Carvana Co. (NYSE: CVNA) — A Magnificent Machine Running on a Subprime Tank of Gas

Company: Carvana Co. | NYSE: CVNA | CIK 0001690820 | FYE Dec 31 | HQ Tempe, Arizona | IPO April 2017 Sector: Consumer Discretionary — Automotive Retail (online used-car retail + vertically integrated reconditioning/logistics + auto finance/securitization). GICS sub-industry: Automotive Retail. Date: 2026-06-07 | Reference price: ~$65.81/sh (2026-06-06; post a 5-for-1 forward split effective 2026-05-07; public market-data aggregators, unofficial) | Market cap: ~$72–73B | Economic shares: ~1.097B (Class A 716.3M + Class B/LLC units 380.5M, post-split) | Enterprise value: ~$75B

⚡ Claude’s Take

This block is the author’s own independent, subjective opinion. It is general information and not investment advice. Everything below it (the analytical body, Sections 1–16) deliberately carries no recommendation and no price target.

Verdict: AVOID at ~$66 — but NOT a short. For owners, HOLD/trim into strength. Accumulate-on-weakness only below ~$45–48 (post-split), with a defensible fair-value zone of roughly $48–$60. Framing: this is momentum / quality-growth-at-a-rich-price into a late-cycle credit backdrop — not deep value, and not a clean short despite the crowded ~13–14%-of-float short position. The zone maps to ~22–28× forward earnings (post-split FY26E EPS ~$2.05) and ~22–28× EV/adj-EBITDA; at ~$66 you are paying ~32× forward earnings and ~33× EV/adj-EBITDA for flawless execution through a subprime cycle that is visibly deteriorating.

Carvana is, operationally, one of the most impressive turnarounds in US retail: from a 2022 near-bankruptcy (negative equity, a ~$5.5B junk-debt stack, an over-built cost base) to FY2025 revenue of $20.3B (+49%), 596,641 retail units (+43%), $2.24B adjusted EBITDA (11.0% margin), $1.0B of operating cash flow, and equity back to +$4.2B. The scale-economies thesis is real: a national logistics + reconditioning network anchored by the $2.2B ADESA acquisition (56 auction sites, capacity toward 3M units) that Vroom and Shift died trying to replicate. What the market is pricing correctly: the genuine operating leverage, the durability of unit-growth momentum (six straight quarters of ≥40% unit growth), and a long runway against a ~1.5% share of a fragmented ~38M-unit market. What I think it is mispricing — or choosing to ignore — is the quality and durability of the earnings. Over 41% of FY2025 gross profit ($1,733M of $4,192M) is “Other” gross profit — overwhelmingly gain-on-sale of securitized auto loans ($1,193M = 28.5% of gross profit) plus ancillary VSC/GAP commissions — i.e. ~100%-margin, credit-cycle- and ABS-spread-sensitive income, much of it routed through related parties. The retail-vehicle margin is a thin ~$3,315/unit (~13.6% of a $24,365 ASP). And the leading indicators have already turned: total GPU fell −2.2% YoY in Q1 2026 and adjusted-EBITDA margin compressed ~110bp, even as units hit records — into a backdrop where US subprime auto 90+ day delinquencies hit 5.2% (highest since 2010). You are being asked to pay a ~33× EBITDA multiple for a thin-margin, capital-intensive, finance-levered used-car flipper at the exact moment its richest profit pool faces its first real cyclical test.

Layered on top is a governance structure I cannot underwrite at a premium: a dual-class “controlled company” where Ernest Garcia II sold ~$2.5B of stock (zero open-market buys) into the very rally his son engineered, a web of DriveTime related-party dealings (FY2025 VSC commissions alone $338M), and a $2.3B Tax Receivable Agreement payable (~$1.7B owed to insiders) that quietly drains future cash to the family. Conviction: medium. Flips bullish if: GPU re-accelerates and the 2025/2026 loan vintages hold their gain-on-sale economics through the credit downturn (proof the finance engine is structural, not cyclical) — at a better price. Flips bearish (toward short) if: unit growth decelerates below ~25% while GPU keeps falling and ABS gains compress — the combination that de-rates the multiple and the earnings simultaneously. One-line tag: a magnificent machine running on a subprime tank of gas.

Standing note for the analytical body (Sections 1–16): The analysis below carries no BUY/SELL recommendation and no price target. Valuation is discussed only as embedded expectations and bear/base/bull scenarios. The single position-taking exception in this document is the Claude's Take block above, which is explicitly the author’s own opinion.

1. Executive Summary

Carvana is the largest online-only used-car retailer in the United States and, increasingly, a vertically integrated automotive platform: it buys, reconditions, sells, delivers, and finances used vehicles, and it monetizes the loans it originates by securitizing and selling them. After a textbook capital-cycle boom-and-bust — a debt- and equity-funded 2017–2021 land grab, a 2022 demand air-pocket that pushed the company to the edge of bankruptcy (negative stockholders’ equity, a 2022 operating loss of $2.35B), and a 2023 distressed debt exchange — the business has staged a genuine operational recovery. FY2025 delivered revenue of $20.3B (+48.6%), 596,641 retail units sold (+43.3%), gross profit of $4.19B, GAAP operating income of $1.88B, adjusted EBITDA of $2.24B (11.0% margin), net income of $1.41B, and operating cash flow of $1.04B. Equity has rebuilt to +$4.2B.

The investment question is not whether Carvana is growing — it plainly is, faster than almost any company its size — but whether the economics of that growth are durable and whether the price discounts a flawless outcome. Three facts frame the debate. First, earnings quality. Over 41% of FY2025 gross profit is “Other” gross profit, dominated by ~$1.19B of gain-on-sale on securitized auto loans — a ~100%-margin stream that depends on functioning ABS markets, tight credit spreads, and benign subprime loss assumptions. The core retail vehicle gross margin is a thin ~$3,315 per unit. Second, the cycle is turning against the richest profit pool. Q1 2026 already showed total GPU down 2.2% YoY and EBITDA margin down ~110bp, while US subprime auto delinquencies reached their highest level since 2010. Third, valuation and governance. At ~$66 post-split (~$72–73B market cap, ~$75B EV), the stock trades at ~39× trailing / ~31× forward earnings, ~33× EV/adjusted-EBITDA, and ~3.7× EV/revenue — multiples that embed years of continued hyper-growth toward management’s stated 3M-unit / 13.5%-margin goal. The Up-C dual-class structure concentrates control and economics with the Garcia family, whose ~$2.8B of stock sales, DriveTime related-party arrangements, and $2.3B Tax Receivable Agreement (≈$1.7B to insiders) are a structural reason to discount, not premium-ize, the equity.

Business quality (Section 3): a real but contestable economies-of-scale advantage (national reconditioning + logistics + ADESA). It is a lead, not a fortress; rapid share gains fail the market-share-stability test for a settled moat. Industry (Section 2): structurally mediocre — commoditized, cyclical, credit-dependent, thin-margin — but enormous and fragmented, rewarding the scaled low-cost winner. Financial quality (Section 5): improving dramatically at the operating line, but margin-flattered by financialized, cyclical income and carrying meaningful (if reduced) leverage at 9–14% coupons. Capital allocation (Section 6): reckless in 2021–22, genuinely well-executed in 2023–25 (distressed exchange, equity raised into euphoria, deleveraging) — but discounted by self-dealing and dual-class control. Valuation (Section 10): prices the bull case; the downside is not cushioned because a credit-cycle disappointment would compress the multiple and the earnings together. No price target or recommendation is offered in this body; scenarios and embedded expectations are in Section 10.

2. Business Overview

What Carvana does. Carvana operates the largest online-only used-vehicle retail platform in the United States, and around that storefront it has built a vertically integrated automotive operating system. A customer can, entirely online, browse a centralized inventory of tens of thousands of reconditioned vehicles (75,683 total website units at FY2025), buy or finance a car in minutes, trade in their existing vehicle, and have the car delivered to their door or picked up from one of the company’s signature glass “vending machine” towers — with a seven-day return guarantee. Behind that interface sits the asset-heavy machine that actually creates the economics: a national network of inspection and reconditioning centers (IRCs), the ADESA physical auction footprint (56 US sites acquired for $2.2B in 2022), a proprietary logistics/transport network, and an in-house auto-finance origination and securitization operation.

How it makes money — four stacked revenue streams per car. Management is explicit that the single retail unit sold is the atomic driver, because each unit unlocks multiple monetization layers (FY2025 figures):

• Retail vehicle sales — $14,537M revenue (71.5% of total), the sale of the car itself at a ~$24,365 average selling price; retail vehicle gross profit ~$3,315/unit (~13.6% of ASP).
• Wholesale sales and revenues — $4,052M (19.9%), sale of trade-ins and vehicles not meeting retail standards, plus third-party wholesale-marketplace activity through ADESA; ~$806 gross profit per retail unit.
• Other sales and revenues — $1,733M (8.5% of revenue but 41.3% of gross profit): predominantly gain-on-sale of the auto loans Carvana originates (sold/securitized into the ABS market — $1,193M in FY2025) plus commissions on complementary products (vehicle service contracts, GAP insurance — many sold via the DriveTime Master Dealer Agreement). This is the high-margin, credit-cycle-sensitive layer.

The finance engine. When a customer finances through Carvana, the company originates the loan, warehouses it, and then sells it — either to a forward-flow buyer (historically Ally) or via its own securitization shelf — recognizing a gain on sale up front. This converts a single car sale into a vehicle margin plus a finance margin plus an ancillary-product margin, and is the principal reason Carvana’s “total GPU” (~$7,026) screens far above CarMax’s retail GPU (~$2,322). It is also the reason a used-car retailer’s earnings are tethered to the consumer-credit and ABS-spread cycle.

Revenue mix and recurrence. Revenue is transactional, not recurring — there is no subscription or contracted backlog; every dollar must be re-earned each period by selling another car. It is also cyclical and credit-sensitive (discretionary big-ticket purchase, financed by a credit-thin customer base). The business is essentially 100% US-domestic. Scale: ~23,100 employees; FY2025 revenue $20.3B; 596,641 retail units; ~18.5M average monthly unique website visitors.

Corporate structure (important). Carvana uses an “Up-C” / umbrella-partnership C-corporation structure: the public entity, Carvana Co., holds a controlling interest in the operating entity Carvana Group, LLC, while the founder Garcia family (and DriveTime) hold the remaining LLC units paired with high-vote Class B shares. This creates (a) a non-controlling interest that takes a large slice of net income ($488M of $1,407M in FY2025), (b) a dual-class voting structure giving the Garcias control (Class B = 10 votes/share) and “controlled company” status, and © a Tax Receivable Agreement obligating Carvana Co. to pay ~85% of certain realized tax benefits to the LLC-unit holders (a $2.3B payable, ~$1.7B to insiders). Following a 5-for-1 forward stock split effective May 7, 2026, total economic shares are ~1.097B (716.3M Class A + 380.5M Class B/LLC units).

3. Industry Dynamics

Market structure and size. The US used-vehicle market is one of the largest consumer markets in the world: roughly 38–39 million used units change hands annually, against a retail value on the order of $850 billion–$1 trillion depending on source and definition (Mordor Intelligence / IMARC, accessed 2026-06-07). It dwarfs the new-vehicle market (~16M units) in unit terms. Crucially for the investment case, it is extraordinarily fragmented and barely digitized: even after a decade of consolidation, the largest single retailer (CarMax) holds only low-single-digit share, the top dealer groups collectively account for under ~10% of dealerships, and an estimated ~66% of 2025 used-car sales still occurred offline, with online penetration growing only ~7% per year. (Industry data, accessed 2026-06-07.)

This combination — vast, fragmented, low online penetration — is the structural bull case. It is precisely the profile that rewards a national, technology-enabled, low-cost consolidator: there is enormous share to take from sub-scale independents and franchised dealers whose used operations are an afterthought to new-car sales and service.

Profit pools and economics. The catch is that the underlying activity — buying and reselling commoditized, price-transparent inventory — is structurally low-margin and cyclical. A used car is a depreciating, individually-priced asset; gross margins on the vehicle itself are thin (Carvana’s retail vehicle GPU is ~$3,315, ~13.6% of ASP; CarMax’s retail GPU is ~$2,322). The profit pools that are attractive sit adjacent to the metal: financing (interest margin or gain-on-sale of originated loans), F&I products (vehicle service contracts, GAP insurance), and wholesale/reconditioning scale. This is why every serious used-car retailer is really a finance company wearing a retailer’s clothes — and why the economics rise and fall with the consumer-credit cycle.

Competitive intensity. The market is intensely competitive at the vehicle-retail level (price transparency via the internet has compressed dealer pricing power) but has thinned dramatically among online-only players. The 2020–2022 capital-cycle boom funded a wave of e-commerce entrants; the 2022–2023 bust killed most of them. Vroom wound down its used-car e-commerce operation in 2024 (≈90% workforce reduction); Shift went bankrupt. Their failures are the single most important piece of competitive evidence in this report: they demonstrate that the online used-car model has no inherent moat at sub-scale — only density and vertical integration make the logistics and reconditioning economics work, and only Carvana reached that density. The serious remaining competitor is CarMax (KMX) — national, now omni-channel, with a captive finance arm (CAF) — plus the omni-channel pushes of franchised groups (Lithia/Driveway, AutoNation, Penske, Sonic/EchoPark).

Regulatory landscape. Auto retail and financing are regulated at both federal (FTC, CFPB) and state (DMV/dealer-licensing) levels. Two 2025 developments matter: the FTC CARS Rule (the “junk-fee”/disclosure rule for auto dealers) was vacated by the Fifth Circuit in January 2025 on procedural grounds and not appealed — lifting a near-term federal compliance burden; and the patchwork of state dealer-licensing and DMV titling rules remains the principal friction for a national online seller (Carvana has faced periodic state licensing disputes, e.g., historical Illinois/Michigan/North Carolina suspensions). Auto-finance practices remain exposed to CFPB/state-AG scrutiny.

Macro backdrop (2024–2026). Used-vehicle prices (Manheim Index) ended 2025 roughly +0.4% YoY and are forecast ~+2% in 2026 — i.e., normalized depreciation, neither the 2021 windfall nor a crash (Cox Automotive). Tariffs (25%+ on imported vehicles from April 2025) raised new-car prices ~10–15%, pushing marginal demand toward used — a modest tailwind to used-vehicle volume and pricing. The genuine headwind is credit: New York Fed data put serious (90+ day) auto-loan delinquencies at ~5.2% in Q4 2025, the highest since 2010, with subprime stress elevated into 2026 — a direct threat to both Carvana’s finance-gain profit pool and to demand from its credit-thin customer base. The September 2025 Chapter 7 collapse (amid fraud allegations) of subprime auto lender Tricolor is a warning flare for the deep-subprime end of the market.

Capital-cycle read. The industry sits in a mid-cycle “high returns attract capital” phase. The prior over-capital phase already purged weak online hands (Vroom, Shift), leaving Carvana with a favorable, less-crowded online supply side today. But Carvana’s own aggressive capacity buildout (ADESA megasite conversions toward >3M units of reconditioning capacity) is itself the next supply wave — and its re-rated equity and ~$7K GPU are the magnet pulling capital (its own capex; dealer groups’ omni-channel spend) back in. Mean-reversion risk on GPU is rising, not falling.

Verdict: structurally mediocre industry, attractive only for the scaled winner. Commoditized inventory, thin vehicle margins, price transparency, cyclicality, and heavy credit dependence make this a bad business for the average participant. Its saving grace — enormous size and fragmentation — disproportionately rewards a national low-cost operator with adjacent finance economics. Carvana is positioned to be that winner, but it is competing for an attractive position within a fundamentally average industry, and the most attractive part of its profit pool (finance) is the most cyclical.

4. Competitive Position

The moat, named. Carvana’s competitive advantage, to the extent it has one, is economies of scale — specifically a national, vertically integrated inspection/reconditioning + logistics + auction footprint that lowers unit cost and shortens days-to-sale as throughput rises. The 2022 acquisition of ADESA US for $2.2B is the cornerstone: 56 physical auction sites placing ~78% of the US population within 100 miles of an inspection-and-reconditioning center (IRC), with reconditioning capacity expandable from ~2M toward >3M units per year. Through Q3 2025, Carvana had integrated 15 ADESA sites, operated 33 retail inventory pools, and was building combined auction-IRC “megasites.” By the end of 2025 it cited fully built-out retail production capacity above 1.5M units per year. The sunk real-estate and reconditioning network is the one asset a new entrant cannot cheaply replicate — and the graveyard of Vroom and Shift is the proof that sub-scale online retail does not work.

Is it durable? Pressure-testing. Three tests, applied honestly:

• Network effects: NO. There is no genuine two-sided flywheel that strengthens with each user. More buyers do not make the platform more valuable to other buyers; inventory is commoditized. Management and bulls sometimes invoke “data” and “selection” advantages, but neither compounds into a defensible network. (INTERPRETATION.)
• Switching costs / brand: WEAK-TO-MODEST. Used-car purchase is an infrequent, high-consideration, price-transparent transaction. Buyer switching cost is ~zero; a shopper compares Carvana, CarMax, AutoNation, and the local dealer on price and selection in the same browser session. Carvana has built real brand awareness (the vending machines, the 7-day return, the delivery experience), which lowers customer-acquisition friction — but brand here is a marketing asset, not pricing power. (INTERPRETATION.)
• Market-share stability (the central test): FAILS. A settled moat shows stable share; a land-grab shows surging share. Carvana’s share is surging (~1.5% and climbing fast). That is wonderful for growth investors but is the signature of an unsettled, contestable market, not a fortress. The honest read: Carvana has a scale lead, not an unassailable moat. (INTERPRETATION.)

Direct comparison vs. CarMax. CarMax is the only competitor at comparable scale: ~765,000 used units (FY ending Feb 2025), ~$26.4B revenue, a national footprint, a maturing omni-channel offering, and a captive finance arm (CarMax Auto Finance) that books income separately from GPU. The headline GPU gap — Carvana’s ~$7,026 “total GPU” vs. CarMax’s ~$2,322 retail GPU — is not apples-to-apples: Carvana stacks retail + wholesale + finance/ancillary into one number, whereas CarMax reports CAF outside GPU. On a like-for-like retail-vehicle basis the margin gap is much smaller; Carvana’s real edge is (a) lower fixed SG&A per unit at scale (no traditional store network) and (b) a larger captured finance/ancillary take per unit. The first is a genuine, defensible structural advantage; the second is the cyclical, credit-sensitive piece.

Cost advantage or capital intensity? Vertical integration is simultaneously a scale cost advantage and a capital-intensity liability. When utilization is high, the IRC/ADESA network drives down reconditioning cost and days-to-sale — a flywheel of operating leverage that the FY2024–25 numbers validate (SG&A per retail unit fell sharply as units grew). When demand falls, that same fixed-cost base becomes a drag — exactly the 2022 near-death experience. The advantage is therefore conditional on sustained throughput growth, which makes it less robust than a true low-cost-producer moat that holds across the cycle.

Verdict: a real but contestable scale lead, not a durable fortress. Carvana has the strongest position in online used-car retail and a structural SG&A-per-unit advantage that should persist if volume holds. But the inventory is commoditized, switching costs are near zero, there are no network effects, share is still being fought for (not defended), and CarMax is a credible scaled rival pursuing the same omni-channel + captive-finance model. The piece of the advantage that is most durable (logistics/reconditioning scale) supports a thin retail margin; the piece that is most lucrative (finance gains) is the least durable. A genuine moat must tie to a financial outcome that would deteriorate without it — Carvana’s would deteriorate (and is beginning to, on GPU) the moment growth slows.

5. Growth History and Forward Opportunities

History — boom, bust, recovery. Carvana’s revenue history is a capital-cycle parable: $0.86B (2017 IPO year) → $5.6B (2020) → $12.8B (2021) at the euphoric peak → a pull-back to $10.8B (2023) as the company slammed the brakes to survive → $13.7B (2024) → $20.3B (2025). Retail unit volume tells the cleaner operational story because it strips out used-car price inflation/deflation: units were cut hard in 2023 (312,847, down from the 2021–22 peak as the company prioritized per-unit economics and cash over growth), then re-accelerated to 416,348 (2024, +33%) and 596,641 (2025, +43%). Q1 2026 marked the sixth consecutive quarter of ≥40% YoY unit growth (187,393 units, +40%). Growth is essentially 100% organic post-ADESA — driven by share gains, geographic densification, and improved selection/days-to-sale — not acquisition.

Quality of the growth. The 2024–25 growth is materially higher quality than the 2019–2021 growth: it has been accompanied by expanding (not collapsing) per-unit economics and positive operating cash flow, rather than the cash-incinerating “growth at any cost” of the pre-2022 era. That is the single most important favorable change in the story. The caution flag is that per-unit gross profit has stopped expanding and begun to slip (total GPU −2.2% YoY in Q1 2026), so incremental growth is now coming at a flat-to-declining margin — the inflection from “margin-and-volume” to “volume-only” growth that often precedes multiple compression in a richly-valued name.

Forward opportunities.

• Share gains in a fragmented market. At ~1.5% of ~38M units, the runway is real. Management’s stated long-term ambition is 3 million units per year at a 13.5% adjusted-EBITDA margin within 5–10 years — implying roughly 7–8% market share and a >5× increase in volume. This is the central bull number.
• Reconditioning capacity already built. The ADESA megasite program means much of the physical capacity for 1.5M–3M units is already in place or in flight, so incremental volume should be capital-light relative to the 2018–2021 buildout — supporting the operating-leverage thesis if demand fills it.
• Finance/ancillary monetization. Deeper attach of financing, VSC, and GAP, plus optimization of loan-sale execution, is a profit lever — but it is the cyclical lever.
• Adjacencies. Wholesale marketplace (ADESA), the auction business, and potential expansion of the finance platform are optionality, not core.

Verdict: high-rate, now-higher-quality growth — but quality is plateauing at the margin. The volume runway is genuine and the organic, cash-generative nature of recent growth is a real positive versus the company’s own history. But the leading edge of per-unit economics has turned down, and the most lucrative growth lever (finance) is the most cyclical. The growth is high-quality for now; whether it stays so depends on the credit cycle and on Carvana’s ability to grow units without buying volume with margin.

6. Financial Quality

Income statement — a genuine operating turnaround. The trajectory is unambiguous at the operating line: gross profit $1,724M (2023) → $2,876M (2024) → $4,192M (2025); GAAP operating income −$2,351M (2022) → −$80M (2023) → +$990M (2024) → +$1,881M (2025); adjusted EBITDA $339M → $1,378M → $2,237M, with margin expanding from 3.1% to 10.1% to 11.0%. SG&A discipline is the engine: total SG&A grew from $1,796M (2023) to $2,308M (2025) while revenue nearly doubled, so SG&A per retail unit fell sharply — textbook operating leverage off the fixed reconditioning/logistics base.

But the margin is financialized. Decomposing FY2025 gross profit:

Within “Other,” gain-on-sale of originated auto loans was ~$1,193M in FY2025 — 28.5% of total gross profit (up from $755M in 2024 and $434M in 2023). This income is recognized when Carvana sells/securitizes the loans it originates; it is ~100% gross margin and depends on (a) functioning, tight-spread ABS markets, (b) benign assumed loss rates on the underlying (often subprime) loans, and © continued origination volume. It is the highest-quality-looking and lowest-durability-quality earnings stream simultaneously. The retail-vehicle margin that anchors the business is a thin ~$3,315/unit (~13.6% of ASP). A serious investor must mentally separate the durable retail/wholesale/SG&A-leverage story (~58% of GP) from the cyclical finance story (~41% of GP). (INTERPRETATION — central to the bear case.)

Quality-of-earnings flags.

• FY2023’s $450M GAAP net income was an artifact of an $878M gain on debt extinguishment from the distressed exchange; underlying operating income that year was −$80M. Anyone anchoring on 2023 “profitability” is misreading the recovery’s start date — the real operating inflection was 2024.
• Net income includes large non-controlling interest. FY2025 net income of $1,407M includes $488M attributable to NCI (the Garcia/LLC-unit holders); only ~$919M is attributable to Carvana Co. (Class A). Per-share and ROE figures must be read against the right denominator and numerator (the Up-C structure).
• Q1 2026 is the canary. Records on units (187,393, +40%) and adjusted EBITDA ($672M) coexisted with total GPU −2.2% YoY ($6,783) and adjusted-EBITDA margin −110bp (10.4%); “Other”/finance was ~41% of Q1 gross profit. Volume up, per-unit economics down. (FACT — Q1 2026 10-Q.)

Cash flow. A real positive: operating cash flow turned firmly positive — +$803M (2023), +$918M (2024), +$1,036M (2025) — and investing outflows are modest (capacity largely built), so the business is now self-funding. This is the strongest single piece of evidence that the recovery is real and not merely accounting. Note, however, that working-capital/inventory swings and the timing of loan originations vs. sales can make quarterly cash flow lumpy, and that a portion of “operating” cash relates to the finance book.

Balance sheet. From the abyss to merely-leveraged: total stockholders’ equity (incl. NCI) was −$1,053M (2022) and −$384M (2023) before recovering to +$1,375M (2024) and +$4,203M (2025) (NCI $762M; Carvana Co. $3,441M). Cash is $2,327M; total assets $13,201M; inventory $2,408M. Total debt is ~$4,976M, of which the bulk is the high-coupon senior secured notes (see Section 7 for the schedule); much of the remainder is asset-based (floor-plan, finance-receivable securitization, real-estate facilities), some non-recourse. Net debt is ~$2.6B (~1.2× adjusted EBITDA) — a dramatic improvement from the existential leverage of 2022. The non-obvious liability is the $2.3B Tax Receivable Agreement payable (≈$1.7B to insiders), a real future cash claim that most screens miss.

Returns on capital. Reported ROE looks spectacular (~60% on the small, recovering equity base) but is flattered by a depressed/recovering denominator, leverage, and the NCI structure — it is not a clean signal of through-cycle return on capital. A more honest read: the business is now generating real operating profit and cash, but a normalized, de-levered, through-cycle ROIC for a thin-margin used-car retailer-plus-lender is likely good-not-great, and is partly manufactured by the finance book. (INTERPRETATION.)

Verdict: economics genuinely improve with scale — but the headline margin is propped up by cyclical, financialized income, and the leading indicators have rolled over. The SG&A operating leverage is real and durable; the gain-on-sale finance margin is real but cyclical and now the swing factor in both directions. The balance sheet is repaired but still carries expensive debt and a hidden TRA claim. This is a much higher-quality financial profile than 2022 — and a lower-quality one than the GPU headline suggests.

7. Capital Allocation

The two eras. Carvana’s capital-allocation record splits cleanly. Era 1 (2017–2022): reckless. Debt- and equity-funded hyper-expansion — culminating in the $2.2B ADESA acquisition (2022) layered on top of a ~$5.5B unsecured note stack — left the company with negative equity and on the brink of bankruptcy when demand softened and rates rose. This was value-destroying growth financed pro-cyclically at the top of the cycle: a textbook capital-cycle blunder. Era 2 (2023–2026): genuinely well-executed. Management navigated out of the hole with a series of intelligent moves.

The 2023 distressed exchange. In September 2023 Carvana exchanged ~$5.5B of unsecured notes into ~$4.2B of new senior secured notes (a ~$1.3B principal reduction) and cash-tendered near-term maturities at a discount — producing the $878M extinguishment gain that flattered 2023 GAAP earnings. Critically, the new notes carried a PIK (payment-in-kind) toggle, deferring cash interest during the recovery.

Debt schedule and deleveraging (FY2025 10-K, Note 9):

Deleveraging has been real: the 2028 tranche was fully retired in 2025 (~$611M redeemed/repurchased for ~$633M, a small premium); the PIK toggle flipped to cash-pay on the 2028 and 2030 notes through 2025 (only the 2031 tranche still PIKs, adding ~$287M of principal in 2025); and net interest expense fell to $505M (2025) from $651M (2024)/$632M (2023) despite the punitive 9–14% coupons. The maturity wall is back-loaded to 2030 ($1.66B) and 2031 ($2.27B), giving runway — but the cost of this debt is high, and the 9% cash step-down is a meaningful ongoing drag.

Equity issuance — using an overvalued currency well. Management raised equity into strength: ~6.83M shares at ~$186.56 in FY2024 (~$1.27B) and ~1.48M shares at ~$364.93 in FY2025 (~$539M) — roughly $1.8B raised at pre-split prices of $187–$365, versus sub-$10 lows in 2022–23. Issuing an expensive currency to delever is exactly what a rational manager should do; it is the mirror image of the 2021 mistake. SBC is modest (~$96M in FY2025 against $1.4B net income). No buybacks, no dividend — appropriate for a company still carrying 9–14% debt and reinvesting in growth.

The governance discount — three structural flags.

1. Insider selling, zero buying. Per the insider-transaction (Form 4) filings, Ernest Garcia II (Chairman) sold ~12.94M Class A shares for ~$2.47B across ~96 filings, with zero open-market purchases, by converting Class B→Class A and selling same-day in laddered lots under successive Rule 10b5-1 plans; CEO Ernest Garcia III sold ~1.03M shares for ~$349M, also zero buys. Combined: ~14M shares, ~$2.82B monetized, no purchases — the family harvested ~$2.8B during the rally while public holders absorbed ATM dilution. 10b5-1 cover mitigates the legal concern but not the signal: this is not insider conviction at the current price. (FACT — SEC Form 4 filings.)
2. DriveTime related-party web. DriveTime Automotive (privately controlled by Garcia II) is a recurring counterparty: FY2025 vehicle-service-contract commissions to Carvana of $338M (Carvana’s largest related-party line, up from $193M/$138M), plus wholesale purchases ($27M), warranty administration ($24M), and servicing. These arrangements are disclosed and may be arm’s-length, but they route a meaningful, growing slice of Carvana’s high-margin ancillary economics through a vehicle the controlling family also owns. (FACT — 10-K Note 6.)
3. Tax Receivable Agreement. The Up-C structure carries a $2.3B TRA payable, ~$1.7B of it owed to insiders — a contractual future cash transfer from Carvana Co. (public shareholders) to the Garcia/LLC-unit holders as tax benefits are realized. A real, often-overlooked drain on per-Class-A-share value. (FACT — 10-K.)

Incentive alignment. Carvana is a “controlled company” (Garcia >50% of votes via 10×-vote Class B) and declines certain board-independence protections. Executive PSU metrics are operationally sensible — adjusted EBITDA, retail-unit thresholds (600K/750K), and Core FCF — with no annual cash bonus; reported CEO summary-comp is modest ($8.4M) though “compensation actually paid” spiked to ~$110M on the 2021 mega-grant’s appreciation. The metric set aligns management with the operating recovery; the dual-class structure and self-dealing align control and value extraction with the family.

Verdict: mixed, trending positive — but premium-disqualifying on governance. The 2021–22 over-expansion was a serious, nearly fatal capital-allocation error. The 2023–25 recovery — distressed exchange, equity raised into euphoria, debt retired, PIK flipped to cash, no value-destroying M&A — was genuinely skillful, and operating reinvestment is now earning real returns. But dual-class control, ~$2.8B of insider selling with zero buys, an expanding DriveTime related-party take, and a $1.7B-to-insiders TRA mean value is being steered toward the controlling family in ways that warrant a structural discount to, not a premium on, the equity.

8. Major Changes — Last Two Years

• Operational recovery and scale-up (2024–2026): from −$80M operating income (2023) to +$1.88B (2025); units +33% then +43%; six straight quarters of ≥40% unit growth into Q1 2026. The defining change. (FACT.)
• Balance-sheet repair: equity from −$384M (2023) to +$4.2B (2025); 2028 notes retired; PIK→cash on 2028/2030; net interest down to $505M. (FACT.)
• 5-for-1 forward stock split (first ever): board-approved March 13, 2026; shareholder-approved May 5, 2026; effective May 7, 2026 (split-adjusted trading May 8). A signal of management’s confidence and a bid for retail-investor accessibility; cosmetically lowered the share price from ~$329 to ~$66. (FACT — 8-K; OCC infomemo #58924.)
• 2026 Omnibus Incentive Plan approved alongside the split (a roll-over of remaining 2017-plan capacity, retaining a 2% evergreen auto-increase). (FACT — DEF 14A / 8-K.)
• ADESA integration milestones: 15 sites integrated, 33 retail inventory pools, megasite conversions, >1.5M-unit built-out capacity by end-2025; new Jacksonville megasite announced Jan 2026. (FACT.)
• GPU/margin inflection (Q1 2026): total GPU −2.2% YoY, EBITDA margin −110bp — the first clear sign per-unit economics have plateaued. (FACT — Q1 2026 10-Q.)
• Macro/credit deterioration: US subprime auto delinquencies at post-2010 highs; Tricolor’s Chapter 7 (Sept 2025) a subprime warning flare; offset by tariff-driven demand shift toward used and the FTC CARS-rule vacatur. (FACT.)

Verdict: net thesis-strengthening on execution and the balance sheet, but with two clear emerging negatives — the GPU/margin rollover and the worsening subprime-credit backdrop, both of which strike at the most valuable (finance) part of the profit pool just as the equity is priced for perfection.

9. Risk Analysis (Risk Matrix)

Catastrophic-loss assessment. A total loss is unlikely from here — the business now generates ~$1B of operating cash flow, has positive equity, and net leverage near 1.2×. But the 2022 experience proves the tail is fat: a sharp demand contraction colliding with a high fixed-cost base, expensive debt, and a frozen ABS market could re-create existential stress. The combination of high operating leverage, financial leverage, and credit-cycle sensitivity makes CVNA a high-beta (3.55) instrument where a moderate fundamental disappointment can produce an outsized equity drawdown.

10. Financial Quality Snapshot (multi-year)

* FY2023 net income includes an $878M debt-extinguishment gain; underlying operating income was −$80M. (Sources: SEC EDGAR XBRL; FY2025 10-K MD&A. Some early-year per-unit fields not disclosed/not comparable.)

11. Valuation Discussion (Embedded Expectations)

Setup. At ~$65.81 post-split, with ~1.097B economic shares (Class A 716.3M + Class B/LLC units 380.5M), market capitalization is ~$72–73B. Adding net debt of ~$2.6B (and noting a further $2.3B TRA claim, ~$1.7B to insiders), enterprise value is ~$75B. Against FY2025 results that yields:

For context, CarMax — a scaled, profitable, far less levered peer — trades at a small fraction of these multiples (high-teens to low-20s P/E, low-single-digit EV/EBITDA in normal periods). Carvana’s multiple is a growth multiple, not a retailer multiple.

Embedded expectations. What must be true to justify ~$75B EV? A useful frame is management’s own long-term target: 3 million units/year at a 13.5% adjusted-EBITDA margin. At ~3M units and a ~$30K blended revenue/unit, revenue would be ~$90B; at 13.5% that is ~$12B of adjusted EBITDA. Against today’s ~$75B EV, that is ~6.3× the target EBITDA — a reasonable multiple if and when the target is achieved. So the bull math is internally coherent: the current price is roughly “fair” against a fully-executed 5–10-year plan discounted back, assuming (a) the unit-growth ramp continues at high rates, (b) the 13.5% margin is reached and held, and © the finance/ABS profit pool that contributes ~40% of gross profit survives the credit cycle intact. The market is, in effect, underwriting all three.

What the market may be getting right: the genuine operating leverage and SG&A-per-unit decline; the durability of unit-growth momentum (six quarters of ≥40%); the long runway against ~1.5% share; and the balance-sheet repair. These are real and well-evidenced.

What the market may be getting wrong: (1) extrapolating the GPU peak when GPU has already turned down (−2.2% in Q1’26) — if the 13.5% target margin proves to be a peak rather than a plateau, the EBITDA bridge to $12B lengthens or breaks; (2) treating ~$1.2B of cyclical, ABS-dependent finance gains as structural recurring earnings when subprime delinquencies are at post-2010 highs; (3) under-discounting the governance leakage (TRA, DriveTime, dual-class) from per-Class-A-share value.

Scenario analysis (illustrative, 2–3 year horizon; not a forecast):

• Bear: Unit growth decelerates to ~15–20%; GPU reverts toward ~$6,000 as finance gains compress in a credit downturn; EBITDA margin slips to ~8–9%. Adjusted EBITDA stalls near ~$2.0–2.3B; the multiple de-rates toward ~12–18× EV/EBITDA. Equity could fall 40–60%+ — the multiple and the earnings compress together, with no valuation floor.
• Base: Unit growth ~25–30%; GPU holds ~$6,800–7,000; margin ~11–12%. Adjusted EBITDA reaches ~$3.0–3.5B in 2–3 years; if the multiple settles to ~20–25× EV/EBITDA, EV is roughly flat-to-modestly-higher than today — the stock “grows into” a still-rich multiple, delivering muted returns with high volatility.
• Bull: Unit growth sustains ~35–40%; GPU stable-to-up on finance + ancillary attach; margin marches toward 13%. Adjusted EBITDA approaches ~$4–5B; the market keeps a ~25–30× multiple, and the equity compounds materially. Requires the credit cycle to remain benign and execution to stay flawless.

Own-history valuation context. On own-history valuation percentiles, CVNA screens far from its own cheapest levels; the de-rating risk is to its own historical lows, which — given the 2022–23 distress — were extreme. The high beta (3.55) and ~13–14%-of-float short interest add a technical layer: a crowded short against a momentum/retail-favorite name with a fresh split creates genuine squeeze potential on any clean beat, while the GPU/margin/ABS deterioration provides the fundamental ammunition the shorts need. That two-sided tension is the defining feature of the stock at this price.

No price target or recommendation is offered in this body. The embedded expectations are demanding: the price requires the bull scenario to substantially play out, the credit cycle to cooperate, and the governance leakage to be tolerated.

12. Variant Perception

Consensus. The sell-side is broadly constructive — average rating “Buy,” post-split price targets clustered ~$92–97 (RBC $92/Outperform, BTIG $97/Buy, both reiterated June 2026). The consensus narrative: a secular-share-gainer with proven operating leverage and a multi-year runway toward 3M units, deserving a premium growth multiple.

The strongest bull case. Carvana is the only company to have cracked profitable online used-car retail at scale; it has a real SG&A-per-unit cost advantage, ~1.5% share of a fragmented ~38M-unit market, capacity already built for >1.5M units, six straight quarters of ≥40% unit growth, a repaired balance sheet generating ~$1B of FCF, and a founder-operator with skin in the game. If it executes toward 3M units / 13.5% margin, today’s price is reasonable and the upside is large.

The strongest bear case. The valuation (~39× P/E, ~33× EV/EBITDA, ~3.7× EV/sales) prices perfection in a structurally thin-margin, cyclical, commoditized business. ~41% of gross profit is cyclical, ABS-dependent finance/ancillary income arriving exactly as subprime delinquencies hit post-2010 highs; GPU and EBITDA margin have already turned down YoY; the company still carries ~$4B of 9–14% debt; and the dual-class/related-party/TRA structure siphons value to the controlling family, whose insiders have sold ~$2.8B and bought nothing. A modest growth or credit disappointment compresses the multiple and the earnings together.

The 3–5 assumptions that matter most, and what would falsify each:

1. Unit growth stays ≥30% for years. Falsifies bear: a record, as-guided Q2 2026. Falsifies bull: deceleration below ~25% as macro softens.
2. GPU/margin holds (13.5% target is a plateau, not a peak). Falsifies bull: GPU re-accelerates with units. Falsifies bear: two-plus more quarters of YoY GPU/margin decline (Q1’26 was the first).
3. Finance/ABS gain-on-sale is durable through the credit cycle. Falsifies bull: gain-on-sale margin stable as subprime losses rise. Falsifies bear: ABS spreads widen / loss rates rise and loan-sale gains compress.
4. The balance sheet stays comfortably serviceable. Falsifies bull: continued deleveraging and a 2030/2031 refi at lower coupons. Falsifies bear: a demand shock that re-stresses leverage before the wall.
5. Governance leakage stays bounded. Falsifies bear: insiders begin buying / DriveTime take shrinks. Falsifies bull: expanding related-party economics or new control-favoring actions.

The crowded short. ~13–14% of float short (short ratio ~5.8) into a high-beta, retail-favorite, freshly-split name is itself a variant-perception signal: it raises squeeze risk on good news but also reflects a deep, fundamentally-grounded skeptic base. The honest synthesis: both sides are right about different things — the bulls about the operating machine and runway, the bears about earnings quality, the credit cycle, valuation, and governance. The resolution turns on the credit cycle and on whether GPU stabilizes — neither of which is knowable in advance, which is why the risk/reward at ~$66 is unattractive without a margin of safety.

13. Fact vs. Interpretation Table

14. Open Questions

1. Finance-gain durability: What loss assumptions underpin the $1.19B gain-on-sale, and how sensitive is it to a 100–200bp widening in ABS spreads or a step-up in subprime losses? (Not fully disclosed.)
2. Like-for-like vs. CarMax: Carvana all-in (retail + finance + ancillary) vs. CarMax all-in (retail GPU + CAF income) on a normalized per-unit basis — how large is the durable economic edge once finance is treated symmetrically?
3. IRC/ADESA utilization: Current throughput vs. the >1.5M-unit built-out capacity — how much operating leverage is left, and how much fixed-cost risk in a downturn?
4. Securitization mechanics: Gross vs. net (non-recourse) finance-receivable facility balances; retained risk; off-balance-sheet exposure.
5. 2030/2031 refi path: Can the 9–14% notes be refinanced materially lower, and on what timeline? What is the cash-interest run-rate once the 2031 PIK ends?
6. Through-cycle normalized GPU and margin: Is 13.5% a sustainable plateau or a cyclical peak?

15. What Must Be True

Bull case — what must be true (and its falsification test):

• Unit growth sustains ~30–40% for several years toward 3M units; the SG&A-per-unit advantage deepens; the 13.5% EBITDA margin is reached and held; and the finance/ABS profit pool survives the credit cycle. Falsification test: two consecutive quarters of decelerating unit growth (below ~25%) accompanied by continued YoY GPU and EBITDA-margin decline would break the “scale + leverage compounds” thesis and de-rate the multiple.

Bear case — what must be true (and its falsification test):

• The current GPU/margin is a cyclical peak; finance gains compress as subprime losses rise and ABS spreads widen; growth slows into a thin-margin, capital-intensive core; and the multiple normalizes toward a retailer’s. Falsification test: GPU re-accelerates while units keep compounding ≥35% and gain-on-sale margins stay stable through a rising-delinquency environment — proof the finance engine is structural and the operating leverage is durable — would invalidate the bear and justify the premium.

16. Source Note

See the Source Appendix (Appendix B below) for the full source list with URLs and access dates. Primary sources: Carvana FY2023/FY2024/FY2025 10-Ks and FY2026 Q1 10-Q (SEC EDGAR, CIK 0001690820); DEF 14A (2026); 8-Ks (5-for-1 split, results); Form 4 corpus; EDGAR XBRL company facts. Market data: public market-data aggregators (2026-06-06, unofficial, reconciled to filings); OCC infomemo #58924 (split). Industry/macro: Cox Automotive/Manheim, NY Fed Household Debt, Mordor/IMARC market sizing, KBB/CNBC (tariffs), Holland & Knight (FTC CARS rule).

The analytical body (Sections 1–16) carries no recommendation and no price target; the sole position-taking view is the labeled Claude's Take block at the top.

Appendix A — Diligence Questionnaire

Supplemental to the research memo. Answers are labeled Fact / Interpretation / Assumption where it matters.

General

What thoughtful questions have other investors asked about this company? The durable ones: (1) Is online used-car retail a real business or a structurally unprofitable model that only looks viable at a cyclical GPU peak? (Vroom/Shift died proving it can fail at sub-scale.) (2) How much of the profit is cyclical finance income (gain-on-sale of securitized loans) versus durable retail margin? (~41% of gross profit is “Other”/finance — the central question.) (3) Can the company survive the next demand/credit downturn given 2022’s near-bankruptcy? (4) How much value leaks to the Garcia family via dual-class control, DriveTime related-party deals, and the $2.3B TRA? (5) Is the ~33× EV/EBITDA multiple a growth multiple that will compress as growth slows? Kerrisdale’s June-2023 short (“equity is a zero”) was the extreme bear version — wrong on timing through the recovery, but its quality-of-earnings critique (finance-gain reliance) remains the live debate.

Cyclicality & Earnings Nature

Are earnings at a cyclical high or low? Interpretation: closer to a high than a low. GPU and EBITDA margin peaked in 2025 and already turned down YoY in Q1 2026; the finance-gain component (~28% of GP) is benefiting from still-tight ABS spreads even as subprime delinquencies hit post-2010 highs — a configuration that historically reverts.

Driven by external environment or internal actions? Both. Internal: genuine SG&A operating leverage, ADESA reconditioning scale, days-to-sale improvement — real and management-driven. External: tariff-driven demand shift to used, still-functional ABS markets, normalized used-car pricing — cyclical tailwinds that can reverse.

How stable are revenues? Fact: highly unstable historically — $12.8B (2021) → $10.8B (2023) → $20.3B (2025). Used-car demand is discretionary and credit-sensitive; this is a cyclical revenue base, not a recurring one.

Outlook for products/services / market size. Fact: US used market ~38M units/yr, ~$850B–$1T, fragmented, ~66% offline. Large and growing (~+2% pricing in 2026), with online penetration rising ~7%/yr — a long runway for share gains, domestically focused (no material international).

Business Quality & Competitive Moat

Is the industry getting more or less competitive? Interpretation: less crowded online (Vroom/Shift exited), but CarMax and franchised dealer groups are intensifying omni-channel competition; vehicle-retail pricing remains commoditized and price-transparent. Net: still highly competitive at the metal.

How profitable is the business (ROIC, ROE)? Fact/Interpretation: reported ROE ~60%, but flattered by a recovering/small equity base, leverage, and the NCI structure — not a clean through-cycle signal. Normalized ROIC for a thin-margin retailer-plus-lender is good-not-great and partly manufactured by the finance book.

How profitable is the industry — competitors, barriers? Structurally mediocre (thin vehicle margins, cyclical, credit-dependent), attractive only for the scaled winner. Barriers: a national reconditioning/logistics/auction footprint (Carvana’s ADESA-anchored scale) is the real barrier; capital intensity is high.

Can the business be easily understood? Mostly — buy/recondition/sell/deliver used cars and finance them — but the finance/securitization accounting (gain-on-sale, Up-C/NCI, TRA) materially complicates a clean read.

Undermined by foreign low-cost labor? No — physical, domestic, logistics-bound service business.

Do brands matter? Modestly. Carvana’s brand lowers customer-acquisition friction (vending machines, 7-day return) but confers little pricing power on commoditized inventory.

Nature of competition / switching costs? Price, selection, and convenience. Buyer switching costs are ~zero (infrequent, comparison-shopped purchase). No network effects.

Financial Condition & Balance Sheet

Assets not fully recognized on the balance sheet? The brand and the reconditioning/logistics network’s earning power (intangible). The ADESA real estate is on-balance-sheet at cost.

Off-balance-sheet liabilities? The $2.3B Tax Receivable Agreement (~$1.7B to insiders) is a real future cash claim that screens miss. Finance-receivable securitization facilities (some non-recourse) warrant scrutiny for retained risk. Operating-lease obligations exist but are capitalized under ASC 842.

How conservative is the accounting? Interpretation: aggressive in spirit — heavy reliance on gain-on-sale finance income (recognizing future loan economics up front), large non-GAAP adjustments, and the $878M 2023 debt-gain that flattered GAAP. Not fraudulent, but flattering; normalize finance gains and one-offs.

How CapEx-hungry? Historically very (the 2018–2022 buildout nearly sank it). Now lighter, as 1.5M-unit capacity is largely built — a key part of the current FCF-positive story — but a re-acceleration of expansion would re-intensify it.

Capital Allocation & Management

FCF generation and use; philosophy? Fact: operating cash flow ~$1.0B (2025), now self-funding; deployed to deleveraging (2028 notes retired, PIK→cash) and organic growth. No dividend, no buyback (appropriate given 9–14% debt). Philosophy: founder-led growth maximization with opportunistic balance-sheet repair.

Significant acquisitions recently? ADESA US ($2.2B, 2022) — strategically central but pro-cyclically timed; nothing material since (organic growth focus).

Buying back shares? No — issuing into strength instead (~$1.8B of equity raised at pre-split $187–$365 in 2024–25). Rational use of an expensive currency.

Issuing large amounts of stock to insiders? SBC modest (~$96M FY25). But insiders are large sellers: Garcia family ~$2.8B sold, zero buys, under 10b5-1.

Compensation policy / motivations of management? PSU metrics = adjusted EBITDA + retail-unit thresholds (600K/750K) + Core FCF; no annual cash bonus — operationally aligned. But “controlled company” dual-class structure (Garcia 10×-vote Class B) plus DriveTime self-dealing and the TRA mean control and value extraction are steered to the family. Motivation: build a dominant platform while monetizing personal stakes.

Valuation & Market Data

ADR, MLP, or K-1 issuer? No — US C-corp (Carvana Co., Up-C structure with an LLC subsidiary), Class A common on NYSE. Not a K-1 issuer for public holders.

Dividend policy? None; none expected near-term.

How profitable is the business? GAAP net income $1.41B (2025); operating margin ~9%; net margin ~6–7% — but ~41% of gross profit is cyclical finance/ancillary income.

Net income vs. cash from operations diverging? FY2025 net income $1,407M vs. CFO $1,036M — broadly tracking, though both are influenced by finance-book timing and one-offs; FY2023’s GAAP profit diverged sharply from operating cash and operating income due to the debt-gain.

Risks & Downside

What would cause the stock to decline? Multiple de-rating from ~33× EV/EBITDA; GPU/margin normalization (already underway); finance-gain compression in a credit downturn; unit-growth deceleration; a demand/credit shock re-stressing leverage; governance/related-party escalation.

Risk of catastrophic loss? Tail risk is real but reduced. 2022 proved the model can approach bankruptcy; today’s positive equity (+$4.2B), ~$1B FCF, and ~1.2× net leverage make a total loss unlikely from here — but high operating + financial + credit leverage make large drawdowns plausible on moderate disappointment (beta 3.55).

Chance of total loss? Low from current footing, conditional on no re-leveraging into a demand shock before the 2030/2031 maturity wall.

Recent News & Events

Has the business environment changed recently? Yes: (1) 5-for-1 split (May 2026) — cosmetic; (2) GPU/EBITDA-margin turned down YoY in Q1 2026 — fundamental; (3) subprime auto delinquencies at post-2010 highs and Tricolor’s Chapter 7 — credit-cycle warning; (4) tariff-driven demand shift toward used — modest tailwind; (5) FTC CARS rule vacated — reduced near-term regulatory burden.

Significant acquisitions? None recent (post-ADESA).

Change in accounting policies? None material flagged; ongoing reliance on gain-on-sale finance accounting.

Recent operational changes? ADESA megasite conversions, >1.5M-unit built-out capacity, new Jacksonville megasite (Jan 2026); continued geographic densification.

Appendix B — Source Appendix

Primary sources first. Market data from unofficial aggregators is flagged and reconciled to filings. Access date for web/market data: 2026-06-06/07 unless noted.

A. Primary — SEC filings (EDGAR, CIK 0001690820)

1. Form 10-K, FY2025 (filed 2026-02-18, period 2025-12-31). Business, MD&A KPIs (retail units, GPU, gross-profit-by-category), non-GAAP reconciliation (adjusted EBITDA), Note 6 (related party / DriveTime), Note 9 (debt instruments), TRA disclosure, cover (Class A/B share counts as of 2026-02-13). https://www.sec.gov/cgi-bin/browse-edgar?action=getcompany&CIK=0001690820
2. Form 10-K, FY2024 (filed 2025-02-19).
3. Form 10-K, FY2023 (filed 2024-02-22). 2023 debt exchange terms; $878M extinguishment gain.
4. Form 10-Q, Q1 2026 (filed 2026-04-29, period 2026-03-31). Q1’26 units 187,393, revenue $6,432M, total GPU $6,783 (−2.2% YoY), adj EBITDA $672M (10.4%), net income $405M; cover share counts as of 2026-04-27.
5. Form 8-K filings, 2023–2026. FY2025 results (2026-02-18); 5-for-1 split (board 2026-03-13; AGM approval 2026-05-05; effective 2026-05-07); 2026 Omnibus Incentive Plan.
6. DEF 14A / proxy (2026). Executive compensation, PSU metrics (adj EBITDA / retail-unit thresholds / Core FCF), controlled-company status, 2026 plan, related-party detail.
7. Form 4 corpus (insiders), 2023–2025 — SEC EDGAR. Garcia II ~12.94M Class A sold (~$2.47B, ~96 filings, 0 buys, 10b5-1); Garcia III ~1.03M (~$349M, 0 buys).
8. SEC EDGAR XBRL company facts — multi-year IS/BS/CF concepts (revenue, gross profit, operating income, net income, NCI, gain-on-loan-sales, long-term debt, equity, cash, inventory, EPS, shares, cash flows).

B. Market & corporate-action data

9. OCC Information Memo #58924 (2026-05-07) — Carvana Co. 5-for-1 stock split. https://infomemo.theocc.com/infomemos?number=58924
10. Price / market data — public market-data aggregators, 2026-06-06 (UNOFFICIAL; reconciled to filings). Reference price ~$65.81; market cap ~$72–73B. Share counts reconciled to 10-K/10-Q covers and adjusted for the 5-for-1 split.

C. Industry, macro & regulatory

11. Cox Automotive / Manheim Used Vehicle Value Index — 2025 ~+0.4% YoY; 2026 forecast ~+2%. (accessed 2026-06-07)
12. Federal Reserve Bank of New York, Household Debt and Credit Report — auto-loan serious (90+ dpd) delinquency ~5.2% in Q4 2025 (highest since 2010). (accessed 2026-06-07)
13. Mordor Intelligence / IMARC — US used-car market sizing (~38.6M units 2025; ~$850B–$1T; ~66.5% offline; online ~+7%/yr). (accessed 2026-06-07)
14. KBB / CNBC — 2025 auto tariffs (25%+) and impact on new-car prices / demand shift to used. (accessed 2026-06-07)
15. Holland & Knight — FTC CARS Rule vacated by the 5th Circuit (Jan 2025, not appealed). (accessed 2026-06-07)
16. CarMax (KMX) public disclosures — FY (Feb 2025) ~765,000 used units, ~$26.4B revenue, retail GPU ~$2,322, CarMax Auto Finance. (accessed 2026-06-07)
17. Vroom / Shift — public reporting on Vroom’s 2024 used-car wind-down and Shift’s bankruptcy. (accessed 2026-06-07)
18. CNBC — Carvana Q1 2026 earnings coverage (2026-04-29). https://www.cnbc.com/2026/04/29/carvana-cvna-earnings-q1-2026.html
19. Businesswire — Carvana 5-for-1 split announcement (2026-03-13). https://www.businesswire.com/news/home/20260313986313/en/