US auto retail, eight ways

What you are actually comparing

A franchised dealer sells new cars at a thin margin in order to earn the service work and the finance and insurance attached to them. Parts and service runs from 41% to 48% of gross profit at these six, at a gross margin in the high forties, and it is the counter-cyclical part — a car sold today is warranty and maintenance revenue for the next six years, performed by the franchise that sold it and nobody else.

Neither CarMax nor Carvana has a meaningful service business. Both replace it with finance: CarMax by holding a consumer loan book of about $16 B on its own balance sheet, Carvana by originating loans and selling them for an up-front gain. That makes their gross profit both higher-margin and more procyclical than a dealer’s, and it puts CVNA at 41% of gross profit from finance against 33% at the most finance-dependent of the dealers, Sonic Automotive.

Two things a screen counts as debt, and shouldn’t

Floor plan is not debt. Floor-plan notes payable finance new-vehicle inventory, are secured on that inventory and liquidate as the cars are sold; they rise and fall with the stock on the lot rather than with anything a lender could call. A standardised screen counts them in enterprise value anyway, which puts a dealer a billion dollars or more above the net debt its own filings report. They are excluded here, and — this is the part that is usually missed — the EBITDA they would be measured against is struck before floor-plan interest, so the numerator and the denominator agree. Get only one of the two right and the multiple moves by about a turn.

Three of the eight consolidate a consumer lender. Lithia Motors, AutoNation and CarMax carry $2.8 B, $2.6 B and $15.7 B of non-recourse securitisation debt. It is unarguably on the balance sheet, but each captive’s segment income already arrives net of what that borrowing costs, so counting the debt again in enterprise value counts the same funding twice. At CarMax the difference is not marginal: it is 2.1× net debt to EBITDA against 17.3×, which is the difference between a retailer and a finance company. Excluded from the headline figures, and shown separately in the leverage section.

Inside the band, these are not the same business

Penske Automotive is being taken private by its controlling shareholder, so its 11.5× is a transaction price rather than a market view and is held out of what follows. The five that remain sit within 1.2 of a turn of each other on EBITDA. Their returns on capital, their leverage and their forecast growth are nowhere near that close. The price is not separating these businesses; what follows is what does.

The two ends of that are worth naming. Group 1 Automotive is the cheapest of the five at 6.3×, and it also carries the least debt at 3.2× EBITDA and earns 7.8% on invested capital — second only to AutoNation. Its discount is not about returns or the balance sheet. It is the most geographically concentrated name in the set and the most exposed to the UK agency transition, and those are the risks being priced.

Lithia Motors is the other end: the dearest of the five at 7.5× while earning the lowest return on capital at 5.7% and carrying the most debt at 4.0× EBITDA. Outside the franchised group, CarMax is on 27.0× trailing earnings with the thinnest EBITDA margin of the eight at 3.5% and no parts annuity at all — the dearest earnings multiple in the set attached to the weakest unit economics in it.

The two multiples also rank the group differently. Sonic Automotive is 6.4× on enterprise value and 9.7× on earnings, moving 3 places dearer between the two measures — more than any of the five — because its interest cover is the thinnest at 3.3×. More of its EBITDA is spoken for before it reaches the equity. Which multiple a screen quotes decides where it ranks.

The franchise is the moat and the leash

A franchise is both a moat and a leash: protected territory and warranty work nobody else may perform, in exchange for facility mandates, inventory allocation and the manufacturer’s right to change the distribution model underneath you. Which badges a dealer holds, and where it holds them, therefore decides most of what the multiples are arguing about.

Concentration varies far more than the franchise counts suggest, and the detail by company is below. The exposure they share is where the badges are held: Group 1 Automotive, Lithia Motors and Penske Automotive have revenue outside the United States, which is where the agency model is already live.

Which badges, and where they are held

What the cycle did, and what each company did

Three years says one thing; seven says another

EBITDA fell at all six franchised dealers over three years — the mildest by 3.0% a year, the steepest by 11% — and compounded at 8.6% to 20% a year over seven. Both are true of the same companies.

Lithia Motors and Asbury Automotive lead revenue over seven years, at 18% and 15% a year, and they are also the only two whose share count is higher than it was six years ago. That is one fact, not two. Asbury Automotive spent 92% of its present equity value on acquisitions in three years and Lithia Motors 51%: dealerships bought partly with stock arrive as revenue and as shares at the same time, which is why the revenue columns flatter them and the earnings columns beside them do not.

AutoNation is the mirror image, and the clearest argument on the page for reading per share. It has the slowest revenue growth of the six over seven years at 3.7% and the fastest earnings per share at 22%, because the count underneath it fell 12% a year. CarMax is the one that went backwards on both — -4.9% on EBITDA and -14% on earnings per share — and it is also the one that bought nothing.

The forecast columns say the same thing about the next two years: consensus has earnings per share growing faster than revenue at every one of the eight. That is a forecast about margin and share count rather than about selling more cars, and it is the same machinery the seven-year record was built on.

Carvana has no trailing EBITDA or earnings rate at all: it was loss-making for most of the window, and a compound rate off a negative base is a number without a meaning.

Revenue mix says nothing; gross profit mix says everything

At the franchised six, new vehicles are about half of revenue and a fifth or less of gross profit. Parts and service is a tenth to a sixth of revenue and more than two-fifths of gross profit. Finance and insurance is single-digit revenue and a quarter to a third of gross profit, because it is booked net of cost. Anybody ranking these businesses on revenue mix is reading the least informative table in the filing.

What they did with the cash

Acquisitions, capital expenditure, buybacks and dividends over three years, each as a share of the four combined, against the policy each company publishes. The two barely agree; the stated policies are below the chart.

Six of the eight have fewer shares outstanding than in 2018. One has more: Carvana raised about $2.3 B through an at-the-market programme over the three years while every other name was retiring stock.

Stated capital allocation, and what followed

GPI Group 1 Automotive Return-based

One of four stated strategic pillars: acquisitions, capex, buybacks, dividends, debt reduction, real estate and organic growth evaluated through a single return-based framework, with active disposal of underperforming or non-strategic stores to recycle capital. It leans on low rent-adjusted leverage as the thing that lets it buy and repurchase at the same time — and it does both.

ABG Asbury Automotive Leverage-anchored

A target transaction-adjusted net leverage ratio of 2.5× to 3.5× in a normal environment, with allocation decided inside that band; it ended FY2025 at 3.2×, up from 2.9×, after Herb Chambers. The stated growth levers are same-store growth, acquired revenue and opportunistic repurchases, in that order, overseen by a board-level capital allocation and risk committee. No dividend.

AN AutoNation Value per share

States the goal as growing long-term value per share and tests every use of capital against the market price of its own stock versus its view of intrinsic value, alongside acquisitions, facilities and debt repurchases. It bought back about 4.1m shares in FY2025 at an average of roughly $193 and had close to a billion dollars of authorisation left in February 2026. No dividend. The most Buffett-shaped policy statement in the set, and the share count backs it.

SAH Sonic Automotive Capital structure

Frames repurchases as maintaining a target capital structure and offsetting equity-compensation dilution rather than as a return-maximising act, with a brand strategy of tilting further toward luxury and mid-line import. It operates under a lease-adjusted leverage covenant of 5.75× and a dividend basket negotiated with its lenders, which is where the constraint actually sits.

LAD Lithia Motors Most explicit

A stated free-cash-flow split: 25–35% to acquisitions, 25% to capital expenditure, innovation and diversification, 40–50% to dividends and buybacks, which it says reflects acquisition valuations relative to its own share price. Acquisition discipline is spelled out too — a target after-tax return above 15%, three to six times intangible investment to annualised adjusted EBITDA, intangibles of 15–30% of annualised revenue. The only one of the eight that publishes a percentage framework, and FY2025 spending matched it.

PAG Penske Automotive Dividend-first

No percentage framework in the 10-K. In practice the policy is visible in the cash flows: by far the largest dividend in the set and raised every year, modest buybacks, steady capex into owned real estate, and opportunistic rather than programmatic acquisitions. Penske Corporation’s controlling interest is the structural reason buybacks stay small — and the reason the company is now being taken private, which ends the question of what it would have done next with public shareholders’ capital.

KMX CarMax Under review

Investing in digital capability and store and capacity expansion, pursuing CarMax Auto Finance’s expansion into the full credit spectrum, growth through investments, partnerships and acquisitions, and returning excess capital. The notable line is the admission that it is working to update its strategy and long-term objectives for fiscal 2027 — a reset in progress. Buybacks are the only cash return; there is no dividend.

CVNA Carvana Reinvest everything

No framework, because there is nothing to allocate yet: no dividend, no buyback, and three years of net equity issuance through an at-the-market programme alongside capex that is a fraction of any dealer’s. Capital goes into inventory, reconditioning capacity and market penetration. The structure matters as much as the policy — Carvana Co. is a holding company whose only asset is an interest in Carvana Group, it depends on distributions from it to meet its own obligations, and the tax receivable agreement directs a large future cash stream to the founding family.

Sources and conventions

Where a convention is a choice rather than a figure, the choice is stated here and applied to all eight. Nothing on this page is adjusted for one company and not another.

• Standardised financial data — Standardised income statements, balance sheets and cash flow statements from 2018, quarterly data through the September 2026 quarter, capitalisation, prices and consensus estimates.
• Forms 10-K and 10-Q, SEC — Revenue and gross profit by line of business, brand mix, franchise and store counts, geographic footprint, captive finance disclosure, stated capital allocation policy and non-recourse debt.

EBITDA

Operating income plus depreciation on a standardised template, struck before floor-plan interest, asset impairments and gains on store divestitures. Penske’s adds the equity earnings of its 28.9% stake in Penske Truck Leasing, because enterprise value carries the stake. At CarMax it is already net of CarMax Auto Finance’s interest expense, which sits inside cost of revenue.

Enterprise value

Equity value plus corporate debt, finance and operating leases and commercial paper, less cash and short-term investments. Floor plan and captive non-recourse debt are excluded.

Return on invested capital

Operating income taxed at a uniform 25%, over equity plus corporate debt, leases and floor plan, less cash.

The regional map’s units

Each retailer is measured in whatever unit its own filing publishes, then normalised to its own US total: new-vehicle units at Group 1 Automotive, share of revenue at AutoNation and Sonic Automotive, franchise counts at Asbury Automotive and Penske Automotive, store counts at CarMax. So the four shares compare within a panel and the shading compares across panels, but the underlying unit is not the same company to company. Lithia Motors publishes no geographic split — its segments are Vehicle and Financing — and Carvana sells one national inventory pool rather than a territory, so both panels are drawn flat rather than estimated. Regions are the four US Census regions.

The finance bar

On the gross-profit chart, finance is one bar rather than two: a dealer’s F&I commissions and, at CarMax, captive finance income. They are different businesses — a commission on a third party’s loan against the spread on a book held on balance sheet — but combining them is what makes the share of gross profit that is neither a vehicle nor a part comparable across the two models.

What the charts carry

Every chart on this page draws seven of the eight, bar the regional map, which draws all eight. Carvana is in the prose, in the brands table, in the policy notes, in the workbook and on that map, where it is drawn flat because national e-commerce has no territory to shade. It is on no other chart here: at 31.0× EBITDA, a 2018 base small enough to index into the thousands and a return on equity running from −153% to +21%, it sets the extent of any shared axis by itself. Leaving it in made the other seven unreadable, which is a worse distortion than leaving it out and saying so. Two further charts drop a name for a reason given in their own captions.

Carvana’s equity value

Total economic equity: 722.7m Class A plus 380.5m Class B shares at the Class A price, both after the five-for-one split of May 2026. Enterprise value also adds the $2.1 B tax receivable agreement liability, of which $1.6 B is owed to related parties.

CarMax’s year

A February year end. Each fiscal year is labelled by the calendar year it mostly covers, so 2025 here is the year ended February 2026, and its trailing figures run to 31 August 2026.

Operating cash flow

Not comparable across these eight, and deliberately not used. Floor-plan borrowing from a manufacturer’s captive sits in operating activities while the same borrowing from a third-party lender sits in financing; captive loan originations run through operating cash flow at Lithia and AutoNation; and Carvana’s originations and sales both pass through it.

Commentary

Judgement, not data. Every figure here is sourced and reproducible from the filings; the conclusions drawn from them are not, and this is not investment advice.