LULU: ten times earnings, no funded debt, and a buyout that turns on one question

lululemon athletica inc. (NASDAQ: LULU)

Struck at the September 4, 2026 close of $100.61 — the first close that reflects the second-quarter print released after the September 3 close. Every figure on this page is computed from filed documents and company guidance.

Each row is the same discounted cash flow — same balance sheet, same cost of equity of 8.69%, same 2.0% terminal growth — run on a different view of the run rate. The two figures in the value column are the perpetuity-growth terminal and the exit-multiple terminal; neither is a target price, and the Institute publishes ranges rather than points because the inputs cannot carry a point. The bottom row is the same model run backwards: the permanent level of EBITDA at which it returns exactly the September 4, 2026 close. Cases and methodology are in Sections 9 and 11.

The whole case reduces to one question, and it has to be answered before any of the arithmetic means anything: is FY2026 the run rate, or is FY2025? The company earned $2,707 million of EBITDA in the year ended February 1, 2026, on a 24.4% margin. The guidance issued September 3, 2026 implies $1,966 million on a 18.9% margin once the non-recurring tariff refund is removed. Those two numbers are 38% apart. The enterprise value sitting on top of them does not change between the two. Every valuation in this case, and every buyout return in Section 9, is a consequence of which denominator the reader accepts.

The market has already answered, and its answer is neither. Run the discounted cash flow backwards — hold the price at the September 4, 2026 close and solve for the operating result that price requires — and it resolves to EBITDA settling permanently at about $1,621 million: 60% of what the company earned last year, 82% of an FY2026 estimate that has already been cut, and a 15.6% margin the company has not printed in any year examined here. That level sits between the Institute’s own bear-case terminal of $1,538 million and its base-case terminal of $1,984 million. It is a claim of permanent impairment, not a claim about a bad year.

The Institute does not accept it. On the Institute’s own operating case — the one in which the decline stops and nothing more — the equity is worth $116 to $136 against a price of $100.61. On the reading that FY2025 is the ordinary year and FY2026 the aberration, it is worth about $205. Both are above the price, and the case is published with both rather than with an average of them, because averaging would hide the disagreement and the disagreement is the finding. The distance between them is also not mainly a demand problem: of the $741 million that separates the two years, $576 million — 78% — is margin and only $165 million is revenue. Revenue recovery requires customers to return, which no owner controls. Margin recovery is assortment, pricing, markdown discipline, sourcing and overhead, which is exactly what a control owner buys the right to fix. That asymmetry is why Section 9 finds a buyout that fails on the Institute’s case and works comfortably on a buyer’s.

What the Market Is Paying, and What It Is Paying It For

The capitalisation is set out first because it is the half of the calculation nobody disputes. There is no funded debt, the cash is real, and the share count is off the 10-Q cover. Whatever disagreement exists about this company, it is not about the numerator.

Measurement dates. Share price, market capitalisation, both enterprise values and the drawdown are struck at the September 4, 2026 close; the one-session change is the September 3 close to the September 4 close. The share count is the Form 10-Q cover of August 28, 2026 and includes the exchangeable shares of Lulu Canadian Holding, Inc. Cash, the absence of funded debt and the lease liabilities are read from the balance sheet of August 2, 2026 in the Form 10-Q filed September 3, 2026 (Note 4 for the debt). Cash net of the post-quarter buyback and net cash per share are struck at August 28, 2026. Sources: Form 10-Q for the quarter ended August 2, 2026; Form 10-K for the year ended February 1, 2026, filed March 17, 2026.

Two features of that table do work later in the case. Net cash of $11.48 a share is 11.4% of the share price, so a buyer of the equity at $100.61 is paying $89.13 for the operating business and taking the cash alongside it — Section 9 sets out what a financial sponsor does with a balance sheet in that condition, and it is not what a public shareholder does with it. And the enterprise value is shown twice because the operating leases are $2,141.1 million of obligation that the accounting keeps off the funded-debt line while the rent that services them sits inside operating expense. Neither presentation is wrong. Pairing one of them against the wrong earnings figure is, and it is the most common error made on retailers: capitalise the leases in the numerator and the rent has to come back in the denominator, or the same obligation is counted twice.

The same enterprise value against two run rates

The denominator is the argument, so it is printed twice. Nothing in the left-hand block changes between these two columns — same shares, same cash, same absence of debt, same leases, same $9.87bn enterprise value excluding leases. Only the year changes. Read the shaded multiple rows against each other and the size of the question becomes visible: on the guided year this is a cheap company, and on last year’s actual result it is priced at a level normally reserved for businesses in liquidation.

The FY2026 column is company guidance issued September 3, 2026, at the midpoint of the stated ranges, with the $134.5 million tariff refund recognised in the second quarter removed — it is a non-recurring recovery and leaving it in would flatter both the margin and the multiple. The FY2025 column is the year ended February 1, 2026 as reported in the Form 10-K filed March 17, 2026; EBITDA is operating income plus depreciation and amortisation. Both enterprise values are struck at the September 4, 2026 close and are identical across the two columns. Multiples are lease-consistent: enterprise value excluding the lease liability is paired with EBITDA after rent, and enterprise value including it with EBITDAR before rent; mixing the two double-counts the leases. Free cash flow is net income plus depreciation and amortisation less capital expenditure on both columns, so that they are comparable. On the cash-flow-statement convention, FY2025 operating cash flow less capital expenditure was $922 million — $473 million lower, the difference being working capital, which is sized in Section 11 and carried in neither cash-flow model.

A distressed multiple on an undistressed balance sheet. That sentence is the entire case in nine words. A company carrying no funded debt whatsoever and $1.27bn of net cash — $11.48 a share, or 11.4% of the stock price — is being valued at 5.0x of clean EBITDA and 11.5x clean earnings. Multiples in that range normally accompany a capital structure under strain. There is no capital structure here to be under strain. Whatever the market is worried about, it is not solvency.

A quarter that beat, and guidance that did not

Second-quarter revenue of $2.42bn was 4.3% below the $2.53bn the company reported a year earlier, and operating income of $453.7 million was 13.4% below the prior-year $523.8 million. Gross margin held at 60.5% and operating margin at 18.8%. Diluted earnings of $2.92 a share compared with $3.10. On its own that is a decelerating quarter at a company that remains substantially profitable.

The guidance did the damage. Full-year revenue guidance was reset to a midpoint of $10.43bn from a prior midpoint of $11.07bn — a cut of 5.9%. Full-year diluted EPS guidance was reset to a midpoint of $9.61 from $11.05 — a cut of 13.1%. Earnings guidance was cut more than twice as hard as revenue guidance, which is the signature of operating deleverage against a fixed cost base rather than of a demand problem alone.

Guidance as issued with second-quarter results on September 3, 2026, against guidance as previously issued. Q3 figures are as guided; Q4 figures are the residual implied by full-year guidance less reported first-half results less guided Q3. Guidance excludes the effect of any further share repurchase.

The residual matters. Backing reported first-half results and guided third-quarter results out of the full year leaves an implied fourth quarter of roughly $3.23bn of revenue and $4.06 a share. The full year is being carried by a quarter the company has not yet entered.

Building the multiple without double-counting the leases

This is the single most common construction error applied to a store-based retailer after ASC 842, and it is worth being explicit about because it is what produces the difference between a stock that looks expensive and a stock that looks cheap.

Under ASC 842 the operating lease cost of $406 million sits inside operating expenses. It has already been deducted before arriving at reported operating income and therefore before arriving at EBITDA. Adding the $2.14bn of capitalised lease liabilities to enterprise value while dividing by an EBITDA that is already net of rent charges the reader for the leases twice.

EBITDAR adds the $406 million of operating lease cost back to EBITDA so that the numerator and the denominator treat leases the same way. All three rows use the same September 4, 2026 close and the same clean EBITDA base.

The two consistent constructions agree with each other to within a tenth of a turn — 5.0x and 5.1x. The inconsistent construction reports 6.1x, which is 21.7% higher than either honest answer. A reader who saw only the third row would conclude the stock trades at a market multiple. It does not.

The leases are real obligations and this page does not pretend otherwise. The $2.14bn of total lease liabilities is a contractual claim on future cash and it is carried at full weight in the second construction above and in every leverage calculation in the buyout chapter. The point is not that leases are free. The point is that they must be counted once.

There is a non-recurring refund sitting inside guided earnings

Guided full-year EPS of $9.61 at the midpoint includes $0.86 a share of benefit from tariff refunds recovered after the Supreme Court’s treatment of the IEEPA duties. That is not an operating result and it will not repeat. Stripping it leaves $8.75 of clean guided earnings, which is the figure the valuation table uses and the figure the 11.5x multiple is built on.

Under ASC 450 a gain contingency is not recognised until realisation is assured, so the $95.5 million of filed-but-unrecognised claims is properly absent from both the income statement and guidance. It is an unpriced asset, not an omission.

Both directions, stated honestly. The $0.86 inside guided EPS flatters the reported number and this page removes it. The $95.5 million of filed refund claims that accounting rules forbid the company from recognising sits outside guidance entirely and this page does not add it back either. The first is an overstatement corrected; the second is an option the reader should know exists and that no figure here depends on.

One line does almost all of the damage

Total comparable sales fell 9.0%, or 10.0% in constant dollars. Read as a single number that looks like a brand in broad decline. Segmented, it is not.

Comparable sales for the quarter ended August 2, 2026, as reported on Form 10-Q filed September 3, 2026. Constant-dollar figures are disclosed where the company provides them.

The Americas comparable at −12.0% is the outlier, and the Americas is the segment that produced $7.85bn of FY2025 revenue against $1.75bn from China Mainland and $1.50bn from Rest of World. A 12.0% decline in the segment carrying roughly three-quarters of the business is not the same problem as a broad brand failure, and it does not have the same remedy.

Fiscal year ended February 1, 2026, as filed on Form 10-K March 17, 2026. Segment income is before $1.40bn of unallocated corporate expense, which is why the segment margins above do not reconcile to the consolidated operating margin.

Meanwhile the store base grew. The company operated 825 stores at quarter end against 784 a year earlier, on 3,880 thousand square feet against 3,511 thousand — square footage up 10.5%. Comparable sales falling 9.0% while selling space grows 10.5% is the definition of operating deleverage, and it explains why the earnings guidance cut ran at more than twice the revenue guidance cut.

4.1  Store roll-forward, FY2022 through Q2 FY2026

The net figure above understates the activity underneath it, and the gross figures are harder to obtain than they should be. The 10-K and the 10-Q disclose net store count only. Gross openings and closings appear in Exhibit 99.1 to the earnings Form 8-K — material that is furnished rather than filed, and therefore carries a different liability standard from the financial statements. The annual figures below are Institute sums of the four quarterly disclosures, not a single reported annual number.

Beginning plus opened less closed foots to ending in every row, and each year’s beginning count equals the prior year’s ending count — both tested in the model rather than eyeballed. Of the 65 gross additions in FY2024, 14 were the Mexico stores acquired from the former franchise partner rather than newly opened locations. FY2026 covers the half-year to August 2, 2026 only.

Two things fall out of that table. The company opens roughly 67 stores a year and closes roughly 8 — a closure rate of 1.11% of the base, which is not a chain pruning underperformers but a chain in continuous expansion. And the stores are getting larger: average gross square footage per store has risen from 4,486 to 4,703 across the period.

Put the space against the revenue and the operating-deleverage problem stops being an inference and becomes a measurement. Revenue per gross square foot has fallen from $3,150 in FY2022 to an estimated $2,687 in FY2026, a decline of −14.7%. The company added 12.4% of selling space while the productivity of that space fell by nearly a sixth. The two ends of that comparison are not measured alike — the FY2022 numerator is a reported figure and the FY2026 numerator is management guidance issued September 3, 2026, measured against square footage at August 2, 2026 rather than at fiscal year end. The direction is not in doubt; the second decimal place is.

This is the line to watch, and it is the only one. The multiple is a consequence. The Americas comparable is the cause. A reader who wants to know whether $100.61 was a cheap price should follow that single disclosure quarter by quarter rather than the valuation ratios, which will move mechanically once it turns.

There is nothing here that forces anybody’s hand

Form 10-Q for the quarterly period ended August 2, 2026, filed September 3, 2026, page 3, condensed. The statement foots: total liabilities of $3.69bn plus stockholders’ equity of $4.79bn equals total assets of $8.48bn. Funded debt is a line with nothing on it — no term loan, no bond and no drawn revolver, confirmed at Note 4. The $600 million revolving facility was amended and restated on October 15, 2025 and matures October 15, 2030; availability of $593.7 million is net of $6.3 million of letters of credit issued under it, and a further $1.00bn accordion is uncommitted. The balance sheet is struck at August 2; the share count on the cover of the same filing is struck at August 28, and between those two dates the company spent $118.8 million retiring 1,000,000 shares. The enterprise values on this page net the later cash figure of $1.27bn against the later share count, because pairing an August 2 cash balance with an August 28 share count would credit the company with cash it had already spent. The income tax prepayment is shown for completeness and is not deducted; it is a receivable from the taxing authority rather than a use of cash, and the case takes no position on its recoverability. This table is broken out one level finer than the same statement in the PDF, which condenses to twenty rows so it fits a single page without splitting; both foot to the same totals from the same underlying figures.

Current assets of $3.95bn cover current liabilities of $1.80bn more than twice over. There is no drawn revolver, no term loan, no bond, and no maturity to negotiate. The only long-dated contractual obligation is the lease book, and its weighted average remaining term is 6.59 years at a weighted average discount rate of 4.4%.

That matters for a specific reason. A company with no debt maturity cannot be forced into a decision by a lender. Whatever happens to the Americas comparable over the next several quarters, the outcome will be decided by management and the board rather than by a credit agreement. Very few businesses trading at 5.0x have that property.

The company is buying the stock the buyout could not afford, and borrowing nothing to do it

There is no dividend. No dividend appears on the financing line of any of the three years in the FY2025 Form 10-K, and the company states at Item 5 of that filing that it does not anticipate paying a cash dividend on its common stock in the foreseeable future. There is also no interest and no principal to pay, because there is no funded debt. Every dollar of free cash flow is therefore discretionary, which makes the record below the whole of capital allocation rather than a selected slice of it.

Operating cash flow and capital expenditure from the Consolidated Statements of Cash Flows — Form 10-K for the fiscal year ended February 1, 2026 (filed March 17, 2026), page 47, and Form 10-Q for the quarter ended August 2, 2026 (filed September 3, 2026), page 7. Free cash flow is operating cash flow less purchases of property and equipment. Repurchases are the financing-activity cash line and include commissions and excise tax. The FY2026E row is the Institute’s estimate and not company guidance: operating cash flow is not guided, so free cash flow is built as guided net income plus estimated depreciation less estimated capital expenditure with working capital assumed neutral, and capital expenditure is the first-half figure of $277.1 million doubled and increased ten percent for the seasonal second-half weighting observed historically. Repurchases are not forecast, because the company has committed to no pace.

Over three and a half years the business generated $4,461.7 million of free cash flow and put $4,069 million of it back through the market — 91.2% of everything it earned. In two of those four periods it spent more on its own stock than it produced, funding the difference out of the cash balance, which is why a company holding $1,155 million at the start of the stretch holds $1,389.7 million now. Nothing was borrowed to do any of it.

Consolidated Statements of Stockholders’ Equity — 10-K page 46 and 10-Q page 6 — and the 10-Q cover page for the August 28, 2026 count. Repurchased shares are retired; the company holds no treasury stock. These are common shares only. A constant 5,116,000 exchangeable shares sit alongside them at every one of these dates and are added to the common count wherever this case computes a market capitalisation, so the decline shown here is not an artefact of which class is counted. Repurchases retired 16,511,000 common shares through August 2 while settlement of stock-based compensation net of shares withheld for tax issued 872,000 back, and the difference is the 15,639,000 net reduction to that date.

16,611,000 shares have gone, 13.6% of the company, and the retirement was paid for out of operating cash flow and the cash balance rather than out of a credit agreement. Calling that a leveraged recapitalisation would be wrong, and the distinction is the argument rather than a quibble. A leveraged recapitalisation borrows against the business to retire equity and then services the borrowing out of the same cash flow the equity used to belong to. This company has retired the equity and skipped the borrowing. The $4,069 million spent retiring 16,511,000 shares is an average of $246.44 a share, against a September 4, 2026 close of $100.61. Across the whole programme the company has been a buyer well above the current price — simultaneously an argument that the remaining authorisation will be spent at better prices than the ones already paid, and an argument that management’s judgement of value has not so far been ratified by the market. Both readings are available on the same number.

The equity story here is the buyback, not the buyout, and the arithmetic is straightforward. The company repurchased 4,918,000 shares in the first half at an average of $141.34. After the quarter closed it bought a further 1,000,000 shares at an average of $118.80. At the September 4, 2026 close of $100.61, the $712.5 million of remaining authorisation buys 7,081,801 shares — 6.4% of the company.

The final row is not an extrapolation of the first-half run rate, which the company is under no obligation to sustain. It states only what the already-authorised dollars buy at the current price. Guidance excludes the effect of any further repurchase, so any shares retired from here are accretive to guided per-share figures.

Note the direction of the average prices. The first-half average of $141.34 sits 40.5% above the current price; the post-quarter average of $118.80 sits 18.1% above it. The buyback has so far been executed at prices the market has since rejected, which is a fact about the buyback and not an argument against it — but a reader assessing capital allocation should hold both.

Why this frames the chapter that follows. A repurchase retires stock against a fixed earnings stream out of operating cash flow, with no control premium, no interest expense, and no interest-deductibility problem. The buyout tested in Section 9 does the same thing with a 29.2% premium, $9.83bn of debt, and a tax code that will not let the buyer deduct most of the interest. That comparison is the finding.

Three numbers older than the document they appear in

Three figures in lululemon’s own disclosure are older than the filing they appear in. None of the three is an error by the company, and all three will mislead a reader who does not notice them.

7.1  Lease maturity schedule versus the lease balance

The maturity schedule is the one disclosed in the Form 10-K for the fiscal year ended February 1, 2026. The Form 10-Q carries no lease maturity note, so the schedule above is the most recent one filed — measured seven months before the liability it is being read against.

Between those two dates right-of-use assets rose 19.4% and lease liabilities rose 19.1%. A further $278.5 million of leases have been committed but not yet commenced and therefore appear in neither figure. Reading the February schedule as though it described the August balance understates the near-term commitment by roughly a fifth.

7.2  Filed refund claims unrecognised under ASC 450

The $95.5 million of filed-but-unrecognised IEEPA refund claims discussed in Section 3 is measured at August 2, 2026 and correctly excluded from every recognised figure. It is worth roughly $0.86 a share pre-tax against a $100.61 stock. No number on this page depends on it.

7.3  Beneficial ownership percentage versus share count

Schedule 13D/A amendments filed by the founder. The first three amendments report an identical share count at three different percentages because the denominator was shrinking underneath them — the buyback, not a purchase.

Three filings, one unchanged holding of 9,904,856 shares, and percentages of 8.4%, 8.6% and 8.7%. Nothing was bought. The company retired stock and the same block became a larger share of a smaller company. A reader who saw only the percentages would conclude the founder was accumulating.

What the founder’s position actually is

The founder held 9,576,564 shares, or 8.4% of the company, at the most recent amendment dated August 3, 2026. A cooperation agreement dated May 26, 2026 carries a standstill running 18 months. The chief executive who departed on January 31, 2026 was succeeded by an appointment announced on April 22, 2026, with a start date of September 8, 2026.

No going-private process is under way. Nothing in the Form 10-Q, the Schedule 13D/A chain, or the May 26, 2026 cooperation agreement describes a going-private proposal. The buyout analysis that follows is a constructed test of whether the arithmetic works at a price a board could be asked to accept, not a report of a transaction anyone is pursuing.

8.1  The buyback is quietly enlarging the founder’s position

The percentage the founder last reported and the percentage he now holds are not the same number, and the difference is not a purchase. Amendment No. 24 reported 9,576,564 shares as 8.4% of the company, computed against the share count in force on August 3, 2026. Against the 110,710,025 shares outstanding at the most recent measurement the identical block is 8.65% — a position worth $963.5 million at the close this case is struck at. The holding did not change; the denominator did. The company is retiring stock underneath a holder who is contractually barred from adding to his position, with the arithmetic result that a standstill signed to hold the founder’s influence flat is running alongside a buyback that increases it.

8.2  A reported marital proceeding, and why it belongs in a valuation case

The following is reported, not filed. Press accounts describe a divorce proceeding between the founder and his spouse, commenced in April 2026 in the Supreme Court of British Columbia, and describe it as proceeding without a prenuptial agreement. Nothing in the Schedule 13D/A chain, the proxy statement or the Form 10-Q filed September 3, 2026 discloses any of it, and the Institute has not obtained a court record. Read this section at the standard of a press report, not of a filing.

It is carried here for one reason and no other. A claim on marital property, if it reaches the shares, is a mechanism by which the largest block on the register can move without the holder having formed any view about the company. That is a different kind of supply than a founder selling because he thinks the stock is expensive, and it is not restrained by the standstill, which governs what the holder may acquire and how he may campaign rather than what a court may divide. An 8.65% block is roughly $963.5 million at the close, set against the buyback authorisation sized in Section 6. The Institute has not obtained a trading volume series and therefore states no view on how long such a block would take to absorb.

The Institute draws no further inference. It does not model a sale, does not forecast an outcome in a court whose proceedings it has not read, and does not treat a private matter as a signal about the business. The valuation on this page is unchanged by it. What changes is the completeness of the register a reader is looking at: the largest holding on it is subject to a restriction that expires, and to a reported claim that has not been adjudicated.

Run twice, because the answer depends on which year you believe

On paper this is the specification a sponsor screens for: a category-leading brand, no funded debt to refinance, $1.27bn of cash on the balance sheet, $1.06bn of free cash flow, and an equity that has fallen 80.3% from its high. This section builds the transaction and runs it twice — once on the operating path the market is pricing, and once on the operating path a buyer would underwrite. Those are different paths, they produce opposite answers, and the distance between them is the entire transaction.

The load-bearing question is whether FY2026 is the level or the trough. Everything that follows — the financing, the tax leakage, the exit, the price a board could refuse — is downstream of that one judgment. The Section 163(j) arithmetic below is real and it binds hard, but it is not what decides the deal. The operating case decides the deal, and Section 163(j) decides how badly a wrong operating case is punished.

9.1  Derivation of the entry price

The $130.00 offer used throughout this section is triangulated from three references, each computed independently of the others and of the buyout model itself.

The three references span $128.63 to $133.29 and average $130.90. The test price is 0.7% against that average. They are independent: a transaction multiple observed in the market, a discounted cash flow built from the company’s own operating drivers, and a trading multiple taken from the closest listed peer. That three methods with nothing in common land within $4.67 of each other.

The convention point cuts against the sponsor, not for it. The Skechers premium was struck against a 15-day volume-weighted average price, not against a single close — and the close this case is struck at fell 17.4% in the session. A board negotiating a sale anchors on an undisturbed reference, not on the trough print. The same 30% premium applied to the pre-earnings close of $121.77 implies $158.30 a share, 21.8% above the price tested here. $130.00 is therefore at the generous end for a buyer and the stingy end for a board, which strengthens rather than weakens what follows: the arithmetic below fails at a friendly price, so it fails harder at a realistic one.

9.2  Sources and uses at $130.00 a share

Offer of $130.00 a share is a 29.2% premium to the September 4, 2026 close of $100.61. The operating cash minimum of $400 million is the working balance a retailer of this size cannot sweep. Entry multiples are 6.7x clean EBITDA excluding lease liabilities and 6.4x clean EBITDAR including them.

A sponsor equity cheque of $4.05bn is 27.5% of total uses. That is already an unusually thick equity layer for a buyout, and it is the first sign of the problem: the business will not carry more debt than this, and the reason is the tax code.

9.3  Section 163(j) carryforward and cash tax leakage

The deduction for business interest is limited to 30% of adjusted taxable income. For tax years beginning after December 31, 2024 that income is computed with the depreciation and amortisation addback — an EBITDA base — and the addback is permanent. The stricter EBIT base that governed tax years 2022 through 2024 is still widely quoted and would overstate the cash tax in every year of this hold, so the table below runs on the correct and more generous base. The limitation binds anyway.

Disallowed interest is deferred under Section 163(j)(2), not destroyed — see the paragraphs below the table. It produces no cash benefit inside the hold period because EBITDA never rises far enough to absorb it. Leverage at the end of year one is 5.3 turns of EBITDA and EBITDAR coverage is 1.80x. There is no change-in-working-capital line, and the omission runs in the sponsor’s favour. Operating working capital stands at $744 million, or 7.1% of estimated FY2026 revenue, at August 2, 2026, and a retailer growing revenue funds that ratio out of the same cash this table is using to retire debt. Held at a constant percentage of revenue, the base case would absorb roughly $79 million over the five years and the bull case roughly $263 million. Those adjusted figures are disclosed rather than deducted, because no forward working-capital guidance exists to derive them from — but the direction is not in doubt: the free cash shown here is the generous case, and the returns computed from it are the high end.

This is the whole finding, in one comparison. The structure pays $835.4 million of cash interest, deducts $547.7 million, and then writes a cheque to the Treasury for $207 million of cash tax on a business whose entire EBIT is consumed by interest. What is left to repay $9.83bn of debt is $173.9 million. At that rate the debt is not retired in a hold period; it is barely dented.

That word disallowed has to be read precisely, because the statute does not throw the deduction away. Under Section 163(j)(2) business interest disallowed in one year is carried forward indefinitely and treated as business interest paid or accrued in the succeeding taxable year. It is a timing difference and it books as a deferred tax asset. This page does not claim the deduction is lost, and a reader who has seen the limitation described that way elsewhere should discard the description.

What makes the limitation bind here is that the carryforward has nowhere to go. A carryforward is only worth something in a later year with spare capacity under the thirty percent test, and this structure has none: interest exceeds thirty percent of EBITDA in every one of the five hold years, so the balance accumulates instead of unwinding. It reaches $1.24bn by exit, a deferred tax asset of roughly $370.7 million at the assumed rate. Meanwhile the cash leaves on schedule — $207 million in year one and $1.07bn across the hold. A sponsor return is computed on cash, and a deduction that becomes usable after the exit does not service debt before it.

The model credits nothing for that asset at exit, which understates value and is stated here rather than hidden. Two things bound how much. An ownership change at the exit brings the carryforward within Section 382, which caps annual use at a rate applied to the equity value of the loss corporation; and a buyer will only pay for a shield it can absorb, while any buyer running comparable leverage inherits the same capacity problem that created the balance. The conservatism is not doing the work in the conclusion either. Handing the sponsor the entire $370.7 million at exit, with no Section 382 haircut and no discount for the buyer’s own capacity, lifts the base case from 2.4% to 4.0%. The hurdle is twenty percent.

9.4  Returns on the Institute’s operating cases

Base case. Comparable sales −7% in year one improving to 1% by year five, square footage 7% slowing to 3%, margin 17.2% against 18.86% today. The EBITDA path that produces is −7.1% in year one and $1.98bn at exit. The decline stops; nothing recovers.

Five-year hold, exit at the entry multiple, 8.5% blended cost of debt, 30% cash tax rate, Section 163(j) applied each year. A dash marks a cell where debt and balance-sheet cash exceed total uses, which is an arithmetic artefact rather than a deal.

Bear case. Comparable sales −9% in year one improving to 0% by year five, square footage 8% slowing to 2%, margin 14.8% against 18.86% today. The EBITDA path that produces is −11.4% in year one and $1.54bn at exit. The decline continues and margin follows it down.

Same structure and same exit convention as the base case.

Bull case. Comparable sales −3% in year one improving to 3% by year five, square footage 6% slowing to 4%, margin 19.8% against 18.86% today. The EBITDA path that produces is 1.4% in year one and $2.79bn at exit. The Americas comparable turns and margin expands past where it stands today.

Same structure and same exit convention as the base case.

The base case never earns a sponsor return. The best cell in the whole grid returns 7.1% — at $110.00 a share and 4.0 turns, which is the lowest price and the least debt on offer — against a hurdle rate no sponsor sets below twenty percent. Read across the base-case rows and note the direction: at $130.00 a share the return goes from 4.5% at four turns to 2.4% at five turns to −4.6% at six. Raising leverage makes the return worse. That is the diagnostic. In a structure where debt is working, more of it amplifies the equity return; here each additional turn adds interest that Section 163(j) will not let the buyer deduct, so leverage is destroying value rather than creating it. The constraint is the earnings, not the capital structure.

Of the three, only the bull case clears a twenty percent hurdle, and it clears by assuming the problem away. The highest price that still returns twenty percent is $130.72 a share at four turns and $147.81 at five — premiums of 29.9% and 46.9% to the September 4, 2026 close. Those are real premiums, and that is the uncomfortable part of the arithmetic rather than the reassuring part: $130.72 lands within $2.09 of the midpoint of the Institute’s own value range below. Both numbers are pricing the same recovery from opposite directions. But the bull case gets there by assuming the Americas comparable turns positive inside eighteen months and that EBITDA margin expands past the 18.86% the business earns today. A sponsor underwriting that is paying full value for a turnaround that has already happened and carrying two years of operating risk for nothing. In every base and bear configuration the deal fails even at the current market price, before any premium at all.

That result is real, and it is also entirely a function of the operating case behind it. A path that never recovers cannot produce a buyout return no matter how it is financed. The question the three grids above do not answer is whether any of those paths is the right one to underwrite.

9.5  The operating case a buyer would underwrite

Every path above is a public-market path. Each extrapolates forward from the quarter ended August 2, 2026 and therefore embeds an assumption a control buyer does not share: that FY2026 describes the level of the business rather than a bad year in it. That assumption is the correct default for a minority holder, who has no ability to change anything. It is the wrong default for a buyer paying a control premium, because paying one is a statement that the current earnings are not the right earnings.

The company’s own record is the argument. FY2026 estimated clean EBITDA of $1,966 million is a margin of 18.9%. The three fiscal years before it ran 26.1%, 27.9% and 24.4%. A single year that sits five and a half points below the weakest of the three preceding it, in a year carrying tariff cost and elevated markdowns, is the profile of a trough rather than a plateau — and the board acted as though it agreed. The chief executive stepped down January 31, 2026 and a successor was named April 22, 2026, starting September 8, 2026. Boards do not replace a chief executive over a normal year. The change of seat is not proof that FY2026 is abnormal, but it is the directors’ own revealed judgment on the question, and it points one way.

Actual EBITDA is reported operating income plus depreciation and amortisation for each fiscal year. The right-hand column is the single number that decides the transaction: what each path recovers against the last full year the company actually reported. The Institute’s Base case reaches 73% of it. The Sponsor case reaches 105%.

The Sponsor case is built on the same three drivers as the others, so it can be argued with on the same terms. Comparable sales trough at −5% in year one, reach zero in year two and hold at +2% thereafter. Square-footage growth decelerates from +7% to +3% as the fleet stops being the growth engine. EBITDA margin climbs from 19.5% to 22.0% over the five years. That exit margin is deliberately short of the 24.4% the company earned in FY2025 and far short of the 27.9% it earned in FY2024. The Sponsor case does not assume the business returns to its best year. It assumes it recovers most of the way back to an ordinary one.

Sponsor case. The same grid, the same financing, the same five-year hold and the same exit-at-entry convention. Only the operating case has changed.

A dash marks a cell where debt and balance-sheet cash exceed total uses, which is an arithmetic artefact rather than a deal.

On its own underwriting the deal works, and comfortably. At $130.00 a share and 5.0x leverage the sponsor case returns 25.9% and a multiple of money of 3.16 times. Every cell in the grid that is financeable at all clears twenty percent except one, and the highest price that still clears at four turns is $134.70 — a 34% premium to the September 4, 2026 close, well above any of the three reference prices Section 9.1 triangulated from. That is what it means to say a sponsor would compete for this asset: the arithmetic supports a bid materially above where a board would be able to refuse one.

Two features of that result are worth separating out, because they are commonly conflated. The first is that the leverage now helps. On the Institute’s base case the return fell as debt rose; on the sponsor case it rises, from 20.9% at four turns to 36.7% at six. That reversal is the signature of an operating case that grows into its capital structure rather than being crushed by it, and it is the reason the same grid produces opposite answers. The second is that the Sponsor case and the Bull case arrive at almost the same exit EBITDA — $2,838 million against $2,794 million — by opposite routes. The Bull case gets there on revenue, requiring $14,111 million of it against the Sponsor case’s $12,898 million. The Sponsor case gets there on margin. Only one of those two routes is something a new owner can act on directly.

The Institute does not adopt the Sponsor case, and it is not the case behind the valuation in Section 11. It is printed because a buyout analysis run only on paths a buyer would reject is not a test of the buyout — it is the valuation restated in different units. The honest conclusion is narrower than either grid on its own: this is not a deal that fails on financing or on price. It is a deal that resolves entirely on whether FY2026 is the trough. A buyer who believes it is has a large return available at a large premium. A buyer who believes it is not has no price at all.

9.6  How a sponsor would try to make the case true

The Sponsor case above is an assumption, not a plan. What converts one into the other is a post-close operating playbook, and those playbooks are not improvised — the levers are standard across the industry and are pulled in a standard order. The table below lists them and states what each would mean at this specific company. None of them is modelled. The grid does not credit a single one of these actions; it simply assumes an outcome that some combination of them would have to produce.

Chapter references are to the Baratelli Institute’s Practitioner’s Guide to Private Equity, which treats the post-close playbook across Part V and Chapter 41, and interest deductibility at the portfolio-company level in Chapter 44. This case is a worked example of the arithmetic that guide sets out; the guide is where the playbook itself lives.

The lever the model cannot price. Revenue in every path above compounds as the prior year grown by comparable sales and by square footage. That identity says “the same products, in more stores.” It has no term for a new category, which means every case in this section — including the Sponsor case — values brand extension at exactly zero. That is a conservative construction and it is also a real omission, because brand extension is one of the two or three things sponsors most often underwrite in a consumer asset. Yeti is the reference: a company built on a cooler that now sells drinkware as its largest category, alongside bags, cargo and a newly launched fitness line. The counter-evidence has to be carried alongside it. Yeti’s drinkware category declined in its most recent fiscal year, and Under Armour’s extension into footwear did not restore its multiple. Extension is a lever that has worked and a lever that has failed, and a reader should treat the model’s zero as a floor rather than as an estimate.

9.7  Exit, founder rollover, and transaction size

Exit is struck at the entry multiple, which is the neutral convention — it assumes no multiple expansion and no compression. Assuming expansion would be assuming the answer.

The founder rollover does not rescue it, and the reason is worth stating. Running the same structure with the founder’s entire 9,576,564-share position rolled into the new entity produces an identical IRR in every cell. A rollover reduces the sponsor’s cheque and the sponsor’s share of the exit proceeds in exactly the same proportion, so the return per dollar is unchanged. A rollover solves a fundraising problem. It does not solve a returns problem, and the problem here is returns.

For scale, the largest branded-apparel take-private on record is 3G Capital's acquisition of Skechers, announced May 5, 2025 at approximately $9.4bn including a founding-family rollover. The base-case uses above come to $14.75bn — roughly 1.6 times the size of the largest comparable transaction ever done in this category. The financing market for a deal that size does not exist for a business whose interest is 34% non-deductible.

Once the leases are handled the same way on both sides

Section 9 concluded that the buyout resolves on a single operating question rather than on price or on financing, and that the answer flips depending on whether FY2026 is read as the level or as the trough. That leaves the question of what the business is worth on the Institute’s own reading, and the first of the two answers is the one the market itself supplies: what comparable companies trade for.

Before any multiple is printed, one convention has to be settled — and getting it wrong would overstate this case by roughly a factor of two. Every peer enterprise value below is lease-inclusive; the data provider adds the operating lease liability to net debt. Deckers proves it arithmetically: market capitalisation of $11.69bn, less $1.60bn of cash, plus $472 million of reported total debt against zero funded borrowings, produces the $10.56bn enterprise value shown. That reported debt is the lease. Meanwhile every EBITDA in the denominators is after rent, because generally accepted accounting principles put operating lease cost in operating expense. So the set pairs a lease-inclusive numerator with an after-rent denominator, and LULU’s headline 5.0x — computed excluding leases — is not the number to set against Deckers. On the peer set’s own convention this company trades at 6.11x, and the discount to its closest structural analogue is 23.2%, not the 36.9% the mismatched pair would imply. The Institute prints the smaller figure.

Peer data from StockAnalysis.com / S&P Global Market Intelligence, measured September 4, 2026 except where the table says otherwise. Enterprise values are lease-inclusive on every line, including lululemon’s, so the column is internally consistent. lululemon’s EBITDA is the FY2026 estimate with the one-time tariff refund removed, as computed in Section 3; the peers are last twelve months as reported.

10.1  Limitations of the peer set

Columbia Sportswear runs on a different clock, and its denominator is contaminated in exactly the way this case corrects for in lululemon. The price is a September 3 close, one day stale against the rest of the set. More consequentially, its last-twelve-months EBITDA of $342 million includes roughly $78 million of non-recurring tariff refunds — 23% of the denominator. Removing them moves the multiple from 8.2x to 10.6x. The reported figure understates by more than two turns. Section 3 performs the identical adjustment to lululemon’s own guided earnings; it would be indefensible to make it there and not here.

Under Armour’s price is an intraday quote, not a close. It was measured at 1:54pm on September 4, 2026, while the market was open, and its market capitalisation, enterprise value and both multiples inherit that clock. It is left in the table because removing an inconvenient name is worse than labelling it, but nothing in this case is argued from it: a multiple of 18.6x on a 3.5% EBITDA margin is not a valuation signal, it is a company whose earnings have nearly disappeared.

On Holding’s multiple is not a single number. Published estimates range from 12.79x to 15.8x. That spread is a definitional disagreement about lease treatment, share-based compensation and international-versus-domestic accounting standards — not a rounding difference — and the figure in the table is one provider’s, not a consensus.

Five names are absent and the reasons differ. VF Corp and Gap were dropped on judgment: a levered multi-brand wholesale turnaround and a mass-market value retailer share neither the direct-to-consumer model nor the debt-free balance sheet. Abercrombie & Fitch, Ralph Lauren and Amer Sports were dropped for cause — no price and multiple measured on the same day could be obtained for any of them, and in one case a September 4 share price arrived beside a market capitalisation implying roughly half that price. They are named here rather than quietly omitted, because a comparables table is as much about which names were excluded as which were kept.

10.2  Forward earnings: the offsetting evidence

On enterprise value lululemon is cheaper than Deckers. On forward earnings it is not cheaper at all. Deckers is the tightest structural analogue in the set — direct-to-consumer led, net cash, no funded debt — and it earns a 24.0% EBITDA margin against lululemon’s 18.9%. On enterprise value to EBITDA lululemon is cheaper by 23.2%. On forward earnings Deckers trades at 11.24x and lululemon at 11.50x: lululemon is the more expensive of the two, by 0.26 of a turn. That is printed before the valuation rather than after it because it is the strongest available objection to this case and a reader is entitled to it first. The reconciliation is that the two metrics treat the balance sheet differently — enterprise value to EBITDA credits lululemon for $1.27bn of net cash and charges it for the lease liability, while the earnings multiple does neither. Anyone arguing this stock is cheap has to argue it on enterprise value, free cash flow and the balance sheet, and has to concede the earnings multiple. The Institute concedes it.

On the precedent, one number is printed and one is refused. The acquisition of Skechers by 3G Capital, announced May 5, 2025 at $63.00 a share, was described by the parties as approximately a 30% premium to the 15-day volume-weighted average price. That is the issuer’s own reference and it is defensible. A premium to the unaffected closing price is not printed here: the widely repeated figure could not be tied to an exchange record or to the merger proxy, and a premium computed from an unverified base is precisely the kind of number this case refuses to carry.

And what the Institute thinks the shares are worth

The comparables say what the market pays for similar businesses. A discounted cash flow says what this one produces. The two are built from entirely different inputs and the case is stronger where they agree, which is the only reason to run both.

11.1  Cost of capital

The company has no funded debt. That is not a detail here — it means the weighted average cost of capital collapses to the cost of equity, because there is nothing to weight it against. One entire source of dispute disappears and the argument lands where it belongs, on beta. And beta is where the honest answer is a range: published estimates for this company run from 0.76 to 1.01 depending on the window, the observation frequency and the reference index. That is a spread of roughly 33% on the single most contested input, and it moves the cost of equity by 110 basis points, from 8.14% to 9.24%. A point estimate would be false precision dressed as rigour, so both ends are carried through to a value.

One input is stale and it is labelled rather than buried. The equity risk premium is the implied United States figure measured July 1, 2026, on a series republished monthly — two months old at the publication date of this case. It is the best available and the reader should know its age. The terminal growth band is capped below breakeven inflation of 2.35%: a mature specialty retailer growing faster than the price level in perpetuity is a claim nobody should make in print.

11.2  Unlevered free cash flow projection

Unlevered free cash flow is EBITDA, less cash tax on operating income, less capital expenditure. There is no interest line, and that is the point of an unlevered figure: it is indifferent to how the business is financed, which matters here because Section 9 has already shown that financing it changes the answer. The EBITDA path is the base case derived in Section 9 from comparable sales, square-footage growth and margin — not a separate set of assumptions invented for this section.

Discounted at the mid-point cost of equity of 8.69%. Capital expenditure is held flat at the FY2026 estimate of $610 million, which is undemanding for a chain still opening roughly 67 stores a year. There is no change-in-working-capital line, on the same basis and with the same direction as the buyout model in Section 9: operating working capital stands at $744 million, or 7.1% of estimated FY2026 revenue, at August 2, 2026, and a business growing revenue funds that ratio out of the cash shown here. Held at a constant percentage of revenue it would absorb roughly $79 million across these five years — sized rather than deducted, because no forward working-capital guidance exists to derive it from. Treat the cash flows above as the generous case, not the conservative one.

11.3  Terminal value: perpetuity growth and exit multiple

Terminal value is 74% of enterprise value in this model. That is high and it is normal for a five-year explicit period, but it means most of the valuation is an argument about year six onward rather than about the next five years. It is also why the terminal is computed twice.

Enterprise value is the sum of the two present-value columns, and the row foots across the page: on the perpetuity-growth terminal, $9,605 million of discounted terminal value plus $3,384 million of discounted forecast-period cash flow is $12,989 million. The forecast-period figure is the same $3,384 million that closes the table above; it is repeated here so the addition is visible on one line rather than split across two exhibits. Both rows use the same five years of cash flow and the same 8.69% discount rate; only the terminal method differs. Equity value adds back cash of $1.27bn, net of the post-quarter repurchase, and there is no funded debt to deduct. Per share divides by 110.7 million shares. The exit multiple is what Deckers — the closest structural analogue in Section 10 — trades at today, and it is used precisely because a perpetuity-growth terminal is so sensitive to two inputs nobody can observe. That the two methods land within $7.17 of each other is the useful result; had they diverged, neither would be worth printing.

11.4  Sensitivity: cost of equity versus terminal growth

Equity value per share, perpetuity-growth terminal, base-case cash flows. The rows are the cost of equity at the low beta, the mid-point and the high beta from the input table above. All 9 cells sit above the September 4, 2026 close of $100.61, with the lowest at $114.39 and the highest at $148.61. That is the honest headline of this section and also its limitation: a grid in which every cell agrees is a grid whose bounds may be too narrow, which is why the bear case below is run outside them.

11.5  Bear case, run outside the grid

The grid above varies the discount rate and the terminal growth while holding the cash flows fixed. The larger risk is not that the discount rate is wrong; it is that the base-case operating path is wrong. Running the same discounted cash flow on the bear path from Section 9 — comparable sales deteriorating for another year, margin giving up most of the operating leverage the growth years built, EBITDA ending the period at $1,538 million against $1,966 million today — produces $94.09 a share. That is 6.5% below the current price, and it is the floor the Institute of the range rather than a figure sitting beneath it.

11.6  The same discounted cash flow, run a second time off FY2025

Everything above compounds from the FY2026 estimate. That is a choice, and it is the choice the whole case turns on. The year ended February 1, 2026 produced EBITDA of $2,707 million on revenue of $11,102.6 million, a margin of 24.4%. The FY2026 guidance midpoint is $1,966 million on $10,425 million, a margin of 18.9%. The gap is $741 million. A buyer of the stock and a sponsor underwriting a buyout have to answer the same question before either can price anything: is FY2026 the level, or is it a bad year measured against a level this business has already demonstrated? The Institute does not think that question can be settled from the outside, so it is answered twice.

What has to be believed about that gap depends on what is in it, and it decomposes exactly, with no residual:

The identity is EBITDA_FY2025 less EBITDA_FY2026E equals revenue_FY2026E times the change in margin, plus the FY2025 margin times the change in revenue. It is exact by construction and carries no plug.

More than three quarters of the decline is margin, not demand. That distinction is the difference between a business with fewer customers and a business with the same customers and worse economics, and it is not a rhetorical difference: a margin point is precisely what a control owner buys the right to attack, and a departed customer is not. It is the arithmetic underneath Section 9’s otherwise strange result that the same buyout fails on the Institute’s operating case and clears comfortably on a buyer’s. Within the margin term, the tariff cost management has quantified but not yet absorbed is $95.5 million, or 0.92 points of guided revenue — 16.6% of the 5.52 points of compression. The remainder is not attributed here to any single cause, because the filings do not support attributing it.

Run the model a second time and hold FY2025 as the level. The schedule below starts from reported FY2025 EBITDA of $2,707 million and grows it at the terminal rate of 2.0% — no recovery, no operating leverage, no store-growth arithmetic, nothing but inflation. It is deliberately the least imaginative version of the recovery case: it does not ask the business to earn back anything it did not already earn in the year just closed. Two conventions cut against the answer and are used anyway. Depreciation is FY2025’s own $496 million rather than the FY2026 estimate of $588 million, which raises taxable operating income and therefore cash tax; capital expenditure is FY2025’s own $681 million rather than the FY2026 estimate of $610 million, which takes more cash out. A schedule anchored on a year is handed that year’s own reinvestment, or the run rate and the reinvestment rate are being measured on two different clocks.

Same cost of equity, same tax rate, same terminal methods and the same five-year structure as the base case above. The only change is the starting EBITDA level and the two reinvestment inputs that travel with it. Year-five EBITDA of $2,989 million is 110% of FY2025 — the excess over 100% is compounding at the terminal rate, not recovery.

The two terminal methods agree to $0.12 a share here, against $7.17 in the base case. That is not a virtue of the model — it is a consequence of a flat schedule, in which the exit multiple and the perpetuity are applied to almost the same terminal cash flow. At the low and high ends of the cost-of-equity range, the perpetuity-growth read runs from $190.41 to $222.74.

On the FY2025 run rate the business is worth $205.24 a share. That is 104% above the close, and it is not a forecast. It is the price of a single proposition: that a company which earned $2,707 million of EBITDA in the year ended February 1, 2026 will, at some point, earn approximately that again. Nothing in the model asks it to earn more.

11.7  The question asked backwards

Two valuations that differ by more than $76 a share are not a range; they are a disagreement about a fact. The way to arbitrate it is to stop valuing the company and ask instead what the market is already paying for. Hold the price fixed at the $100.61 close, hold every other input in this section unchanged, and solve for the one number that makes the model return exactly that price — the permanent EBITDA level the marginal buyer is underwriting.

Solved by bisection on a flat EBITDA level growing at the terminal rate, at the mid-point cost of equity, with FY2025 depreciation and capital expenditure — the identical machinery as the run-rate table above, with the price held and the level released instead of the reverse. On an exit-multiple terminal the implied level is $1,375 million; the higher of the two is used in the text because it is the more forgiving reading of what the market believes.

At $100.61 the market is not paying for a bad year. It is paying for a permanent level 17.5% below the FY2026 figure the company itself has guided to for the year already in progress — a level that sits between the bear terminal and the base terminal of Section 9, and implies a margin of 15.55% the business has not printed and management has not guided to. That is a defensible thing to believe. It is not the same thing as believing FY2026 is a difficult year, and a reader who holds the milder view is not holding the view the price expresses.

This run-rate read is not folded into the stated range below, and the omission is deliberate. The published bracket stays anchored on FY2026, because FY2026 is the only year for which the company has given guidance and the Institute has no defensible basis for asserting when, or whether, the FY2025 margin returns. Averaging a $205.24 read into a $128.63 bracket would produce a midpoint nobody holds and would bury the disagreement, which is the finding. The two are set side by side at the top of this page instead, with the reverse-DCF level beneath them, so the reader can see which proposition each number is the price of.

11.8  Valuation summary and stated range

Four independent reads of the same business — two from the discounted cash flow and two from the comparables. What they form is a bracket rather than a point estimate, because a point estimate would imply a precision the inputs cannot carry.

Discounted cash flow reads are at the mid-point cost of equity of 8.69% and terminal growth of 2.0%. Comparable reads convert an enterprise value at the stated multiple back to equity per share by removing the lease liability of $2.14bn and adding cash of $1.27bn.

The Institute’s position, stated plainly. lululemon athletica is worth between $116.43 and $135.98 a share, with a midpoint of $128.63, against a September 4, 2026 close of $100.61. That is 16% to 35% of upside, and it is a statement about value rather than about price: nothing in this case predicts when or whether the market will agree, and the bear path at $94.09 is a live possibility rather than a rhetorical one.

The peer median is not used, and the reason matters. The median multiple of the peer set is 12.01x, which applied to this company implies $205.38 a share — more than double the current price. It is arithmetically available and it is not defensible: the median is Nike, a business carrying net debt on a 10.6% EBITDA margin against lululemon’s 18.9%. The upper bound of the range above is the discounted cash flow on an exit-multiple terminal instead, and the lower bound is a discount to Deckers — 7.00x against 7.95x — taken because lululemon’s margin is thinner and its comparable sales are falling while Deckers’ are not.

11.9  Reconciliation with the leveraged buyout analysis

Section 9 ran the buyout twice. On the Institute’s own operating case no combination of offer price and leverage returns a sponsor’s cost of capital; this section concludes the equity is worth materially more than it trades for. Those look like opposing findings and they are the same one. A sponsor offering $130.00 would be paying 1.1% above the midpoint of the Institute’s own estimate of intrinsic value — inside the range, with no discount left to fund a return. A company can be simultaneously undervalued by the public market and unbuyable by a financial sponsor, and for the same reason: the gap between price and value is real but not wide enough to pay a control premium, a 2.5% fee load and 8.5% interest on the debt required to close it.

The sponsor case in Section 9 breaks that reconciliation, and it is worth being explicit about how. A buyer who treats FY2026 as a trough is not valuing the same earnings stream this section values. It is buying a business it expects to earn 105% of FY2025 EBITDA by year five rather than 73%, and against that stream $130.00 is not a full price. The disagreement is not about the discount rate, the multiple, or the financing. It is about which year is representative.

11.10  A reported third-party position

Michael Burry is widely reported as long lululemon and bullish on it. Both halves of that sentence need qualifying, and the qualifications are the reason this passage runs to several paragraphs rather than one line.

What the filings show is a position being cut, not built. Scion Asset Management’s Form 13F for the second quarter of 2025 disclosed 50,000 shares alongside a 400,000-share call position — 450,000 share equivalents in total. By the third-quarter filing the calls had gone and the disclosed position was 100,000 share equivalents. That is a reduction of 77.8%. Any account describing the position as having been doubled or added to across that period is contradicted by the filings themselves.

And there is nothing newer, because the filing obligation ended. The third-quarter 2025 report, filed November 3, 2025 (for the quarter ended September 30, 2025), is the last one. The investment adviser’s registration was terminated November 10, 2025 and no 2026 Form 13F exists. The disclosed evidence chain stops there, and it stops with the position smaller than it had been.

The bullish 2026 commentary is therefore not a filing. It comes from a subscription newsletter post dated September 3, 2026, reported second-hand by four outlets and not verified against the primary text, and it describes a personal portfolio, not Scion fund capital. Those are different pockets of money carrying different disclosure obligations and they must not be conflated. The Institute has not verified the primary text.

The honest construction, stated once. A well-known investor is reported to hold a personal position at a materially higher cost basis than the current price, on evidence that is a newsletter rather than a filing, after his fund’s disclosed position had already been cut by more than three quarters and his filing obligation had ended. It is not corroboration of anything in this case. The value range above is derived from the cash flows and the comparable companies; it would read the same if no well-known investor had ever mentioned the name, and a reader who finds the range persuasive because of who else is reported to hold the stock has been persuaded by the wrong thing.

The conditions, stated so they can be checked

The Institute does not forecast the Americas comparable and does not claim to know when it turns. What can be stated precisely is what the current price requires, and each of the following is observable in a subsequent filing rather than a matter of opinion. The first item is also the buyout condition: it is the line that decides whether FY2026 was the trough, and therefore which of the two grids in Section 9 a buyer would have been right to run.

1. The Americas comparable, at −12.0% in the quarter ended August 2, 2026, stops deteriorating. This is the single load-bearing condition and everything else is secondary to it.
2. Square footage growth of 10.5% year over year slows toward the rate of comparable-sales recovery, so that operating deleverage stops compounding. The company opened 41 net stores in a year in which comparables fell 9.0%.
3. The implied fourth quarter of $3.23bn of revenue and $4.06 a share is delivered. It is the largest single component of full-year guidance and the company has not entered it.
4. The remaining $712.5 million of repurchase authorisation is actually deployed. At the current price it retires 6.4% of the company, and guidance excludes it.
5. The incoming chief executive, starting September 8, 2026, does not reset expectations further. A new arrival has both the opportunity and the incentive to rebase, and the 18-month standstill in the May 26, 2026 cooperation agreement means the founder cannot force an alternative in the meantime.
6. China Mainland, comping −2.0% reported and −8.0% in constant dollars against $1.75bn revenue in FY2025, continues to hold. It is the segment that is not currently the problem, and the thesis does not survive it becoming one.

Stated, with its limits stated too

The balance sheet is not the risk. A company with no funded debt, $1.27bn of net cash, current assets covering current liabilities more than twice, and an undrawn revolver cannot be forced into a decision by a creditor. Every outcome from here is a management and board decision rather than a credit event, and that is a materially different situation from most equities trading at 5.0x.

On the buyout question the arithmetic does not settle the argument. It locates it. On the Institute’s own operating case the deal fails and fails hard: Section 163(j) disallows 34% of the interest at five turns, the structure pays $207 million of cash tax while its entire EBIT is consumed by interest, $173.9 million of free cash flow will not deleverage $9.83bn of debt, the best cell in the grid returns 7.1%, adding leverage makes it worse, and a full founder rollover changes nothing. On the case a control buyer would underwrite — FY2026 as a trough, margin recovering to 22.0% against the 24.4% the company earned in FY2025 — the same structure returns 25.9% at $130.00 a share and five turns, and clears a twenty percent hurdle up to $134.70. Same financing, same tax code, same exit convention. Only the operating case differs.

So a leveraged buyout of lululemon is not a financing question and it is not a price question. It is a bet on whether the year now closing was the trough, and a reader who wants to know whether a sponsor would compete for this asset has to answer that first. What the Institute will say is the narrower thing: on its own reading of the operating path, the company is already executing the part of the buyout thesis that requires no assumption at all — retiring shares against a fixed earnings stream, out of operating cash flow, with no premium paid, no debt raised and no interest to deduct.

What the Institute will not claim is a view on when the Americas comparable turns, or whether $100.61 proves to have been a low. Neither is knowable from a filing, and asserting either would be the kind of unearned conviction this house does not publish. The eight-year-low framing circulating in secondary coverage is also reported here as attributed rather than established — the Institute has not verified it against primary price records. The figure that is verifiable, and that carries the point without needing the other, is the 80.3% drawdown from the December 2023 high of $511.29.

The conviction, and its boundary. The balance sheet is not the risk. The Americas comparable is. A reader who wants to know whether this was cheap should watch that single line rather than the multiple. The multiple is a consequence. The comparable is the cause.