COST Learning Companion: How to read the Costco research

The COST research family (page counts are those of the approved PDF editions, which this site does not host):

• Deep Dive: A very good business, priced for a better one (58 pages). Public preview; the Deep Dive page says how to read the full report.
• Research Note: Costco at the $1,040.55 valuation close: the franchise is not in doubt, the price is (7 pages).
• Technical Appendix: Method, schedules, bridge and sensitivities (29 pages).
• Learning Companion: How to read the Costco research (12 pages).

---

This edition teaches the reasoning behind our Costco Deep Dive and Research Note. It uses the same numbers, shows where each one comes from, and says where our judgment starts. If you read only one idea, make it this one: a valuation is an argument with arithmetic inside it, and the arithmetic is the easy part.

1. Five kinds of statement

Company-level dollar amounts print in billions of U.S. dollars, or in millions where an amount is smaller than a billion, and the unit is always written; per-share values, prices and dividends are in dollars.

Every number or claim in our research is one of five things, and a good reader asks which. A reported fact is a figure in a company filing, such as FY2026 net sales of $297.2 billion, or a figure published by an official source, such as the 4.91% Treasury yield. A management statement is something the company says about itself or its plans: Costco's intention to spend about $7.5 billion on capital in FY2027 is a plan, not a result. A calculation is our arithmetic on reported facts, such as the 1.94% merchandise margin. An assumption is an input we choose because nothing reports it, such as the beta of 0.85 and the equity premium of 4.5% behind our discount rate. A judgment is our opinion, and we try to say so in the sentence that makes it.

A checksum on a file, or a tidy table, proves which document we read and that we added correctly. It does not prove the assumptions are sensible. That is why we keep the five kinds apart.

2. How Costco makes money

Most retailers earn their profit from the gap between what they pay for goods and what they charge. Costco earns a thin version of that gap and collects membership fees on top. In FY2026 membership fees were $5.91 billion against $11.69 billion of operating income, a ratio of 50.6%. Read that ratio carefully: it divides one revenue line by a profit total, so it is arithmetic and not a profit margin that Costco reports for its fees.

Here is the reconciliation, which you can repeat. Operating income of $11.69 billion less membership fees of $5.91 billion leaves $5.78 billion. Divide by net sales of $297.2 billion and you get 1.94%. We call that the merchandise margin, and it is an analytical residual: it places all of the shared cost of running the warehouses on the sale of goods, so it is not an audited profit for the goods on their own. It is defined as gross margin less SG&A, with the fees kept apart: gross margin was 11.09% of net sales and SG&A was 9.15%, and 11.09% less 9.15% is the same 1.94%. Costco's own operating margin on net sales, 3.93%, is higher because it includes the fees. Do not compare our 1.94% with a gross margin you see elsewhere, because they measure different things; the three numbers fit together as a bridge, and the amount left unexplained is zero.

Why model the two lines separately? Because they behave differently. Fees are collected once a year from paid members, whose annual fee includes a free household card, and the renewal rate is 92.3% in the U.S. and Canada and 89.8% worldwide, so the fees are predictable and need little capital. The merchandise margin is a choice management makes each year about how much of its efficiency to pass on in lower prices. Ten basis points of that margin on FY2026 net sales are worth $297 million of operating income, 5.0 times the $59 million that one point of fee growth adds, which is why our scenarios move both.

3. A scenario is not a probability

We show three paths: a downside of $305.08 a share, a reference of $428.58 and an upside of $592.69. They are conditional. Each says "if these assumptions prove right, the business is worth this much". We attach no probabilities and we do not say the price will stay between the ends. The reference is not an average of the other two. It is the path we think the record supports best, and the downside and upside are chosen to be coherent rather than extreme: the downside keeps the margin at 1.55% for good, which is an enduring impairment, and the upside combines several things going right at once.

A common mistake is to read the middle scenario as a forecast and the price as a prediction error. Ask instead which assumptions would have to be true for each value, and whether the evidence supports them. Notice also that all three scenarios use the same discount rate. The scenarios already differ in their cash flows, and adding scenario risk to the rate as well would count the same risk twice.

4. Free cash flow and why Costco's grows when it grows

Free cash flow is the cash left after the business pays for the capital it needs. For FY2026 it was operating cash flow of $15.82 billion less capital expenditure of $6.44 billion, or $9.38 billion. Two features make Costco unusual. Suppliers fund more than all of its inventory, so growth does not tie up cash, it frees some. And members pay fees in advance, which also grows with the business. Both help cash flow in a growing year and would hurt it if growth stopped.

We were careful not to overstate that. FY2026's working-capital inflow does not repeat at that size: our reference carries a working-capital and deferred-fee flow of $1.28 billion in FY2027 and between $1.14 billion and $1.33 billion in each of the ten years, so free cash flow falls to $6.08 billion in FY2027, when capital spending rises to the planned $7.5 billion, before recovering to $9.94 billion in FY2029.

5. The discount rate: building it, and saying what is assumed

A dollar received later is worth less than a dollar today. The discount rate is how much less. Value depends on it so strongly that, in this model, it matters more than any operating assumption. At 8.66% the reference is worth $428.58. At 7.65% it is worth $515.76 and at 10.31% it is worth $337.89.

We build the rate from parts. The first is the risk-free rate: the yield on a 10-year U.S. Treasury security, which the Treasury publishes every day. On 11-02-2026 it was 4.91%. The second is a premium for owning equities rather than government bonds. We use 4.5%. The third is how risky this company is relative to the market, called beta; we use 0.85, a little less than the market. The cost of equity is the risk-free rate plus beta times the premium: 4.91% plus 0.85 times 4.5% is 8.735%. Debt costs the risk-free rate plus a spread of 0.60%, or 5.51% before tax, and debt is only 1.61% of Costco's market-value capital ($7.58 billion of debt and finance leases against $463.3 billion of equity at the close). Blending the two gives 8.66%.

Be clear about which parts are measured. The Treasury yield is a published fact. Beta, the premium and the debt spread are our assumptions. Nobody reports Costco's beta for the future, and we have no regression beta to measure it, so we show a band instead: beta from 0.70 to 1.00 and a premium from 4.0% to 5.5% give 7.65% at the low end and 10.31% at the high end. The honest way to use a built rate is to say which inputs are judgment and let the reader see the neighbors.

Two consistency rules are worth learning. Flows and rate must match: our cash flows are the whole firm's, before interest, so we use a rate that blends equity and debt, and they are nominal, so the rate is nominal too. Terminal growth of 3.0% is a little more than the 2.29% inflation the bond market implies (the 4.91% nominal yield less the 2.62% inflation-protected yield), and below the risk-free rate, which is where it belongs.

6. The terminal value, and a check worth repeating

A ten-year forecast cannot capture a business that lasts longer, so we add a terminal value for everything after year ten. We use the value-driver form: next year's after-tax operating profit times one minus growth divided by the return on new capital, all divided by the discount rate less growth. The part to understand is the reinvestment term. A company that wants to grow at 3.0% a year has to put money back, and how much depends on how much each new dollar earns. At an 18% return on new capital, the reinvestment rate is 3.0% divided by 18%, or 16.7% of profit.

The responsible check is to put the last forecast year beside the first terminal year. In the last explicit year the reference reinvests 8.5% of its profit; in the first terminal year it reinvests 16.7%. The terminal is more demanding than the forecast, and that is on purpose. It reflects our judgment that returns on new capital will fade from today's levels (Costco measured 45.0% on operating capital in FY2026 on the broad basis the model uses and 29.8% on a narrower one). We do not measure where they settle, so we also show 12%, 25% and 30%: the reference is worth $404.17, $442.24 and $448.10 at those returns.

The terminal value is 60.2% of the reference enterprise value. Most of any long-lived business's value is in the years beyond the forecast, so a model that does not examine its terminal has not finished.

7. From enterprise value to a share price

The forecast values the operating business: $177.8 billion for the reference, which is $70.7 billion from the ten forecast years and $107.1 billion from the terminal. To get to the shareholders you add what they own and subtract what they owe. We add cash and short-term investments of $21.3 billion but remove $2.84 billion of unsettled card receivables, because those are money owed by card processors for sales made in the last few days and are part of running the business. That leaves $18.5 billion. We subtract $6.16 billion of long-term debt and $1.41 billion of finance-lease liabilities. The result, $188.7 billion of equity divided by 445.266 million diluted shares, is $423.83 a share at the balance-sheet date, 08-30-2026.

The price we compare it with is dated 11-02-2026, 64 days later. Between the two dates the value of the business grows at the discount rate, so we multiply by 1.01467, which is 8.66% compounded for 64 days. Costco also went ex-dividend for $1.47 a share, with the record date on 10-30-2026. Someone buying on 11-02-2026 no longer receives it, so we subtract it. The arithmetic is $423.83 times 1.01467 less $1.47, which is $428.58. We treat 11-02-2026 as ex-dividend from the settlement convention, and we say that is our reading and not a statement in the filing.

A value has a date and a price keeps moving, so the two are best read with their dates beside them. The November 9, 2026 regular-session closing price of COST was $1,066.18, 2.5% above the $1,040.55 we compared our values with, and the 10-year Treasury yield was 4.86% on 11-06-2026, the last reading we have, against the 4.91% in our rate. We have not carried the values forward or rebuilt the rate. The reference of $428.58 is 59.8% below the later close instead of 58.8% below the earlier one, a difference that is small beside the gap itself, and the yield's five basis point fall is worth only a few dollars a share on the reference.

8. Reverse valuation: what the price fits

Instead of asking what the business is worth, you can ask what combination of assumptions makes it worth the price. We keep the reference's other drivers fixed and solve for one. The 11-02-2026 close of $1,040.55 matches the reference at a discount rate of 5.24%, or at a long-run merchandise margin of 7.72%, or with FY2036 operating income 2.49 times the reference's. These are conditional break-even combinations. They do not say what investors believe and they do not say the price is wrong. Solving the same way at the November 9, 2026 regular-session closing price of $1,066.18 moves each a little in the direction a higher price implies: 5.19%, 7.96% and 2.55 times.

Translate the margin into the company's own terms to see the size of the ask. A 7.72% merchandise margin in FY2036 is an operating margin on net sales of 9.71% against 3.93% in FY2026. It needs a gross margin of 16.9% with SG&A unchanged at 9.15%, or SG&A of 3.4% with gross margin unchanged at 11.09%. That is how far the cash flows would have to move, not a difference in definitions. And compare the 5.24% rate with the 4.91% Treasury yield: it is only about 33 basis points more than the government pays for ten years, and less than the 5.31% and 5.28% on 20-year and 30-year bonds. A fair reading is that the price fits many combinations, and that each one asks for something the record does not yet show.

9. Comparing companies on one definition

A comparison is only as good as the definition behind it. To compare Costco with its club competitors we put each on the merchandise margin used in this edition: net sales less merchandise costs less SG&A, over net sales, with membership income left out. Worked example for Sam's Club U.S. (Walmart's segment, year to 01-31-2026): its gross margin was 11.35% of net sales and its SG&A 11.44%, so the merchandise margin is 11.35% less 11.44%, which is -0.09%. BJ's comes to 1.51% and Costco to 1.94%. The lesson is that Sam's Club roughly breaks even on goods and relies on its $2.525 billion of membership and other income, while Costco earns 1.94% on goods before its fees. Read the ratio of fees to operating income the same way: 50.6% at Costco, 61.2% at BJ's and 103.4% at Sam's Club U.S. are arithmetic, not disclosed fee profits, and for Sam's Club the income includes "other" items. Two companies' fiscal years end months apart, and each allocates shared costs its own way, so these comparisons show scale and not decimals.

10. A member and a warehouse as units

Two units sit behind the revenue lines. For the member, divide by the people who pay, which is paid members and not the free household cards. In FY2026, on average paid members, fee revenue was $71.56 a member, net sales $3,601 and operating income $141.55. Fee growth of 11.0% can be taken apart, and the pieces multiply: average paid members grew 5.0% and fee revenue per member grew 5.7%, and 1.050 times 1.057 is 1.110, or 11.0%. Costco says about 30% of the fee growth came from the September 2024 increase, which helps separate price from members. For the warehouse, Costco's cohort table shows that a warehouse opened in FY2026 sold $187 million (annualized) against a fleet average of $292 million, or 64%, and that older cohorts grow into the average and beyond. A new warehouse therefore adds less to sales in its first year than its share of the count suggests, and the table shows no cost to build one, which we leave unknown.

11. Bridges for growth

Reported growth mixes things, and a bridge separates them. FY2026 net sales grew 10.1%. Costco reports that comparable sales, at warehouses open a year or more, added $22.5 billion and that new warehouses and other differences added the rest, which is 8.3 points and 1.8 points, and 8.3% plus 1.8% is 10.1%. Costco also says higher gasoline prices added $3.50 billion (1.3 points) and currencies $1.33 billion (0.5 points), so total net sales growth without those two was 8.3%. A bridge is only valid between measures of the same kind: that 8.3% and our 8.0% for FY2027 are both full-year totals including new warehouses, but a nine-week comparable rate excludes new warehouses and covers a fraction of the year, so it cannot validate or refute either.

12. Reading the bull and bear tables

The Deep Dive prints lenses for the strongest case on each side. Each is a model value under stated inputs, not a scenario and not a probability. To read one, find the input that differs from the reference and ask whether the record supports it. A 6.00% discount rate gives the reference $785.29 and the upside $1,276.41, but the rate is below the band we support, so the table says what that belief implies and not that it is right. The capital-spending lens works the same way: if capital spending stayed at FY2026's share of total revenue, which is 2.12%, the reference would lose $37.12 a share. It is read from the stored schedule and changes no value, and it shows how a single assumption can matter as much as a visible one.

13. What a rating is

The Blockhaven Rating runs from 1 to 5 and is printed with words: COMPELLING, FAVORABLE, BALANCED, CAUTIOUS and POOR. Costco is rated 1 out of 5, POOR, at $1,040.55 on 11-02-2026. A few properties matter. It is tied to a price, so the same business at a lower price could deserve a higher rating and nothing about the business would have changed. It is tied to a horizon, here the ten years of the model. It is not a price objective and it is never a recommendation to transact. And it is separate from two other things we print, the thesis (our one-sentence reading of where the business stands) and the near-term outlook.

Here the rating is 1 because the price sits above even our upside case at every rate in the band, including the most favorable end, and because reaching it takes economics well beyond what the record shows or a required return barely above a government bond. The market's two-year record above $850.00 is a fact about the market, and we treat it as the strongest argument against us. Your own view may differ, and you can see which inputs would make it so.

14. Test yourself

These exercises use only numbers in the research.

First, rebuild the 1.94% merchandise margin from the three figures in section 2, then rebuild it again as 11.09% less 9.15%. Second, find the P/E at the close: $1,040.55 divided by FY2026 diluted EPS of $20.76 is 50.1 times. Excluding the $0.15 tariff benefit, EPS is $20.61 and the multiple is a little higher. Third, ask what the dividend yield is: four quarters of $1.47 divided by the close is 0.57%, so almost all of an investor's return would have to come from the business growing. Fourth, rebuild the discount rate: 4.91% plus 0.85 times 4.5% is the cost of equity, and the blend with debt gives 8.66%. Then change beta to 1.00 and the premium to 5.5% and see the high end of the band. Fifth, take the reference value of $428.58 and the low close in the data we captured, $850.00. The reference is 49.6% below that low. What would have to be true of the discount rate, the margin or the duration of growth for the market's two-year range to be reasonable? Our answer is that the rate would have to be near 5.24% or the profits would have to be about two and a half times ours, and you should decide whether you find either convincing.

Glossary

• Net sales: sales of goods and services before membership fees.
• Merchandise margin: net sales less merchandise costs less SG&A, divided by net sales, which equals operating income less membership fees over net sales; our measure and not a reported one.
• EV (enterprise value): the value of the operating business before cash and debt.
• NOPAT: after-tax operating profit.
• Return on new capital: profit earned on each additional dollar the business invests.
• Free cash flow: operating cash flow less capital expenditure, in the history; in the forecast, after-tax operating profit plus depreciation less capital spending and working-capital growth.
• Beta: how strongly a stock has tended to move with the market; a judgment input here.
• Equity risk premium: the extra return investors expect from equities over government bonds; a judgment input here.
• Ex-dividend: from this date a buyer no longer receives the declared dividend.
• IEEPA tariffs: import duties imposed under the International Emergency Economic Powers Act, which the U.S. Supreme Court ruled unlawful in February 2026; Costco says it paid about $500 million of them.
• Comparable sales: sales at warehouses open at least a year, which leave out new warehouses.
• Cohort: the warehouses opened in one year.
• Paid member: a member who pays the annual fee; each paid membership includes a free household card.

Assumptions and disclosure

Company-level dollar amounts print in billions of U.S. dollars, or in millions where an amount is smaller than a billion, and the unit is always written; per-share values, prices and dividends are in dollars. The scenarios are conditional and carry no probabilities, and the rating is our assessment at one price and horizon, never a recommendation to transact.

Important disclosures

Blockhaven research is provided for general informational and educational purposes. It is not individualized investment advice and does not take into account any reader's financial circumstances, investment objectives, risk tolerance, tax situation or portfolio.

Nothing in this report is an offer or a solicitation to buy or sell a security or other financial instrument. Reading or accessing Blockhaven research does not by itself create an investment-advisory, fiduciary, brokerage or client relationship.

Investing involves risk, including the possible loss of principal. Scenario values, terminal values, valuation ranges, sensitivities and forward-looking assumptions are analytical tools, not guarantees or promises of future performance, and actual results may differ materially.

Blockhaven uses information from company filings, company publications, market-data providers and other sources it believes to be reliable, but that information may be incomplete, delayed, revised or incorrect.

Opinions, ratings and interpretations reflect Blockhaven's research judgment as of the report date and may change as new evidence becomes available.

Positions and conflicts

As of November 6, 2026, Blockhaven's owner held no shares, options, or other direct derivatives in Costco Wholesale Corporation; had no material positions in Costco Wholesale Corporation through entities the owner controls; and had no paid financial relationships with Costco Wholesale Corporation.

© 2026 Blockhaven LLC. All rights reserved.