Record Print, Industry-Leading Margins, Guide Raised — Initiating at Hold on Stretched Valuation

Initial Read. A near-perfect operational quarter that confirms Carvana is the structurally most-profitable retailer in U.S. used cars and that the model levers in both directions; the issue is not the business, it is the price you have to pay for it after a parabolic year.

Results vs. Consensus

A clean, broad-based beat against an already-elevated set of expectations. Magnitudes are largest where the model levers most: unit growth, adjusted EBITDA dollars, and EBITDA margin.

Quality of Beat

• Volume: Pure organic. 41% retail unit growth into a market the company estimates grew <5%; share gain is the entire story. ADESA integrations are facilitating the inventory ramp (+50% YoY).
• Margins: ~$100 of the retail GPU YoY gain is tariff-window pull-forward (March/April); the remaining ~$100 is structural (reconditioning, inbound transport). Other GPU +$126 is largely cost-of-funds + VSC attach, both replicable. SG&A leverage of -$460 per retail unit is overwhelmingly fixed-cost dilution against a 41% volume base — the dominant economic point of the quarter.
• EPS: Operational, not below-the-line. $308M of net income on $511M GAAP operating income; the GAAP-to-EBITDA conversion ratio rose to ~85% from 73% Y/Y, signaling the non-cash overhang is shrinking with scale.

Operational Performance

Key Topics & Management Commentary

Overall Management Tone: Confident, with a consistent “keep marching” cadence. Garcia leaned into the long-term unit and margin goals (3M units / 13.5% adj EBITDA margin in 5–10 years) and was unhurried about each individual lever. Notably, Q&A pushback was minimal — the most pointed exchange (Sharon Zackfia, William Blair, asking for aided brand awareness data) drew a deliberately humorous Garcia non-answer (“of course we do, but we very purposely didn’t provide them”) which itself signals a level of comfort with the story.

Margin Expansion Drivers: How Much is Structural vs. Cyclical

The most important analytical question in the print. Jenkins explicitly bridged the +$195 retail GPU walk: roughly half (~$100) is transitory tariff-related, half is reconditioning + inbound transport. Both Garcia and Jenkins were careful not to claim the tariff piece as recurring.

“That higher April retail GPU really linked to the announcements of auto tariffs in late March that drove stronger demand and higher margins.”
— Mark Jenkins, CFO

Read-through: Underwrite ~$100 of structural GPU improvement Y/Y as the durable piece. The other $100 fades as the comp normalizes. The Q3 print is the first read on whether the underlying trajectory continues without the tariff tailwind.

ADESA Integration & the Reconditioning Capacity Plan

Garcia’s response to Brian Nagel (Oppenheimer) on ADESA cadence was the clearest medium-term roadmap of the call: 12 ADESA sites integrated, ~3 per quarter pace, with full ADESA build-out estimated at ~$1B (caveat: some inflation since the original number) over a multi-year horizon. Capex-light integrations are the front end; full build-outs are the back end. The strategic logic is unchanged from the original ADESA acquisition rationale, but the integrated-site count is now large enough to be meaningfully driving inbound transport efficiency.

“We are currently using Phoenix as a test market to optimize our finance verifications, registration processes, vehicle staging, delivery scheduling systems and staffing models for speed.”
— Ernest C. Garcia III, CEO

Assessment: ADESA is delivering. The inbound-transport mileage reduction and inventory-availability ramp are quantifiable proof-points. The risk is execution at scale — Garcia’s own framing identifies reconditioning as the most operationally complex part of the business, and that’s precisely where the next 18 months of growth has to come from.

The 3M-Units / 13.5%-Margin Goal — What It Implies

Garcia restated the 5-to-10-year framework: ~40% CAGR if hit at 5 years, ~20% CAGR at 10. Q2’s 41% print sits at the fast end. Michael Baker (D.A. Davidson) pressed on whether unit growth needs to accelerate from current levels — Garcia’s answer was carefully agnostic on the timing.

“Growing at 40% for five years is an ambitious target, but we’re an ambitious group, and we’re going to try to get there somewhere between that 5- and 10-year target.”
— Ernest C. Garcia III, CEO

Assessment: The valuation case requires the fast end. The operational case — capacity, inventory, advertising re-investment — is consistent with the fast end. The unanswered question is whether market unit growth (sub-5%) leaves enough share for CVNA to keep the cadence without compressing pricing or reaching credit-quality limits in financing.

Vertical-Integrated Finance Platform & Other GPU

Multiple analysts probed the durability of Other GPU. Jenkins’ framing of the platform’s structural advantages — intimate knowledge of the car (150-point inspection) and the customer (direct interaction vs. indirect finance) — was the most analytically substantive part of the call.

“By vertically integrating finance and other parts of the transaction, it affords a lot of advantages versus a non-vertically integrated model. ... We intimately know the car that we’re selling to the customer. We’ve done 150-point inspection on that, ensured its quality. ... We intimately know the customer because we’re interacting with them directly.”
— Mark Jenkins, CFO

Read-through: The 300–400bps APR premium / 60bps incremental delinquency math (raised by Jeff Lick, Stephens) is the clearest articulation of Carvana’s finance-platform edge. Cost of funds will continue to compound as the buyer base for the company’s loan-sale program expands. This is a real fundamental gain, and it’s genuinely difficult for traditional retailers to replicate without scaling their own loan-sale infrastructure.

Marketing & Brand Awareness

Garcia signaled a deliberate sequential step-up in Q3 advertising. Strategically, the framing was about the gap between total GPU and operating expenses being “very, very large” relative to advertising spend — a way of saying there’s room to invest more without breaking the bottom line. Tactically, this is a margin headwind in Q3 and the reason guide doesn’t imply 12.4% maintained sequentially.

Tariff & Macro Read

Garcia’s read on demand was largely benign — some tariff-related pull-forward and modest follow-on weakness, but no quarter-defining macro shift. The peer-set growth comparison (other public auto retailers ~1% Y/Y unit growth vs. CVNA’s 41%) was repeated as the consistent framing.

Guidance & Outlook

Implied Q3+Q4 EBITDA cadence: $2.0–2.2B FY25 less the implied H1 ~$1.04B (Q1 ~$439M + Q2 $601M, with the Q1 figure inferred from the disclosed Q2 sequential framework) leaves roughly $0.96–1.16B for H2 — consistent with sequential unit growth and a meaningful Q3 advertising bump.

Guidance style: Carvana has a multi-quarter pattern of raising as the year progresses. The midpoint here ($2.1B) is the kind of range a company sets when it wants flexibility to lean into Q3/Q4 advertising without compromising the headline annual number. We’d call it conservative-realistic, not sandbagged.

Analyst Q&A — Notable Exchanges

Q&A skewed toward operational mechanics — capacity, marketing, finance platform, GPU sustainability. Notable threads:

• Daniela Haigian (Morgan Stanley) opened on the >17% incremental EBITDA margin and probed sustainability; Garcia’s 200bps margin gain at 41% growth framing implicitly answered yes. Her follow-up on Avis/Waymo and ADESA’s hub-and-spoke optionality drew a deliberately focused Garcia answer (“part of the sickness” of seeing opportunity everywhere; the answer was “stay focused on the core”).
• Jeff Lick (Stephens) raised the 300–400bps APR / 60bps delinquency spread — the clearest framing of CVNA’s finance edge on the call. Jenkins’ vertical-integration response was the strongest part of the prepared-meets-Q&A material.
• Brian Nagel (Oppenheimer) got the cleanest ADESA cadence framing (12 sites, ~3/quarter) and the explicit framing of the demand environment as “relatively flat” tariff effects.
• Sharon Zackfia (William Blair) pressed for aided brand-awareness numbers; Garcia’s playful refusal (“of course we do, but we very purposely didn’t provide them”) was the lone Q&A friction-point of the call. Read as a deliberate hold-back rather than a dodge — the awareness data is presumably favorable but management isn’t volunteering a specific metric they’d then have to maintain.
• Brad Erickson (RBC) drew the clearest articulation of capacity-expansion phasing (capex-light integration first, then full ADESA build-outs at ~$1B over multiple years).
• Andrew Boone (Citizens) extracted the retail GPU walk that quantified the tariff-vs.-structural split (~$100/$100).
• Rajat Gupta (JPMorgan) probed the “more holistic” framing in the letter; Garcia’s answer was that essentially every line item is a lever, which is operationally accurate but not analytically informative.
• Michael McGovern (Bank of America) got the operational-cost composition breakdown (fulfillment, customer care, T&R, limited warranty) and AI-in-customer-care framing.
• Chris Bottiglieri (BNP Paribas) identified the inventory-build dynamics — sales/selection growth + a marketplace-partner contract change moving inventory onto CVNA’s balance sheet + ASP mix shift — the cleanest Q&A disclosure of the call.
• Michael Baker (D.A. Davidson) pressed on the H2 implied EBITDA deceleration and on the 40% CAGR sustainability question. Garcia’s response was “an ambitious group” framing — deflective but not evasive.
• Chris Pierce (Needham) got the most useful long-term framing on the auto-loan buyer base — the suggestion that CVNA may be expanding the total ABS / pool buyer universe rather than fighting for share of an existing one. If true, this is an under-discussed structural advantage.
• Marvin Fong (BTIG) probed the lower sell-through rate and cost-of-funds runway; Jenkins’ framing positioned the buyer-base expansion + asset-performance feedback as the durable drivers.
• Alex Potter (Piper Sandler) asked the best operating-philosophy question of the call — what could break under sustained 40% growth. Garcia’s answer (reconditioning is most operationally intense; logistics second; market ops third; customer care fourth) is the implicit risk roadmap.
• Ron Josey (Citi) drew the ADESA-vs.-Carvana utilization comparison — ADESA sites are running at less than half the utilization of legacy Carvana sites, which itself runs at less than half facility capacity. The asymptotic catch-up is the medium-term cost-improvement runway.

What They’re NOT Saying

1. No specific aided brand awareness number. Garcia explicitly acknowledged the data exists and was withheld. Read: management is comfortable holding the specific metric back, presumably because (a) it’s favorable but they don’t want to anchor expectations, or (b) it suggests a longer awareness-building runway than analysts would model in. Either way, it’s a deliberate non-disclosure that we’ll watch for in subsequent calls.
2. No quantified Q3 advertising step-up. Jenkins flagged a “larger sequential increase” in Q3 vs. the Q2 increase, but didn’t size it. The implicit margin headwind is real but unsized; we model it as a 100–200bps Q3 EBITDA-margin drag, with full-year guide absorbing it.
3. No detailed cohort data for older markets. Rajat Gupta asked specifically about Atlanta (2013/2014 cohorts); Garcia stuck with “all the trends remain.” The legacy-cohort economics — the case for what mature CVNA markets can earn — would be powerful disclosure if it’s good. The non-disclosure is conspicuous.
4. No 2026 framework. The 5-to-10-year goal was reaffirmed, but no near-term FY26 unit or EBITDA goalposts were teased. This is not unusual for CVNA at this stage, but it leaves the model open-ended past FY25.
5. No update on the long-tail balance-sheet workout. The 2023 debt restructuring is increasingly historical, but the Class A / LLC unit conversion mechanics and warrant overhang still meaningfully complicate the share count math. Management didn’t touch it on this call — not a red flag, but the next conference where they walk it would be useful.

Market Reaction

• Pre-print setup: CVNA had rallied substantially YTD on the back of three consecutive quarters of accelerating margin expansion and continued unit-growth outperformance vs. the public-retailer peer set. Sentiment heading into the print was constructive but elevated.
• Initial reaction: The stock traded sharply higher on the print — reports cited intraday moves of +13% to over +20% on Thursday, July 31, with the print landing after the close on July 30. The combined headline (record print + raised guide) cleared an already-high bar.

The price action is logical given the magnitude of the EBITDA dollar beat ($601M vs. expectations clustered closer to $480–500M) and the FY25 guide raise. What it does, however, is push the multiple back toward the upper end of the historical range — and that is the principal reason we initiate at Hold rather than Outperform. The print itself is unambiguously bullish; the asymmetry at the post-print price is what we’re questioning.

Street Perspective

Debate: Is the 12.4% adj EBITDA margin a new floor or a high-water mark?

Bull view: Operating leverage is just getting started; the SG&A walk shows -$460/unit Y/Y, and most of that is fixed-cost dilution. As the fleet of integrated ADESA sites scales toward Carvana-equivalent utilization, the recon and logistics cost takeout has multiple years of runway. 13.5% is the stated long-term goal; on this trajectory, it could come in faster than the 5–10 year window.

Bear view: Q2 included ~$100 of transitory tariff-related GPU benefit and was lapping easy comps; Q3 will absorb a larger advertising step-up that compresses incremental margins. ADESA build-out capex of ~$1B+ over coming years pressures FCF even as EBITDA holds. Variable-margin compression as ADESA sites continue to ramp at sub-half-utilization is mechanical.

Our take: The structural piece of the GPU gain is real; the cyclical piece will fade. Margins likely settle in a 10–12% adj EBITDA band over the next 12 months as advertising re-investment + ADESA dilution offset some of the legacy-IRC efficiency. That’s still industry-leading and consistent with the long-term trajectory — just not 12.4% sustained.

Debate: Can CVNA hold 40% retail-unit growth into 2026?

Bull view: At ~1.5% U.S. used-car market share with a structurally differentiated offering, the unit math says yes — the constraint is operational scaling, not demand. ADESA integrations are unlocking the inventory and transport capacity to support it. Word-of-mouth flywheel + advertising re-investment + same-day-delivery capability set up a multi-year runway.

Bear view: Comps get progressively tougher; Q2 24 was already a strong quarter to lap. Macro / used-car demand sub-5% growth means CVNA must keep taking share to keep growing — possible, but not guaranteed. Reconditioning is the binding operational constraint, and the next 12–18 months are the test.

Our take: 40% sustains for one or two more quarters comfortably; the binding question is the FY26 number. We’d underwrite something in the 25–35% range for FY26 vs. the consensus that may already be at the high end of that. Not a thesis-breaker, but the magnitude of the 2025 outperformance is unlikely to repeat in 2026.

Debate: Has the stock priced in the long-term goal already?

Bull view: Compounding 30–40% units with margins expanding 100bps+ a year creates an EBITDA trajectory that justifies a premium multiple. Investors who held through the 2023 distress are appropriately compensated for accepting the structural-value optionality back when nobody else would.

Bear view: The current multiple already prices the 3M units / 13.5% adj EBITDA scenario at the fast end. Any execution stumble — reconditioning miss, GPU normalization quarter, FY26 unit deceleration — recalibrates the multiple sharply lower. The asymmetry has compressed.

Our take: This is the correct framing for the rating. The business is exceptional; the price you pay for it is the variable. We initiate at Hold because the post-print level prices the upside scenario without enough margin of safety for the operational risk inherent in five more years of 40% growth.

Model Implications

• FY25 adj EBITDA: Anchor at the midpoint of the new $2.0–2.2B guide ($2.1B); we expect a Q4 update toward the higher end if Q3 unit growth holds.
• FY25 retail units: Implied roughly 530–560K based on Q2 trajectory and Jenkins’ sequential growth language; consistent with the prior FY24 baseline plus the disclosed cadence.
• Retail GPU: Underwrite ~$100 of structural Y/Y gain as durable; the tariff-window piece (~$100) fades by Q4 as we lap normalized periods.
• Advertising: Q3 step-up implies ~$25–40M sequential increase based on the “larger than Q2 sequential” framing; full-year advertising spend at ~5% of revenue.
• Capex: ADESA full build-out of ~$1B+ phased over multiple years; incremental capex per integrated site remains light until full build-out begins.
• Balance sheet: Net debt to TTM adj EBITDA is improving; on $2.1B FY25 EBITDA, leverage drops further. Refinancing optionality is back; investment-grade is a multi-year aspiration but no longer fanciful.

Thesis Scorecard

Overall: Operating thesis confirmed; valuation thesis is the rating gate. The business is structurally advantaged and the model is delivering above any reasonable bull-case extrapolation from a year ago. The price you pay for that performance is the variable that determines the rating.

Action: Initiating at Hold. We are not selling into the post-print rally; we are simply waiting for either a meaningful pullback (call it 15–25% from the post-earnings level) that re-creates margin of safety, or one to two more quarters of execution at this trajectory before the underwriting case for Outperform improves. Holders should size CVNA appropriately given the elevated implied volatility on the name — +13–22% and -25%+ moves on print are now both within the range of plausible outcomes.

Net: Excellent quarter, deserved rally, but the asymmetry has compressed enough that we’d rather pay attention than pay up.