supported assumption on the page — comparable sales are running at +10% headline and +7% on core merchandise, and 29 warehouse openings are already contracted for fiscal 2026.
• EPS reaches $24.971 by the fiscal year ending August 2028, i.e. +37.1% over the FY2025 actual of $18.21. (Consensus-derived, 21 analysts.)
• Membership fee revenue keeps growing at a double-digit rate after the September 2024 fee increase fully laps at the 2026-09-24 print. (Our number, derived from the filings' own disclosure that the increase contributed ~40% of membership income growth in FY2025, ~35% in Q2 FY2026 and ~25% in Q3 FY2026.) This is the most fragile assumption in the price. The mechanical arithmetic points to high-single-digit fee growth once the increase is out of the base, on a line that is 51.3% of operating income.
• The market keeps paying roughly 40x forward earnings for a business consensus expects to grow EPS 10.4% and then decelerate to 4.8% by fiscal 2030. (Our reading.) Note the market has already declined to keep paying 47x-plus: the stock is 13.4% below its 52-week closing high with earnings up 10-15% over the same period, which is by definition a de-rating in progress.
• U.S. and Canada renewal rates hold above 92% while the online-sold mix keeps rising. (Our number, from the filings: 92.3% at FY2025-end, 92.1% at Q2 FY2026, 92.2% at Q3 FY2026, with the company attributing the drift to online-sold memberships that "renew at a slightly lower rate on average.")

6b. The return bridge (why the multiple moves)

Expected return over 12-24 months decomposes as: EPS growth (+10.4% FY2027E, +10.1% FY2028E on consensus) + multiple drift + shareholder yield (~1.0% on a four-year average including specials; 0.62% on the current dividend run rate alone).

Our base case assumes multiple COMPRESSION, and we state the size of it precisely. Today the market pays 41.8x the fiscal year ending August 2027 — i.e. 41.8x the year-ahead estimate. Our base case of $930 pays 37.2x the fiscal year ending August 2028 — 37.2x the year-ahead estimate at that future date. On a like-for-like year-ahead basis that is roughly 11% of multiple compression over twelve to twenty-four months. The arithmetic reconciles exactly: +10.1% earnings growth − 11% multiple compression + 1.0% yield ≈ −1.9%, which is our base fair value against spot.

Four reasons the compression is our base rather than our bear. (a) The largest recent earnings tailwind expires on a known date — the fee increase laps at the 2026-09-24 print, and the filings document its decay from 40% to 35% to 25% of membership income growth. (b) Consensus itself models deceleration, from +12.8% (FY2026E) to +10.4% to +10.1% to +8.9% to +4.8% by fiscal 2030; a multiple in the forties is not usually sustained through a visible deceleration to single digits. (c) The compression has already started and the market is telling us so — down 13.4% from the closing high, below both moving averages, negative on every relative horizon while earnings grew double digits. (d) The justification most often offered for the multiple — that this is really a subscription business — is 24 points weaker than the version everyone quotes (Section 2b: fees are 51.3% of operating income, not ~75%).

Our knowledge base supplies dated third-party readings of the multiple, and they are the closest thing to a historical anchor we are permitted to use. One source recorded in January 2022 that "Costco re-rated from ~25x to over 40x earnings — stock outpaced the business itself." Another in August 2025 that "low-growth staples trade at Nvidia-like 35-42x multiples." Another in October 2025 quoted Costco at 50x and in December 2025 used it as the expensive benchmark against which technology looked cheap. The band the tracked sources record over the last four years is roughly 35x to 50x. Today's 47.7x trailing sits in the top quartile of that band; our base of 46.8x trailing / 37.2x forward sits in the middle of it.

The bull case at $1,090 requires 48.1x FY2027E, essentially the multiple the stock carried at its 2025 peak, and requires either a new fee increase, a special dividend, or both.

And here is the single sharpest observation available about the street. Consensus of $1,102.23 is 48.6x FY2027E. The published low target of $1,000 is 44.1x FY2027E. Today's price is 41.8x. Every published target in the file — including the most bearish one — requires the multiple to EXPAND from here. Not one of the 58 covering analysts has a target that assumes the de-rating continues, on a stock that has been de-rating for twelve months and closed today below both its moving averages. We think that band is anchored to where the stock has been rather than to what it now earns, and it is the reason our base case sits below the entire published range.

6c. Variant perception (where we differ, what would surprise)

• We differ on the durability of membership fee growth, and the filings — not our judgement — supply the evidence. The disclosure is explicit and it decays on a schedule: the fee increase effective 2024-09-01 contributed "approximately 40%" of membership income growth in FY2025, ~35% in the second quarter of fiscal 2026 and ~25% in the third, and membership fee growth itself has already stepped from +14% to +11%. At the 2026-09-24 print the contribution is zero. Consensus models FY2027 EPS growth of 10.4% on a line that is 51.3% of operating income and about to lose several points of growth. Watchable number: membership fee revenue growth at the 2026-09-24 and December 2026 prints — high single digits confirms us, sustained low double digits refutes us.
• We differ on what the headline growth rate is. The third quarter's 12% net sales growth was 2.21 points gasoline price and 1.04 points currency; core merchandise sales grew 7% and core merchandise gross margin, measured on core sales, fell nine basis points. Meanwhile the comparable-sales mix inverted from 5% frequency / 1% ticket in FY2025 to 2% frequency / 7% ticket. A screen showing 12% growth on a 47.7x multiple is showing you a petrol price. Watchable numbers: the frequency-versus-ticket split, and core merchandise gross margin on core sales, both disclosed every quarter.
• We differ from the consensus framing of the business in a way that is FAVOURABLE to Costco and still bearish on the stock. Membership fees are 51.3% of operating income, down from 65.2% in fiscal 2020, and 65.7% of net income, down from 72.3% in fiscal 2022. The merchandise operation is materially more profitable than the folklore says — operating margin has expanded from 3.35% to 3.77% in two years. But the corollary is that the "recurring subscription" share of profit is shrinking, and that is the specific argument used to justify a forties multiple. We think the business is better than its reputation and the stock is worse than its multiple, simultaneously. Watchable number: membership fee revenue divided by operating income, computable from every filing.
• We differ on the balance sheet and on the shareholder yield, both in Costco's favour, and neither changes the verdict. Net cash is roughly $11.7B on the 2026-05-10 10-Q, against a vendor headline of $5.988B — nearly double. And the true multi-year cash return is about 1.0% a year averaged across the special-dividend cycle, not the 0.58% the trailing dividend field shows. These are real corrections and they are worth roughly one turn of multiple, not eight.
• Positive surprise that would force us to re-rate upward: an announced membership fee increase, which the 10-K flags as a matter its forward-looking statements address and which no estimate in the file contains. On roughly $5.5B of annual fee revenue at very high incremental margin, a fee increase of the historical size is worth a material step in operating income. A second positive: a special dividend of the fiscal-2024 kind, which with $11.7B of net cash is affordable and would reset the shareholder-yield arithmetic. A third: renewal rates re-accelerating above 92.5% despite the online mix. Watchable events: the 2026-09-24 release; any 8-K under Item 8.01.
• Negative surprise that would break even our base case: core merchandise gross margin on core sales declining for a second and third consecutive quarter — it fell nine basis points in the third quarter of fiscal 2026 and the 10-K warns that "higher tariffs are more likely to adversely impact rather than improve our results." A second: shopping frequency falling below the +2% of the third quarter, which would mean the comp is entirely price. A third: membership fee growth below 8% at the September print. Watchable numbers: all three are in the quarterly MD&A.

Synthos fair values

All three anchors are multiples of FY2027 consensus EPS of $22.682 (fiscal year ending August 2027, 27 analysts), cross-checked against FY2028E of $24.971 (21 analysts) and the trailing $19.88.

• Bear ~$770 — 34.0x FY2027E. Cross-check: 30.8x FY2028E; 38.7x trailing. The scenario: the fee increase laps and membership fee growth falls to mid-single digits, core merchandise margin keeps slipping under tariff pressure, the frequency component of the comp stays at 2%, and the market re-rates a 7-8% grower toward the bottom of the 35-42x band the tracked sources record for staples. Note $770 is 9.4% below the 52-week closing low of $850.00 — this is a genuine bear case, not a whisper. −18.8%.
• Base ~$930 — 41.0x FY2027E. Cross-check: 37.2x FY2028E; 46.8x trailing; 12.3x book. The scenario: everything works, membership fee growth settles at high single digits, comps run 5-7% on core merchandise, EPS lands on consensus, and the multiple gives back roughly 11% on a year-ahead basis over 12-24 months. Sits 15.6% below the street consensus of $1,102.23 and 7.0% below the street's lowest published target. −1.9%, plus roughly 1.0% of shareholder yield — call it flat.
• Bull ~$1,090 — 48.1x FY2027E. Cross-check: 43.7x FY2028E; 54.8x trailing. The scenario: a new membership fee increase is announced, renewal rates hold above 92.5%, digitally-enabled sales keep compounding above 20%, and the market re-rates back to the multiple it paid at the 2025 peak. Roughly the street median of $1,100 and 1.1% below the street consensus. +15.0%.

Base is 1.9% BELOW spot; asymmetry is negatively skewed, roughly 0.8:1 (18.8% down, 15.0% up). You are being asked to accept slightly more downside than upside to own an outstanding business. That is the definition of a Watch, and there is a specific price — the $850-870 area, roughly 38x FY2027E and the 52-week closing low — at which the same arithmetic produces a Buy.

7. Knowledge base — 22 claims, near-unanimous on the business, near-silent on the stock

Tracked knowledge-base claims naming Costco: 28 raw hits, of which 22 are entity-level. Search covered COST (case-sensitive, with word boundaries) and Costco, across all 51,928 records.

Discards, and the specific false-positive risk we were asked to test. The search pattern included \bCOST\b case-sensitive, which can match the ordinary English word "COST" wherever a claim's stored text is capitalised. We inspected the match attribution on every one of the 28 hits. All 28 matched on \bCostco\b; the \bCOST\b pattern returned zero matches. There are no capitalised-common-word false positives in this lane and nothing was discarded on that ground. The six discards are all of a different kind: six hits name Costco in the claim text but do not carry it as a tagged entity, and they are excluded from the entity count. Four of the six are worth describing because the pattern is itself informative:

• Two use Costco as a valuation yardstick for another company — one arguing a pharmaceutical name "at ~30x is not expensive when Costco and Walmart trade at ~40x" (2026-01-16), another that a semiconductor name at "27x forward earnings" is "cheaper than Costco (50x) or Walmart (34x)" (2025-10-23), with a December 2025 repeat of the same construction. These are claims about Costco's multiple made in service of a thesis about something else, so they do not count as conviction on Costco — but they are the only dated external readings of the multiple in the file and we cite them in 6b as evidence, labelled.
• One is a competitive-risk claim about brands being disintermediated by retailer private label (2026-02-10), where Costco/Kirkland appears as the threat rather than the subject.
• One is an anecdote about grocery inflation (2024-01-09) with no analytical content.
• One uses Costco as an analogy for an airline's cost model (2023-05-27).

Stance mix among the 22 entity-level claims: 16 bullish, 3 neutral, 3 bearish, across only 6 distinct sources — business_breakdowns (9), we_study_billionaires (4), invest_like_the_best (3), compound_and_friends (3), lyn_alden (2), forward_guidance (1) — spanning 2020-12-27 to 2026-03-28. No management-lane (skill 0.5) claims appear in this file, so no half-weighting was required. One source carries skill 1.1; the rest are 1.0.

Two limitations, stated up front. Six sources is one of the two thinnest breadths in this batch, and nine of the 22 claims come from a single 2023 episode. And nothing in the lane is dated later than 2026-03-28 — the last four months of the de-rating, including the earnings miss and the loss of both moving averages, are entirely unrecorded.

The lane's central argument, and it is coherent:

> 2023-05-01 · business_breakdowns · bullish · conviction 90

> "Costco's moat is scale-economies-shared: run merchandise margin near zero, monetize via recurring membership fees, pass all savings to members in a self-reinforcing loop."

> 2023-05-01 · business_breakdowns · bullish · conviction 85

> "~75% of Costco operating income comes from membership fees at ~90% incremental margin, plus negative cash conversion cycle funds the customer experience."

> 2026-03-28 · we_study_billionaires · bullish · conviction 80

> "Costco's scale economics and exclusive bulk-supplier relationships let it undercut rivals on price; shrinking would destroy that cost differentiation."

> 2025-10-27 · compound_and_friends · bullish · conviction 70

> "Walmart and Amazon wipe out competitors; only Costco reliably survives the onslaught."

We have checked the two quantitative claims in that set against the filings and both need updating. The 75%-of-operating-income figure is now 51.3% and has fallen every year since fiscal 2020 (Section 2b) — the claim is closest to correct if read against net income in fiscal 2022 (72.3%), and even on that reading it is now 65.7%. The "negative cash conversion cycle" is now +0.46 days — still remarkable, but Costco no longer collects meaningfully before it pays. Both are the kind of drift that happens when a canonical description of a business outlives the data behind it, and it matters here because the description is load-bearing for the multiple.

The supporting structural claims we do accept, because the filings corroborate them:

> 2022-07-18 · invest_like_the_best · bullish · conviction 80

> "Costco earns profit almost entirely from membership fees; ~4,000 narrow SKUs drive massive volume and cost-plus pricing, insulating it from Amazon."

The stock-keeping-unit figure is confirmed verbatim by the 10-K: "We carry less than 4,000 active stock keeping units (SKUs) per warehouse."

> 2023-05-01 · business_breakdowns · bullish · conviction 75

> "Kirkland (~25-30% of sales, growing) delivers equal-or-better quality than branded peers, giving Costco a durable margin and loyalty advantage."

The 10-K corroborates the direction if not the percentage: "We believe that Kirkland Signature products are high quality, offered at prices that are generally lower than national brands, and help lower costs, differentiate our merchandise offerings, and generally earn higher margins. We expect to continue to increase the sales penetration of our private-label items." The 25-30% figure itself is not disclosed in the filings and we do not use it.

> 2023-05-01 · business_breakdowns · neutral · conviction 65

> "Costco could support meaningful leverage given cash-flow durability but keeps a conservative, self-funding balance sheet (Munger/Berkshire ethos), returning cash via special dividends."

Confirmed exactly, and it is the source of the special-dividend option we price in the bull case. The 10-K documents the $15-per-share special paid in fiscal 2024 at a cost of approximately $6,655M.

The three bearish and one neutral claims are all about the same thing — the price — and they are the most useful claims in the file:

> 2022-01-27 · lyn_alden (skill 1.1) · bearish · conviction 60

> "Costco re-rated from ~25x to over 40x earnings — stock outpaced the business itself, exemplifying investors paying record multiples for large-cap quality."

> 2025-08-08 · forward_guidance · bearish · conviction 65

> "Low-growth staples trade at Nvidia-like 35-42x multiples — a sign 401k/passive inflows are distorting valuations across the market."

> 2026-02-26 · we_study_billionaires · neutral · conviction 55

> "Great business (22% ROIC) but less attractive as a stock; can't reinvest all profits so must pay out dividends/buybacks, capping intrinsic-value compounding."

That last claim is the closest thing in the entire knowledge base to this dive's conclusion, and it is four months old. It is also nearly right on the return metric — the vendor's trailing return on invested capital is 19.0% against the claimed 22%, with return on equity at 28.3% and return on capital employed at 25.3%.

And the strongest counter-argument in the file, which we take seriously:

> 2026-03-26 · we_study_billionaires · bullish · conviction 70

> "Truly exceptional businesses with ever-increasing earnings (like Costco) can be held even while optically expensive; quality can justify a persistently high multiple."

This is the honest bull case for paying 47.7x, and it is the reason our verdict is Watch rather than Avoid. A business with a 92.3% renewal rate, 28.3% return on equity and a fifty-year record of not disappointing does deserve a premium, and the historical record of selling such businesses on multiple grounds is poor. Our disagreement is narrow: not that the multiple should be low, but that it should not be at the top of its own recorded four-year band at the exact moment its largest earnings tailwind expires.

Why Moderate rather than High. The lane is genuinely positive (16 of 22 bullish), coherent, and corroborated by the filings on every structural point we could test. But it is only six sources; nine claims come from one 2023 episode; nothing is dated after 2026-03-28; its single most-cited quantitative claim is 24 points stale; and it contains almost no information about the stock as opposed to the business. A knowledge base that loves a company and is silent on its price is a business endorsement, not an investment case, and we score it accordingly.

8. Data integrity — what we rejected from the vendor file and why

Nine items. Three are direct filing-versus-vendor contradictions, and in every one the filing wins.

1. FILING OVERRIDES VENDOR — the vendor reports deferredRevenue = 0.00; the 10-K reports $2,854M of deferred membership fees. The vendor's balance sheet shows deferredRevenue = 0.00 in FY2023, FY2024 and FY2025. The 10-K states: "Deferred membership fees at the end of 2025 and 2024 were $2,854 and $2,501." This is the forward book of the single most important revenue line in the business, growing 14.1%, and the vendor reports it as nil. We use $2,854M.

2. FILING OVERRIDES VENDOR — the balance sheet is eleven months stale and the true position is much stronger. The vendor's most recent balance sheet is dated 2025-08-31. The 10-Q filed 2026-06-03 reports "Cash and cash equivalents and short-term investments were $19,996 and $15,284 at May 10, 2026, and August 31, 2025" — a $4.712B, 30.8% increase that appears in no vendor field. Combined with the $69M of long-term debt repaid in March 2026 and $96M of short-term borrowings, net cash is approximately $11.7B rather than the $5.988B the vendor reports — a 95% understatement. We use the 10-Q.

3. FILING OVERRIDES VENDOR — the vendor's longTermDebt omits the current maturity. The vendor reports $5.713B. The 10-K states "As of the end of 2025, long-term debt with fixed interest rates was $5,805", with the current portion "included in other current liabilities." The $92M difference is the current maturity; total debt on the filing basis is ~$8.265B rather than $8.173B, and corrected FY2025 net cash is $7.019B.

4. netDebt ignores $1.123B of short-term investments — an 18.8% understatement of net cash. Vendor netDebt of −$5.988B is total debt ($8.173B) less cash and cash equivalents only ($14.161B). Correctly computed on the vendor's own balance sheet: net cash of $7.111B. enterpriseValueTTM of $409.638B carries the error; rebuilt it is $413.241B, a −0.9% residual — immaterial, so the vendor EV is accepted for scale with the caveat noted.

5. The est EBITDA and EBIT rows are systematically inflated and are NOT used. They pass the crude sanity tests — EBITDA exceeds EBIT and EBIT exceeds net income in every row — but they do not reconcile to reported results. The FY2025 estimate row carries ebitAvg of $12.798B against actual operating income of $10.383B (+23.3%) and actual EBIT computed as pretax plus interest of $10.972B (+16.6%); the FY2024 row carries $11.867B against $9.285B (+27.8%). The bias is consistent and material. Every forward multiple in this dive is computed on epsAvg only, in line with the standing rule, and no EBITDA figure from the estimates block appears anywhere on this page.

6. The vendor's gross margin is not the company's gross margin, and the difference is the membership fee. The vendor reports gross margin of 12.8% (FY2025), computed as gross profit over total revenue including membership fees, which have no cost of goods. The 10-K's own measure is gross margin as a percentage of net sales excluding membership fees, and it is 11.03% for FY2025 excluding gasoline deflation. Both are internally correct; they are not the same number and the vendor never labels which it is reporting. We use the filing's basis whenever discussing merchandise profitability and say which is which.

7. inc_q and earn_cal disagree on first-quarter fiscal 2026 EPS by $0.16, and we cannot settle it. The vendor's inc_q block reports diluted EPS of $4.50 for the quarter ended 2025-11-23 (net income $2.001B over 444.5M diluted shares — internally consistent), while earn_cal records the reported actual for the 2025-12-11 release as $4.34. Every other quarter in the series agrees exactly across the two blocks. The first-quarter 10-Q is not in the filing set, so the contradiction cannot be resolved from primary sources and we do not pretend otherwise. We use the inc_q basis, giving a trailing four-quarter diluted EPS sum of $19.88 and a 47.7x multiple; on the earn_cal basis the trailing sum is $19.72 and the multiple is 48.1x. The choice does not change any conclusion in this dive, and we state both.

8. The dividend fields are a quarter behind the filings. profile.lastDiv and ratios_ttm.dividendPerShareTTM both report $5.54, a trailing twelve-month figure. The 8-K filed 2026-04-15 discloses the increase "from $1.30 to $1.47 per share, $5.88 on an annualized basis", confirmed at $1.47 again by the 8-K filed 2026-07-08. The forward yield is 0.62%, not the vendor's 0.584%. Separately, neither figure captures the special-dividend cycle — the 10-K records a $15-per-share special costing approximately $6,655M in fiscal 2024, which makes the four-year average cash return approximately 1.0% of today's market capitalisation rather than 0.58% (Section 4).

9. Four minor internal inconsistencies, flagged and resolved. (a) quote.yearHigh of $1,096.50 and yearLow of $844.06 are intraday extremes; tech.hi52 of $1,094.32 and tech.lo52 of $850.00 are closing extremes. We use the closing series throughout. (b) quote.sma50 reports $959.83 while tech.sma50 reports $955.38 — a 0.47% disagreement on the same statistic. We use the tech series, which is computed from the same closing-price history as the rest of the technical block, and note that the conclusion is unaffected: the price of $947.85 is below the 50-day average on either figure. The 200-day figures agree to four decimal places. (c) inc_a reports FY2022 net income of $5.844B while cf_a reports $5.915B for the same year; the $71M difference is the non-controlling interest in the Taiwan operation, consistent with the $842M acquisition recorded in the FY2022 cash-flow statement. Both figures are correct on their own basis and we use the attributable figure of $5.844B. (d) Three of the eight insider records are null — the same officer, with a blank transaction type, zero shares, zero price — and are excluded from Section 10.

Not defects, correctly reported: the FY2023 growth rate of +6.8% and FY2024 growth of +5.0% look like a step-down but reflect the fact that fiscal 2023 was a 53-week year, per the 10-K. The seg_prod and seg_geo blocks are exceptionally clean — core merchandise growth of $19,086M and ancillary growth of $1,201M both tie to the 10-K's MD&A to the dollar, and U.S. and Canada revenue of $200,046M and $36,923M tie to the filing's critical-audit-matter paragraph exactly. After the segment defects found elsewhere in this batch, this is worth stating positively.

Non-equity tripwire: checked and clear. COST is common stock, beta 0.872, average volume 2.3M shares, dividend variable and rising (a 13.1% increase in April 2026 plus a documented history of large specials), and a 52-week closing range of $850.00-$1,094.32 — a 29% band. Nothing resembling a preferred or baby-bond profile.

9. Technicals — the quietest de-rating in the batch

• Price $947.85, down 0.65% (−$6.23) on the day, on volume of 2.2M against a 2.3M average — a slightly light session. The day's range was $937.01 to $950.86 against a $945.00 open and a $954.08 previous close: the stock opened lower and never recovered the prior close.
• −13.38% from the 52-week closing high of $1,094.32; +11.51% above the 52-week closing low of $850.00. Maximum drawdown from peak over the trailing year is −13.38%, identical to the current distance from the high — meaning the stock is sitting at very close to its worst point of the year.
• BELOW BOTH moving averages — −0.79% under the 50-DMA of $955.38 and −1.05% under the 200-DMA of $957.88. The two averages are within 0.26% of each other and the price is just beneath both. This is a coiled, unresolved structure, not a crash: a stock that has spent months converging on its own averages and has now slipped fractionally underneath them. It is one of only two names in this batch trading below both.
• RSI 62.3 — upper-neutral, and notably NOT oversold. MACD +0.24 — positive but negligible. The combination is unusual and worth naming: an RSI in the low sixties on a stock below both moving averages means the recent short-term trend has been up while the medium-term trend has been down. The stock has bounced 11.5% off its low without repairing the structure. There is no capitulation here to buy.
• Relative performance, and this is the whole story: 12-month −0.49% against SPY +24.26% and QQQ +30.80% — deficits of 24.7 and 31.3 points; 6-month −2.12% against SPY +11.09% and QQQ +15.71% (−13.2 and −17.8 points); 3-month −6.41% against SPY +7.59% and QQQ +7.67% (−14.0 points, and the worst three-month relative reading of the three horizons). The underperformance is not only large, it is accelerating.
• Practical read. A stock that has gone nowhere for a year while its earnings grew 10-15% has de-rated by roughly the amount of that earnings growth — from about 53x to about 48x trailing. That process is not obviously finished, and nothing in the chart says it is. There is no oversold reading to buy, no positive divergence, no reclaimed average, and no catalyst for 51 days. The first genuinely interesting technical level is the 52-week closing low of $850.00, 10.3% below spot — which is also, not coincidentally, close to where our arithmetic turns a Watch into a Buy. A reclaim of both moving averages on volume, with the September print behind it, would be the other route in.

10. Insiders — small, one-directional, and mostly mechanical

Three further records — all Sarah Catherine George, dated 2026-03-02, each with a blank transaction type, zero shares, zero price and zero securities transacted — are corrupt rows and are excluded (Section 8).

The honest reading is that this table contains very little signal, and what little there is points one way. Not a single open-market purchase appears in the file. Two genuine open-market sales occur, and they are small in both share count and dollar value: 885 shares at $957.45 and 700 shares at $993.00, together roughly $1.5 million. The Millerchip disposal is an in-kind tax withholding on vesting, which is not a decision. The Raikes transaction is a gift, which is estate planning rather than a market view.

Two observations, both weighted lightly. First, the two discretionary sales were both executed above today's price — at $993 in April and $957 in June, against $947.85 now — so on a pure mark-to-market basis both insiders sold well. Second, and more usefully: through a twelve-month period in which the stock fell 13.4% from its high and underperformed the index by 24.7 points, no officer or director bought a share on the open market. That absence is not evidence of anything on its own; management at a company like this is compensated overwhelmingly in restricted stock and has little reason to add. But we would have recorded conspicuous buying as a positive, and honesty requires recording its absence as a mild negative.

11. Verdict, kill-criteria and flip conditions

Watch.

The business is close to unimprovable. The price is close to unrewarding. Those two facts do not cancel; they define the position.

What is genuinely excellent here, stated without hedging. A 92.3% U.S. and Canada member renewal rate and 89.7% worldwide. 914 warehouses, of which Costco owns the building at 866. Fewer than 4,000 stock-keeping units per warehouse driving a 0.46-day cash conversion cycle — the merchandise is sold before it is paid for. Return on equity 28.3%, return on invested capital 19.0%, return on capital employed 25.3%, interest coverage 76.9x. Net cash of approximately $11.7 billion on the 2026-05-10 10-Q, nearly double the vendor's headline. Deferred membership fees up 14.1% to $2,854M. Digitally-enabled comparable sales up 21%. Operating margin expanded from 3.35% to 3.77% in two years — and, contrary to the folklore, membership fees are now only 51.3% of operating income, down from 65.2% in fiscal 2020, because the merchandise business genuinely got better. Capital expenditure at 2.4x depreciation. A dividend raised 13.1% in April and a documented history of large special dividends. Our knowledge base is 16-to-3 bullish and every structural claim in it that we could test against the filings held up.

What you are being asked to pay, stated without softening. 47.7x trailing diluted earnings of $19.88 and 41.8x the fiscal-2027 consensus of $22.682, on a business consensus expects to grow EPS 10.4% next year and 4.8% by fiscal 2030, with a total shareholder yield of about 1%. 12.56x book. The vendor's own quality model gives price-to-earnings and price-to-book sub-scores of 1 out of 5. And the largest single driver of the last two years' earnings growth expires on a known date: the September 2024 membership-fee increase supplied ~40% of membership income growth in fiscal 2025, ~35% in the second quarter of fiscal 2026 and ~25% in the third, and zero from the 2026-09-24 print onward.

Why Watch rather than Hold, and why not Buy. Base fair value of $930 is 1.9% below spot, and the asymmetry is negatively skewed at roughly 0.8:1 — 18.8% of downside against 15.0% of upside. There is no return in the base case. But this is not a Hold in the "fairly priced, own it if you own it" sense either, because the direction of travel is against the price: the stock is below both moving averages, has underperformed by 24.7 points over twelve months with the gap widening (the three-month deficit is 14.0 points), the multiple is at the top of the four-year band our tracked sources record, and the earnings tailwind is expiring. Watch is the accurate label: we want to own this business, at a price roughly 10% below today's, and we have specified where.

On the analyst target band, because it is the most striking single fact in the file. Consensus is $1,102.23, the median $1,100, the high $1,275 — and the low is $1,000, itself 5.5% ABOVE the current price. Expressed as multiples of the fiscal-2027 consensus, the street's most bearish published target is 44.1x and the consensus is 48.6x, against today's 41.8x. Fifty-eight covering analysts, and not one has a target that assumes the multiple stays where it is, let alone falls — on a stock that has been de-rating for twelve months and closed today below both its moving averages. We read that band as anchored to where the stock traded in 2025 rather than to what it will earn in 2027, and it is the single largest reason our base case sits below the entire published range.

On the 2026-09-24 print, 51 days away. This is outside any banner rule and we say so plainly: there is no earnings banner on this dive. But it is not a neutral date either. It is the 16-week fourth quarter, the full fiscal year, the first print with no fee-increase contribution at all, and the point at which membership fee growth reveals its underlying rate. Anyone establishing a position before 2026-09-24 is taking that specific risk deliberately, and the honest advice is to wait for it.

Pre-registered KILL criteria — what would remove this from the watch list entirely:

• Membership fee revenue growth below 6% at either of the next two prints, which would indicate the underlying sign-up-and-upgrade rate is materially weaker than the fee increase has been masking.
• U.S. and Canada renewal rate falling below 91.5%. It has run 92.3% / 92.1% / 92.2% across the last three disclosures; a break below 91.5% would be a change in kind, not a mix effect.
• Core merchandise gross margin, measured on core merchandise sales, declining for three consecutive quarters. It fell nine basis points in the third quarter of fiscal 2026 after rising 16 basis points in fiscal 2025, and the 10-K warns that "higher tariffs are more likely to adversely impact rather than improve our results."
• Shopping frequency growth falling below +2%, which would mean comparable sales are entirely price and mix.
• The warehouse opening cadence falling below the mid-twenties, against 24 net new in fiscal 2025 and 29 planned for fiscal 2026 — square-footage growth is the only volume lever and it is already the slower part of the algorithm.
• A leveraged acquisition or a change in the self-funding balance-sheet policy, which is one of the specific things the knowledge base identifies as constitutive of the business.

Pre-registered FLIP TO BUY criteria — what would make this a position, and at what price:

• A price in the $850-870 area — the 52-week closing low is $850.00 — which is roughly 38x FY2027E and produces double-digit expected returns on unchanged consensus. This is the primary trigger and it requires nothing from the company.
• An announced membership fee increase. The last was effective 2024-09-01; the 10-K explicitly flags "membership fee changes" as a matter within its forward-looking statements; no consensus estimate in the file contains one. On roughly $5.5B of fee revenue at very high incremental margin this is the largest available upside surprise, and it would justify buying at a materially higher price than $870.
• A special dividend of the fiscal-2024 kind. Net cash is roughly $11.7B and rising $4-5B a year; the fiscal-2024 special cost $6,655M and is fully documented in the 10-K.
• Membership fee growth holding at or above 10% at the 2026-09-24 print with the fee increase fully lapped, which would directly falsify the central bearish argument in this dive and would warrant an immediate upgrade.
• A reclaim of both moving averages (currently $955.38 and $957.88) on above-average volume after the September print, which would signal the de-rating had run its course.

Where COST fits in the Synthos Framework Portfolio. The defensive compounder sleeve, at a 0% weight today and a 2-3% target weight on a trigger, with the trigger pre-registered above at $850-870 or on an announced fee increase. The house standing rule that mega-cap incumbents are guilty until proven innocent on a forward basis is precisely what produces this verdict: on trailing quality Costco is close to the best business in the batch, and on forward arithmetic — decelerating consensus EPS growth, an expiring tailwind, a top-of-band multiple and a 1% yield — it does not clear the bar for capital today. Note also the low beta of 0.872 and the near-zero correlation of the business to anything cyclical: when this does become a position, it is a portfolio stabiliser, not a return driver, and it should be sized as one.

On overlap within the batch. Costco and Home Depot (also covered 2026-08-04) are both large-cap American retailers trading below their 200-day averages after twelve months of underperformance, and they are not the same trade. Home Depot is 17.8% below its closing high with a 24.7x trailing multiple, 21 hold and 4 sell ratings, an earnings report in fourteen days and $63.961B of net debt — a cyclical name whose problem is the housing cycle. Costco is 13.4% below its high with a 47.7x trailing multiple, one sell rating in fifty-eight, no report for 51 days and $11.7B of net cash — a quality name whose problem is arithmetic. The Costco question is "what multiple"; the Home Depot question is "when does the cycle turn." Logged as a tracked Synthos call (Watch) as of 2026-08-04 at $947.85.

Single biggest risk: that a 47.7x multiple meets a 10% grower whose largest recent tailwind has just expired. Nothing needs to go wrong operationally. If consensus is met exactly and the market simply pays the middle of the band our tracked sources record for large-cap staples rather than the top of it, the stock is worth $930; if it pays the bottom, $770. The business risk in this name is close to nil and the valuation risk is the whole position.

Most fragile assumption in the price: that membership fee revenue keeps compounding at a double-digit rate after 2026-09-24. The filings document the fee increase's contribution decaying from ~40% to ~35% to ~25% of membership income growth, and the growth rate itself stepping from +14% to +11%. That line is 51.3% of operating income. If it settles at 7-8% rather than 11-14%, roughly two points come off operating income growth — and at 41.8x forward earnings, two points of growth is worth considerably more than two points of price.

Provenance & disclosures

• Traceability: 22 entity-level knowledge-base claims name Costco (28 raw hits; 6 discarded as thesis-only mentions where the company is named in the claim text but not carried as a tagged entity; breadth 6 distinct sources; net conviction positive). Stance mix 16 bullish / 3 neutral / 3 bearish, spanning 2020-12-27 to 2026-03-28. The search pattern deliberately included \bCOST\b case-sensitive, which risks matching the ordinary word "COST" in capitalised claim text. We inspected the match attribution on all 28 hits: every one matched on \bCostco\b and the \bCOST\b pattern returned ZERO matches. No capitalised-common-word false positives were found and none were discarded on that ground. Of the 6 non-entity discards, three use Costco's multiple as a valuation yardstick for a different company (a pharmaceutical name, a semiconductor name twice), one is a private-label competitive-risk claim about brands, one an inflation anecdote and one an airline analogy; the three multiple-yardstick claims are cited in Section 6b as dated external readings of Costco's multiple, explicitly labelled as such and excluded from the conviction count. No management-lane (skill 0.5) claims appear in this file, so no half-weighting was required; one source carries skill 1.1 and the remainder 1.0. Searches covered COST (case-sensitive) and Costco across all 51,928 records. All quoted claims are verbatim from the stored claim text.
• Data as-of: fundamentals — the latest income statement used is the fiscal third quarter ended 2026-05-10, 10-Q filed 2026-06-03; the most recent annual report is the FY2025 10-K filed 2025-10-08 for the year ended 2025-08-31 · estimates 2026-08-04 · prices 2026-08-04 ($947.85; 50-DMA $955.38; 200-DMA $957.88; RSI 62.3) · knowledge-base claims 2026-08-04. Costco's fiscal year ends on the Sunday closest to 31 August, so every estimate row labelled 2026-08-31, 2027-08-31 and so on is an August fiscal year-end and is labelled as such throughout. Fiscal 2026 is a 52-week year ending 2026-08-30; fiscal 2023 was a 53-week year. Every figure comes from the Synthos vendor data file for COST or from the SEC filings in filings/COST/; no figure comes from memory, recall or external retrieval.
• Filings consulted: 10-K filed 2025-10-08 (fiscal year ended 2025-08-31); 10-Q filed 2026-06-03 (Q3 FY2026, ended 2026-05-10 — the primary source for the current position); 10-Q filed 2026-03-11 (Q2 FY2026, ended 2026-02-15); 8-K filed 2026-04-15 (dividend increased $1.30 → $1.47); 8-K filed 2026-05-28 (Q3 results release); 8-K filed 2026-07-08 (dividend maintained at $1.47). Targeted searches were run across all six for membership fee revenue, renewal rates, paid-member and cardholder counts, fee-increase timing and contribution, comparable sales, digitally-enabled sales, gross margin drivers, warehouse counts and openings, long-term debt and senior notes, operating and finance leases, not-yet-commenced leases, dividends and special dividends, share repurchase authorisation and remaining capacity, capital expenditure plans, legal contingencies and changed risk factors.
• Filing-versus-vendor overrides (three, all resolved in favour of the filing): (1) deferred membership fees — the vendor reports deferredRevenue = 0.00 in all three balance-sheet years; the 10-K reports $2,854M at FY2025-end, up 14.1%, and we use the filing; (2) the current balance sheet — the vendor's latest balance sheet is dated 2025-08-31, while the 10-Q filed 2026-06-03 reports cash and short-term investments of $19,996M at 2026-05-10 against $15,284M, so net cash is approximately $11.7B rather than the vendor's $5.988B, and we use the 10-Q; (3) long-term debt — the vendor's $5.713B omits the current maturity, and the 10-K states long-term debt with fixed rates was $5,805M, so we use $5,805M and corrected FY2025 net cash of $7.019B. A fourth override is smaller but affects the yield: the vendor's dividend fields report the trailing $5.54 while the 8-K of 2026-04-15 established a $5.88 annualised rate.
• Other data rejections (detailed in Section 8): netDebt corrected from −$5.988B to −$7.111B on the vendor's own balance sheet (omits $1.123B of short-term investments), with enterpriseValueTTM correspondingly $409.638B against a rebuilt $413.241B (−0.9% residual, accepted for scale); est EBITDA and EBIT rows rejected outright as systematically inflated (FY2025 ebitAvg of $12.798B against actual operating income of $10.383B, +23.3%; FY2024 $11.867B against $9.285B, +27.8%) — all forward valuation runs on epsAvg only and no EBITDA figure from the estimates block appears in this dive; the vendor's gross-margin basis (total revenue including membership fees) distinguished from the company's (net sales excluding fees, 11.03% ex-gasoline for FY2025); an unresolved $0.16 disagreement between inc_q ($4.50) and earn_cal ($4.34) on Q1 FY2026 EPS, with both trailing multiples stated (47.7x and 48.1x); closing-price 52-week extremes preferred over intraday ($1,094.32/$850.00 rather than $1,096.50/$844.06); tech.sma50 of $955.38 preferred over quote.sma50 of $959.83; the FY2022 net-income difference between inc_a ($5.844B) and cf_a ($5.915B) identified as the Taiwan non-controlling interest; three corrupt insider records excluded.
• Estimate coverage: FY2027E, the primary forward anchor, rests on 27 analysts on EPS and 25 on revenue; FY2026E on 24 and 26; FY2028E on 21 and 21; FY2029E and FY2030E on 9 analysts each on EPS and 10 on revenue. No conclusion in this dive rests on the FY2029 or FY2030 rows, which are cited only to show the shape of consensus deceleration to +4.8%. Note also that FY2026E is effectively three-quarters actual ($14.01 reported plus a $6.57 consensus fourth quarter), so the 46.1x "forward" multiple on it is functionally trailing.
• Peer note: the vendor peer set contains three genuine club/warehouse comparables — Walmart ($887.7B), BJ's Wholesale ($12.3B) and PriceSmart ($6.0B) — plus Target ($67.3B) as a broadline retailer, and six consumer-staples or discount names of a materially different type: Coca-Cola ($372.4B), Procter & Gamble ($351.9B), Philip Morris ($291.2B), Dollar General ($28.1B), Dollar Tree ($25.1B), Ollie's Bargain Outlet ($4.7B). No peer multiples are supplied in the file, so no peer-multiple comparison is drawn. The only dated third-party readings of Costco's own multiple available to us come from the knowledge base and are cited, labelled, in Section 6b.
• Fair-value caveat: the $770 / $930 / $1,090 anchors are multiples of FY2027 consensus EPS of $22.682 for the fiscal year ending August 2027 (34.0x / 41.0x / 48.1x), cross-checked against FY2028E of $24.971 (30.8x / 37.2x / 43.7x) and the trailing $19.88. Stated arithmetic, not a discounted cash flow. The base case assumes roughly 11% of multiple compression on a like-for-like year-ahead basis over 12-24 months, which is named explicitly in Section 6b as the load-bearing assumption; at an unchanged 41.8x year-ahead multiple, consensus FY2028 earnings imply approximately $1,044, or +10.2%. Readers who believe the multiple holds should use that figure rather than ours.
• Timing: the third quarter of fiscal 2026 was reported 2026-05-28 with the 10-Q filed 2026-06-03, and the FY2025 annual report was filed 2025-10-08. The next print is 2026-09-24 — 51 days away — and no earnings banner applies. That print is nonetheless a specific input to the verdict, because it is the first quarter with no contribution from the September 2024 membership-fee increase.
• Accessibility note: all tables are labelled, all emphasis is carried by text rather than colour, and every figure is stated in full rather than encoded in a chart.
• Not investment advice. Independent research, educational and informational only, never personalised. No recommendation to buy, sell or hold any security is made to any person.
• Version: 2026-08-04-full.