A Record Quarter Trips Our Own Per-Unit Trigger: Downgrading to Hold Into a 16% Rally

Results vs. Consensus

The scorecard shows beats on units, revenue, total GPU and adjusted EBITDA, and misses on wholesale GPU and EPS. The composition matters more than the headline: the dollar beat came from volume and a richer retail mix, while per-unit economics declined both year over year and sequentially.

A note on EPS. The $0.39 diluted Class A figure is on the post-split basis that began with Q2 (the 5-for-1 split took effect May 7, 2026), and the $0.21 prior-year comparative is already restated. The quarter carried a one-time $53M pre-tax loss on extinguishing the 2030 senior secured notes. At the quarter's 12.5% effective tax rate and the share of net income attributable to Class A holders, that charge cost roughly $0.04 per diluted share. Excluding it, EPS would have been about $0.43, still a penny short of the most common consensus figure. The small remainder reflects SG&A growing faster than units sequentially, which we treat in detail below. As in prior quarters, valuation math uses the 1.127 billion shares that would be outstanding on full conversion of LLC units, not the 740 million diluted Class A shares.

Year-over-Year Comparison

Per-Unit Economics: Year over Year and Sequential

The year-over-year comparison shows where per-unit profit went. The sequential comparison shows that the slippage is still happening, and it moved from gross profit into SG&A.

Revenue assessment. This is a clean revenue line. The accounting distortion that inflated Q2 growth to 52% has lapped as promised, and what remains is genuine volume plus a mix shift toward newer, higher-priced cars. Unit growth accelerated against an industry that the company describes as approximately flat, which is share gain at the rate the thesis needs. The one soft spot is wholesale revenue per vehicle, down 15.9%, which management did not address.

Margin assessment. At Q2 we decomposed a 199bps adjusted EBITDA margin decline into roughly 111bps of revenue-denominator inflation and 88bps of real per-unit erosion, and argued the denominator half would disappear in Q3. It did: total revenue per retail unit rose only 2.1% this quarter. The margin still fell 133bps, to 10.0%. With the denominator gone, nearly all of this quarter's compression is per-unit economics, and that is the line our thesis needed to turn. Management attributes the decline to pricier mix and the customer rate giveback, and estimates fuel and rising benchmark rates cost about $200 per unit. That bridge is credible but it is not new: Q2's bridge also leaned on fuel and benchmark rates, and the macro portion of the explanation has grown, not shrunk.

EPS assessment. The GAAP picture is better than the EPS headline. GAAP operating income converted at 89.3% of adjusted EBITDA, up from 86.7% a year ago, because depreciation is flat while EBITDA grows. Net income margin rose to 6.0% from 4.7%. Retained earnings reached $840M from $551M at Q2. The refinancing charge explains about $0.04 of the $0.05 miss against $0.44. The GAAP view matters more than usual, because from 2027 the GAAP lines become the company's primary reporting metrics. On that basis, operating income per unit fell $255 year over year, a smaller decline than adjusted EBITDA per unit, but still a decline.

The Per-Unit Question

This section exists because the per-unit trend is the reason we are changing the rating on a quarter in which almost everything else went right.

Four consecutive year-over-year declines, each larger than the last. At Q2 we wrote that two specific distortions explained part of the decline and that both would lift in Q3: the revenue-denominator inflation from a marketplace partner's gross-revenue treatment, and reconditioning costs that had not yet fully normalized through cost of sales. Both lifted. Non-GAAP retail GPU rose $85 year over year, and management attributed it to wider retail-to-wholesale spreads with no reconditioning drag mentioned. Total revenue per retail unit rose only 2.1%. The decline widened anyway.

The year-over-year bridge is simple. Non-GAAP GPU fell $410 and non-GAAP SG&A per unit was flat, so adjusted EBITDA per unit fell $405. Within GPU, retail rose $85, wholesale fell $196 and Other fell $299. Management named the causes plainly: Other GPU fell because of "our decision to give back to customers in the form of lower interest rates relative to benchmark rates as well as by rising benchmark rates," and wholesale GPU fell because retail unit growth outpaced wholesale gross profit growth. The first is a deliberate, ongoing choice. The second is structural for as long as retail outgrows the wholesale marketplace. Only the benchmark-rate component is plausibly cyclical.

The company's own two-year comparison is the strongest counterargument, and it deserves a fair statement. The letter says Q3 2024 was helped by about $250 per unit of loan-sale timing and falling rates, and Q3 2026 was hurt by about $200 per unit of fuel and rising rates. Adjusting for both, Q3 2024's $3,948 becomes about $3,698 and Q3 2026's $3,680 becomes about $3,880. That supports the claim of "consistent unit economics" over two years while the business doubled. Two caveats limit it. The comparison explicitly excludes the wholesale marketplace, which is the line that is diluting. And "consistent" is not "expanding," which is what a move from roughly 10% to 13.5% adjusted EBITDA margin requires.

*At Q3's 217,385 units, which is the floor because Q4 units are guided up sequentially; any higher unit count lowers the per-unit figure. Nine-month adjusted EBITDA is $2,241M. Q3-to-Q4 adjusted EBITDA fell 16.3% in 2024 and 19.8% in 2025.

Our take. The guide moved up and the market celebrated, correctly, because it settled the Q2 debate about whether $2.7–3.0B was a floor or a forecast. But the same arithmetic says Q4 per-unit profit is roughly flat year over year at the upper-half midpoint, against Q4 2025, a quarter that was itself depressed by the reconditioning stumble. A Q4 decline in line with the last two years (16–20% from $800M) would land in the lower part of the upper half, not the top. A fifth consecutive per-unit decline is therefore a live possibility, not a tail. Management also now says units versus margin is "largely a choice," and that it prefers units as long as the supply chain can scale. We believe it. That is a coherent strategy for building a 3-million-unit business, and it is the reason we no longer expect the per-unit inflection inside our 12-month horizon.

Operating and Unit Economics

Segment Performance

Retail Vehicle

Retail revenue of $6.144B grew 53.8% on 39.4% more units, with revenue per retail unit up 10.3% to $28,263. Retail gross profit of $773M grew 43.4%, and retail GPU of $3,556 rose $100 year over year and $9 sequentially, $157 above the Street. Management credited wider spreads between retail and wholesale market prices, partly offset by higher non-vehicle costs, including inbound fuel. This is the line where Q4's reconditioning stumble did its damage, and it is now clearly repaired.

"I think if we look on a year-over-year basis, I think we break it down as we see retail to wholesale spreads up year-over-year, partially offset by higher non-vehicle costs; some of those higher non-vehicle costs were driven by inbound fuel costs." — Mark Jenkins, CFO

Assessment: Retail GPU held Q2's level, which included an FTC-driven pricing tailwind that management had said was largely spent. Holding it with no windfall to lean on is a better result than the sequential gain looks. Retail is not where the per-unit problem lives.

Wholesale

Wholesale revenue of $1.302B grew 10.6% on 38.3% more wholesale units. Wholesale vehicle gross profit per wholesale unit fell 6.8% to $963, and wholesale GPU per retail unit fell $158 to $708, $96 below the Street. This is the second consecutive quarter in which wholesale per-unit profit missed by roughly $100 or more. Management gave a structural explanation: retail unit growth is outpacing wholesale gross profit growth, and the letter's two-year comparison excludes the wholesale marketplace because it "does not scale at the same rate as our retail offering."

Assessment: This is an admission that the wholesale marketplace is a per-unit dilutor by construction while retail grows near 40%. Wholesale gross profit dollars still grew 14.1%, so the franchise is not failing. But measured per retail unit, the way the company reports it, this line will keep falling as long as the growth plan works. We keep the ADESA Clear pillar at risk and now treat wholesale as a steady per-unit headwind in our model.

Other (Finance and Ancillary)

Other revenue and gross profit of $594M grew 25.3%, well below unit growth of 39.4%. Other GPU fell $307 year over year to $2,733 but rose $67 sequentially, $36 above the Street. Management attributed the year-over-year decline to the late-2025 decision to give back more than 100 basis points of rate to customers, plus benchmark rates that rose through the year, partly offset by a higher average amount financed and hedging gains. A new cash-flow hedge appears in other comprehensive income this quarter ($13M), and nine-month gain on loan sales rose to $1,084M from $878M.

"often premiums go down, but they're applied against a bigger balance." — Ernie Garcia, CEO

Assessment: Sequential stability in a quarter of sharply rising benchmark rates is a genuine positive and suggests the giveback is now in the base rather than still widening. The finance platform pillar holds. But the giveback is permanent by design, and management's description of its rate posture leaves room to absorb more benchmark increases rather than pass them through in full.

Key Topics & Management Commentary

Overall Management Tone: Assured and strategic, with a deliberate turn from quarterly mechanics toward the long-term goal. Unlike Q2, when management drew hard lines around the guide, most questions drew a substantive answer. The exceptions were new-car expansion and the timing of per-unit expansion, which got only a long-term answer. The answers were consistently framed against the 2030–2035 target rather than the next quarter, and that framing was most visible when the question was about near-term per-unit margins.

1. Units Over Margin, Stated Out Loud

The most consequential sentence on the call came in answer to a question about the FTC's pricing guidance. Asked whether the benefit would show up in units or in pricing power, the CEO named a preference, with one condition.

"I think whether it shows up in units or margin, I think, is largely a choice. ... And we -- given where we are and given how clear the path is to 13.5%, I think that we would prefer for that value to show up in unit growth, but that requires that we continue to scale the supply chain as quickly as possible." — Ernie Garcia, CEO

The sentence explains the last four quarters. The rate giveback, the advertising step-up and the willingness to sell richer-mix cars to keep inventory moving are all the same choice. It also means that the operating leverage we were waiting to see in per-unit profit is being spent on growth on purpose.

Assessment: We think the choice is right for the long-term value of the business and wrong for our 12-month rating. An Outperform needs a catalyst inside a year. If management keeps choosing units, the per-unit inflection moves beyond that horizon, and the stock is left to rerate on growth alone from about 27x EBITDA.

2. The 2027 Reporting Shift: GAAP Metrics and a 12.5% Operating Margin Target

The letter announced that from 2027 Carvana's primary reporting moves from non-GAAP GPU, non-GAAP SG&A and adjusted EBITDA to their GAAP counterparts, including operating income and net income. The mid-term goal becomes 3 million retail units at a 12.5% operating margin between 2030 and 2035, which the company calls equivalent to the 13.5% adjusted EBITDA target.

"I think it's a natural step in our evolution as a company. I think adjusted EBITDA has been an important metric for tracking our progress over the past many years. I think the -- as we become more profitable on a GAAP basis, I think the value of adjusted EBITDA goes down." — Mark Jenkins, CFO

The equivalence nearly checks out on current numbers. GAAP operating income was 89.3% of adjusted EBITDA this quarter, and 89% of 13.5% is about 12.0%, so 12.5% assumes conversion keeps improving as depreciation is levered. The quarter's operating margin was 8.9%, leaving a 360bp gap to the target.

Assessment: A clear positive for quality of earnings. Adjusted EBITDA excludes depreciation, stock compensation and interest, all of which are real for a capital-intensive retailer, and the company is volunteering to be measured on the stricter line. Two cautions follow. The new metric hides the per-unit trend we track less well, because operating income per unit fell $255 this quarter against $405 for adjusted EBITDA. And a change of measuring stick during a stretch of per-unit decline deserves to be read carefully, even when the reasons are sound.

3. No More Financial Outlook

Beginning with the Q4 2026 call, Carvana will not provide quarterly or annual financial outlooks, and will focus forward-looking communication on its longer-term goals. The company has historically given a sequential unit direction each quarter and a full-year adjusted EBITDA range at mid-year. Both go away.

The timing is notable. The Q2 guide caused a 7.4% decline, and the upward revision of the same guide caused a 15.7% gain. The guide has been the single largest driver of the stock's earnings-day moves for two quarters running, and management is removing it.

Assessment: Defensible for a company at this scale, and consistent with a management team that has always preferred long-term targets to quarterly commitments. For investors it raises the cost of being wrong between prints. We add a new bear point on visibility. Without a guide to anchor to, per-unit economics, which management will not forecast, become the main quarterly signal.

4. Production Outruns Sales; Listed Inventory Still Lags

At Q2 the binding constraint was reconditioning throughput: inventory had undergrown sales and capped conversion. This quarter brought measurable progress on production and limited progress on the inventory customers actually see.

"We set a record for units produced in the quarter. It was our first quarter producing more than 200,000 units in a single quarter. That's a nice milestone. Our year-over-year growth in production also accelerated slightly relative to Q2, which I think is the direction that we want to head, and we slightly outgrew retail units sold from a -- or we slightly outproduced retail units sold in Q3, which is also definitely a direction that we want to head." — Mark Jenkins, CFO

In the same answer, the CFO added the qualification that matters for our trigger: growth in the number of cars on the website immediately available to customers "is still up year-over-year at a rate that is much below what our sales growth is year-over-year." Balance-sheet inventory of $3.456B rose 5.9% sequentially against 10.2% unit growth, and 49.2% year over year in dollars, a figure inflated by the shift to pricier cars.

Assessment: The flow has turned and the stock has not. Producing slightly more than you sell is the first step toward rebuilding selection, and the unit acceleration suggests conversion improved even with thin inventory. But our Q2 trigger was inventory growth at least matching sales growth, and by management's own account it did not. We score that trigger as missed, with the direction improving.

5. Fuel and Rates: About $200 per Unit

The letter estimates that elevated fuel prices and rapidly rising benchmark rates reduced unit economics by about $200 per unit in Q3. At Q2 management put fuel at about $75 per unit across the income statement and cited higher benchmark rates without a number. Logistics expense rose 58.1% year over year to $68M, and operations expense rose $109 per unit, which the CFO attributed mainly to fuel and logistics.

"This quarter, it was pretty significant moves in gas prices and in interest rates, which impacted us to the tune of about $200 per unit. But in the grand scheme of things, those things tend to zero out and we're just kind of steadily on this march." — Ernie Garcia, CEO

Assessment: We accept that fuel and benchmark rates are cyclical and will eventually reverse. We note that the macro explanation has now been offered for two consecutive quarters, at a larger size each time. We move the macro bear point from contained to emerging. Even adding back the full $200, adjusted EBITDA per unit would have fallen about $205 year over year.

6. SG&A: The Advertising Step-Up Arrived

Management said at Q2 that it would raise advertising dollars in Q3 to offset thin inventory, and it did. Advertising rose to $152M from $121M, a 25.6% sequential increase on 10.2% more units, taking advertising per unit to $699 from $613 at Q2 and $654 a year ago. Total SG&A rose 14.3% sequentially. Compensation and benefits grew only 7.6%, so overhead is still levering. The CFO separated the long-run picture from the near term.

"advertising expense is a place where not necessarily in the near term, but certainly in the longer term, see opportunities for meaningful advertising leverage as we continue to raise awareness, understanding and trust, build out our selection, improve speeds, et cetera, enhancing the customer offering." — Mark Jenkins, CFO

Assessment: The phrase "not necessarily in the near term" carries weight here. Advertising leverage is one of the two named routes to the 13.5% target, along with overhead leverage, and management is saying it will not arrive soon. Overhead leverage alone ($159 per unit year over year) was fully spent this quarter on operations and advertising.

7. Mix Moves Further From the Traditional Car Buyer

At Q2 management called the shift to newer, pricier cars "definitely not structural" and said it expected metrics to move back toward the average car buyer. In Q3 retail revenue per unit rose again, to $28,263 from $27,908 at Q2, and the CFO said the adjusted EBITDA margin decline was "primarily driven by selling a mix of higher-priced vehicles and giving back to customers in other GPU." The CEO described the mechanism directly.

"The most notable area where that's occurring, right now, is in prime where we've seen our prime mix really shift up as we've focused on producing more expensive later-model cars as we try to catch up." — Ernie Garcia, CEO

Assessment: The Q2 commitment to revert mix was not met. A richer mix raises revenue and retail GPU dollars but lowers margin percentages and Other GPU premiums, so it flatters the top line and depresses the ratios we track. Mix should normalize as production catches up. That is the same "as production catches up" condition that underlies most of this quarter's explanations.

8. The Balance Sheet: 0.8x and a Refinanced Capital Structure

Early in the quarter Carvana refinanced the remaining $1.7B of senior secured notes due 2030 with a senior secured Term Loan B, for about $45M of annual interest savings and a one-time $53M loss on extinguishment. Net debt to trailing adjusted EBITDA reached 0.8x and trailing adjusted EBITDA to interest reached 6.9x, both company records. Liquidity rose to $7.503B and nine-month free cash flow reached $734M.

"This transaction, along with the successful execution of our operational plan in Q3, resulted in a net debt to trailing 12-month adjusted EBITDA ratio of 0.8x and a trailing 12-month adjusted EBITDA to interest expense ratio of 6.9x, both new company records and again placing us in our strongest financial position ever." — Mark Jenkins, CFO

Assessment: The balance-sheet pillar is fully delivered. Leverage is no longer a variable in the investment case, and the remaining question is how fast credit ratings follow the ratios toward investment grade.

9. Capacity: 21 Sites, Three Full Buildouts, 58% Utilization

Two more ADESA locations gained integrated retail production, bringing the total to 21, and construction began on two more full buildouts, for three under construction, all expected to begin production in 2027. Built-out capacity stays at about 1.5 million retail units, with real estate for 3 million. At an 870,000-unit run rate, the company uses about 58% of its built-out capacity and is 3.45x from the 3-million-unit goal, down from 3.8x at Q2.

At Q2 the first full buildout was described as producing in early 2027. The letter now places all three in 2027 without a quarter. Asked whether the full conversions would bring inefficiencies, management confirmed that new facilities run at higher cost until they mature, while saying this year's management and systems work leaves it comfortable bringing sites on "at a reasonable level of efficiency."

Assessment: The physical runway is intact and growing. The new-site cost point matters for our per-unit view: three full facilities ramping in 2027 add another start-up cost layer to a year in which we already expect advertising to stay elevated.

10. Agentic AI: "We Would Build Carvana"

The CEO used the opening remarks to argue that agentic commerce favors a vertically integrated retailer whose decisions are automated and deterministic. In Q&A, the CEO described a future in which a customer's agent shops, values a trade-in, applies for credit and schedules delivery in one exchange with Carvana, while the same agent exposes every dealer fee and add-on in a traditional purchase.

"we think that it also, as a general matter, has kind of the capacity to effectively reduce digital friction to 0 while leaving physical friction largely unchanged." — Ernie Garcia, CEO

Assessment: The argument is logically sound and, unlike most AI commentary on earnings calls, tied to a specific structural advantage: pricing and credit decisions that are already automated and exposed through a single system. It is also unquantified and early. One analyst noted that today's chatbots still surface competitors often, and the answer was mostly about scale. We note it as a long-run tailwind and give it no value in the model.

11. New Cars: Stellantis Goes Live

The letter disclosed that new vehicles from Carvana's seven Stellantis franchise dealerships are now listed on Carvana.com, and used it as the example of platform leverage: the customer base, merchandising, financing, checkout, fulfillment and title workflows already existed. Asked about expansion to other manufacturers, the CEO declined to add detail beyond "The economics work."

Assessment: This is a more concrete disclosure than Q2's qualitative one. It confirms that the optionality is real and running on shared infrastructure at low incremental cost. It remains too small and too undisclosed to model.

12. Competition and Market Position

Asked whether a larger competitor's improving growth was visible in Carvana's results, management said it could not identify any competitor's effect and described its own position in new terms.

"I think for a long time, we've been the second largest seller of used cars in the country. I think we're knocking on the door of becoming the largest." — Ernie Garcia, CEO

Assessment: Growing 39% in a flat industry is the clearest evidence that no competitor is taking share from Carvana at present. The "largest" milestone is symbolic, but it frames the next phase: once Carvana is the market leader, growth comes from expanding the online share of the market rather than from taking share from a weaker rival.

Guidance & Outlook

The outlook comes with the qualifier "as long as the environment remains stable," the same wording as the Q2 guide. The CFO framed the year as a "strong finish," and the CEO answered the question about margin expansion from Q4 by restating the long-term path rather than giving a near-term view.

Implied Q4 ramp: With $2,241M of adjusted EBITDA through nine months, "higher end" implies Q4 adjusted EBITDA of roughly $609–759M, or 19–49% above Q4 2025's $511M. The midpoint of that band ($684M) is a 14.5% sequential decline, which is milder than the 16.3% and 19.8% Q3-to-Q4 declines of the last two years.

Street at: Published FY26 adjusted EBITDA estimates sat around $2.93–2.95B before the print, so the guide moved up to meet the Street rather than above it. Next-quarter EPS consensus sits near $0.35.

Guidance style: Conservative, as we argued at Q2. The mid-year range was treated as a floor and is now being raised within its own bounds. This is the last time investors will receive this kind of guide, which makes the Q4 print a clean test of the practice before it ends.

Analyst Q&A Highlights

Q&A was cooperative and wide-ranging, touching rates, SG&A, AI, the FTC, new cars and competition. The two exchanges that mattered most for the rating were the direct question on when per-unit profit expands and the question on whether the FTC benefit lands in units or margin. Both got long-term answers.

When does per-unit profit start expanding again?

The most direct question on the call asked whether Q4 would mark a return to margin or per-unit expansion, with the questioner noting it might be the last chance to ask under the old guidance practice. Management answered with the long-term target and the $200 macro estimate rather than a near-term view.

Q: "I wanted to ask about just EBIT margin expansion or just EBIT per unit going forward. I mean, last few quarters, there have been a lot of different type of margin headwinds, fuel costs, benchmark rates, reconditioning issues. Do you think starting fourth quarter, you should be back to margin expansion or EBIT per unit expansion for the company, assuming all else equal, just curious to get your thoughts there."
— Rajat Gupta, JPMorgan

A: "I would put that question back in kind of the context of our goal of 13.5% EBITDA margin leverage, which kind of maps to 12.5% operating income margins. ... We can see our way very clearly to 13.5%. And we plan to get there slowly but surely continually over time."
— Ernie Garcia, CEO

Assessment: The second consecutive quarter in which a direct invitation to date the per-unit inflection was declined. "Slowly but surely continually over time" is a commitment to a destination, not a schedule. Combined with the guide arithmetic, it supports our view that Q4 is unlikely to show per-unit growth.

How the company is handling rising benchmark rates in Other GPU

The opening question asked whether the 100bp customer giveback continues and whether rising benchmark rates are being passed through. Management separated the strategic giveback from the cyclical rate response and said pass-through is happening with a lag.

Q: "Are you a -- continuing to talk about the 100 basis points of improvement to consumers? Are you passing along the full increase in the benchmarks? Or how should we kind of think about how you're approaching that?"
— Marvin Fong, BTIG

A: "benchmark rates continued to rise meaningfully over the course of the quarter. But we didn't see a meaningful change in other GPU quarter-over-quarter, in part because we had sort of started the motion on passing some of those benchmark rates through."
— Mark Jenkins, CFO

Assessment: Credible, and consistent with the +$67 sequential Other GPU move. The answer also confirms the giveback is a standing decision rather than a promotion, so the year-over-year Other GPU comparison is likely to stay negative until the giveback laps in late 2026.

Why growth accelerated while the market slowed

A question asked management to rank the causes of the acceleration, given that industry data and the SAAR were both down in the quarter. Management put supply-chain scaling first by a wide margin and connected it to the prime-customer mix shift.

Q: "your units were up 39.4%, which is actually an acceleration, whether it's a 1-, 2-, 3-year basis. And that's in an environment where whether it's NADA or Cox, the used car business was actually down and SAAR was down in the quarter as well. I was wondering if you could maybe just force rank the factors as to why you think the acceleration might have happened."
— Jeff Lick, Stephens

A: "our general premise and kind of belief over a multiyear period is that our rate of growth is constrained by our ability to scale the supply chain. And so we think that is by far and away the biggest driver."
— Ernie Garcia, CEO

Assessment: Consistent with Q2's regional data, which tied sales growth to production capacity almost one for one. If supply is the constraint and production just accelerated, unit growth near 40% should hold into 2027. That is why we move the growth-sustainability bear point to contained.

Units or margin: where the FTC tailwind will land

A question asked whether the FTC's fee-disclosure guidance benefits units or pricing power. The answer was the clearest statement of capital-allocation priority on the call.

Q: "Do you feel like your position is strengthened by this? And do you view this as a benefit more to units or maybe even your ability to push ASP and maybe GPU higher?"
— John Healy, Northcoast Research

A: "I think whether it shows up in units or margin, I think, is largely a choice. ... And we -- given where we are and given how clear the path is to 13.5%, I think that we would prefer for that value to show up in unit growth, but that requires that we continue to scale the supply chain as quickly as possible."
— Ernie Garcia, CEO

Assessment: The single most important exchange for the rating. Management is telling investors that incremental value will be taken as volume wherever the supply chain can carry it, and as margin only where it cannot. That is a long-term positive and a near-term reason to expect per-unit margins to stay flat or down.

Why move to GAAP reporting now

A question asked why the company is shifting its reporting emphasis to GAAP operating and net income in 2027. The answer framed it as a sign of maturity and pointed to large technology companies that do not report adjusted EBITDA.

Q: "I was curious the note on 2027 reporting, could you speak to kind of why now is the right time for that change, just focusing more on GAAP earnings than non-GAAP earnings, maybe focusing more on GAAP operating income than EBITDA, for example?"
— Michael McGovern, Bank of America

A: "I think it's a natural step in our evolution as a company. I think adjusted EBITDA has been an important metric for tracking our progress over the past many years. I think the -- as we become more profitable on a GAAP basis, I think the value of adjusted EBITDA goes down."
— Mark Jenkins, CFO

Assessment: A good answer to the question asked. The question not asked was about the decision to stop giving an outlook, which drew no direct question on the call.

Start-up costs at the full ADESA buildouts

A question asked whether the coming full buildouts would carry inefficiencies into the P&L, and a follow-up pressed on the people side. Management conceded new facilities run at higher cost and pointed to this year's investment in manager reporting, AI-built operating visibility and leadership development.

Q: "As you kind of embark upon the full ADESA conversions or build out, are there any inefficiencies that we should expect to kind of see hit the P&L?"
— Sharon Zackfia, William Blair

A: "those facilities tend to have higher costs, other things being equal, than facilities that are at a greater level of maturity and scale. ... There's always more to do there, but I think we've made gains that leave us feeling good about our ability to bring on new sites at a reasonable level of efficiency."
— Mark Jenkins, CFO

Assessment: An honest answer, and a useful one for modeling 2027. Late 2025's facility additions caused the reconditioning cost stumble that hit Q4 2025. Three larger facilities ramping in 2027 are the same risk at a larger scale, against better management infrastructure.

Whether AI agents erode the finance margin

A question raised the risk that AI agents shopping for lower rates could raise competition and prepayments in auto finance. Management accepted the premise for a static analysis and argued that markets adapt, for example through fees, and that vertical integration keeps Carvana's cost advantage.

Q: "could agents hypothetically drive higher competition proactively searching for lower rates, either at competitors or third-party financing options? And how much does your moat in financing help with that risk going forward?"
— Michael McGovern, Bank of America

A: "it's reasonable to assume there will be things like higher prepayments. ... I think that it is unclear what the sum of all of the market evolutions will be, but it is unlikely to be as dramatic as the comparative static kind of approach might cause you to think."
— Ernie Garcia, CEO

Assessment: The most candid risk acknowledgment of the call. Higher prepayments would lower the value of loans Carvana sells and therefore Other GPU. This is a slow-moving risk rather than a near-term one, but it names a specific mechanism by which AI could cost the company money.

Whether a stronger competitor is visible in the results

A question asked whether a larger competitor's improving growth was affecting Carvana. Management said it could not detect any individual competitor in its results.

Q: "you do have what I view as a significant or a larger competitor out there now that's starting to show much better growth, still nowhere as near to what Carvana is putting up. The question I want to ask is as you're seeing this dynamic change, are you -- is there more -- are you seeing anything in your business that you have a stronger competitor now reemerging?"
— Brian Nagel, Oppenheimer

A: "So I think we have not seen anything in particular from any given competitors that we believe has had any notable impact on our results."
— Ernie Garcia, CEO

Assessment: Unit acceleration in a flat market supports the answer. Competition is not part of our case for the downgrade.

What They're NOT Saying

1. A date for per-unit expansion. Asked directly for the second quarter running, and answered with the 2030–2035 target. Management will discuss the destination but not when per-unit profit starts rising.
2. Why the outlook is being withdrawn now. The decision to stop giving quarterly and annual outlooks drew no direct question and no explanation beyond the letter's link to long-term goals. The guide has driven the last two earnings-day moves.
3. The listed-inventory number. Management said growth in retail-ready listed cars is "much below" sales growth but gave no figure. One question cited roughly 30% inventory growth and management did not confirm or correct it.
4. Wholesale revenue per vehicle. Down 15.9% year over year with no explanation. Wholesale GPU missed estimates by roughly $100 or more for the second straight quarter, and the only commentary was that the marketplace does not scale with retail.
5. A full-year retail unit figure. Five consecutive quarters without one. With the outlook being withdrawn, there may never be one.
6. Mix reversion timing. The Q2 commitment to move back toward the average car buyer was not met, and no new timeline was given.
7. A capex figure. Nine-month capex was $166M, up from $96M, with three full buildouts now under construction. No annual or 2027 figure has been provided.
8. Related-party balances. Other current liabilities due to related parties rose to $79M from $31M at year-end. The release lists the amount with no explanation, and the call did not address it.

Market Reaction

• Pre-print setup: CVNA closed at $59.05 on October 28, down 30.0% year to date against the S&P 500's 13.1% gain, down 18.5% over twelve months and down 2.3% over 30 days. The 52-week closing range was $56.26 to $95.69, so the stock entered the print about 5% above its low and below the $61.43 close that followed the Q2 guide.
• After-hours move: Shares rose roughly 8% in extended trading, to about $63.99, as the letter's outlook crossed.
• Reaction session (October 29): Opened at $65.28, a 10.6% gap. Traded between $63.81 and $69.23 (+8.1% to +17.2%). Closed at $68.31, up 15.7%, near the session high.
• Volume: 23.2 million shares against a 30-day average of 8.3 million, a 2.8x multiple.
• Relative: The S&P 500 fell 0.3% in the same session, so the relative move was about 16 points.

The guide moved the stock both ways. In July the first FY26 guide sent the stock down 7.4%. In October an upward revision of the same guide, plus a clear unit beat, sent it up 15.7%. The stock now sits 11.2% above its post-Q2 close and roughly 3% above where it traded before the Q2 print. Over two quarters the market has round-tripped the guide scare. In doing so it has paid for the growth half of the thesis, which delivered, without pricing in the margin half, which did not.

A de-rated setup met a strong number. The stock entered the print near its 52-week low after underperforming the S&P 500 by 43 points year to date, which left it exposed to a sharp move on any upside surprise. The intraday path, a 10.6% gap that extended through the session to close near its high, shows buyers adding through the day rather than fading the open. That shape matches the Q2 reaction in reverse, when the stock closed at its high after an 11% gap down.

What the tape ignored. Nothing in the reaction reflected the fourth consecutive per-unit decline, the EPS miss or the withdrawal of future guidance. That is the gap between our reading and the market's. We think the market priced the quarter as a de-risking event. We think it was a de-risking of growth and a deferral of margin, and the 15.7% move used up much of the valuation cushion that justified Outperform at $61.43.

Street Perspective

Debate: Is the per-unit decline a choice or a ceiling?

Bull view: Carvana is spending margin to take share, and at 39% growth in a flat industry with record adjusted EBITDA dollars, investors should let it. The decline is explained by a deliberate rate giveback, about $200 of fuel and rate headwinds, and a wholesale marketplace that simply does not scale with retail. Adjusted for the company's stated distortions, per-unit profit is higher than two years ago at double the volume.

Bear view: Four straight declines, each larger than the last, is a trend regardless of the reasons. The operating margin fell 90bps to 8.9% and sits 360bps below a target the company just restated. The EPS miss came without operating leverage: SG&A grew in line with units. Every quarter has a new explanation and the line keeps falling.

Our take: Mostly a choice, and that is the problem for a 12-month rating. We accept the bull decomposition: very little of the decline looks like competitive erosion. But management has now said it prefers to take incremental value as units wherever supply allows, and it is spending to make supply allow it. A ceiling would be a reason to sell. A choice is a reason to stop expecting per-unit expansion on our timetable, which is a reason to hold.

Debate: Does withdrawing guidance matter?

Bull view: Large, profitable growth companies rarely give quarterly guidance, and moving to GAAP metrics and long-term targets is a sign of maturity. The guide created two quarters of volatility around a number that management had consistently beaten. Removing it lets investors focus on the multi-year compounding.

Bear view: The guide was the main tool investors had to judge near-term earnings power, and it disappears in a stretch when per-unit economics are falling and fuel and rates are moving against the business. A company comfortable with its trajectory has less reason to stop sharing it.

Our take: Neutral on intent, negative on visibility. We see no sign the withdrawal is defensive; it fits this management team's long-standing preference for long-term targets, and it arrives with a more conservative accounting basis. But it removes one input to our quarterly assessment, so we need more margin of safety in the price to hold an above-market rating.

Debate: Is valuation still undemanding after the rally?

Bull view: Even after the rally the stock is down about 19% year to date and nearly 29% below its 52-week high. At about 21x our 2027 adjusted EBITDA estimate, investors pay a reasonable multiple for 30%-plus volume growth, a 0.8x balance sheet and a 3-million-unit runway.

Bear view: At $68.31 the enterprise value is about 27x FY26 adjusted EBITDA and nearly 29x trailing, for a business whose per-unit profit is shrinking and whose new GAAP margin target is 360bps away. The price-target changes published on the reaction day were mostly small cuts, to a band of roughly $82 to $100, with no rating changes, which does not suggest the Street sees a step change in value.

Our take: Fair, not cheap. At Q2 the 25x multiple on a falling stock was our reason to stay Outperform through a weak guide. At 27x after a 16% move, with the margin catalyst deferred, the same multiple argument no longer carries an above-market rating on its own. A return toward the pre-print level near 23.5x would.

Model Update

Valuation impact. At the $68.31 reaction close and 1.127 billion fully converted shares, market capitalization is about $77.0B and enterprise value about $79.2B against $2.18B of net debt. That is 28.8x trailing adjusted EBITDA, 27.1x the FY26 upper-half midpoint, 26.4x the top of the range, 24.7x the Q3 annualized run rate and 20.6x our FY27 estimate. Before the print, the same FY26 multiple was 23.5x. We do not regard 27x as expensive for a business compounding units near 40% with a 0.8x balance sheet. We do regard it as fully reflecting the growth we can see, while the margin expansion that would justify a higher multiple has been deferred by management's own choice. That is the definition of a Hold in our framework.

Thesis Scorecard Post-Earnings

Overall: The thesis split in two. The growth and balance-sheet pillars strengthened. Units re-accelerated, production broke a record, the guide moved up, leverage fell to 0.8x, and the growth-sustainability bear point receded. The margin pillars weakened. Per-unit profit fell for a fourth straight quarter by the largest amount yet, after the two distortions we expected to flatter Q3 had lifted. Management also said it prefers to take incremental value as units, as far as the supply chain allows. Our Outperform rested on both halves: growth at a reasonable price, with operating leverage arriving to justify the multiple. One half delivered and the other was deferred, while the price rose 15.7%.

Action: Downgrading to Hold from Outperform. We set three downgrade triggers at Q2: Q3 inventory growth again failing to match retail unit growth; a fourth consecutive year-over-year decline in adjusted EBITDA per unit after the reconditioning normalization and the denominator lap had both landed; or full-year adjusted EBITDA tracking to the lower half of the range. The third did not occur: the guide moved to the higher end, and our reading of the guide as a floor was right. The second occurred clearly, and the first occurred by management's own account, with the trend improving. We said we would move to Hold on any of them. We upgraded at Q1 on pre-committed triggers, and the same discipline applies when they point the other way. This is a Hold on a business we admire at a price that now reflects its growth, not a call that anything is broken.

What would bring us back to Outperform. Any of the following: (a) a year-over-year increase in adjusted EBITDA per unit, or in GAAP operating income per unit once 2027 reporting begins, which on our arithmetic needs Q4 near the top of the guided range; (b) management confirming that retail-ready listed inventory is growing at least as fast as sales, which would release the conversion constraint and reduce the need for advertising support; or (c) a pullback toward the pre-print level near $59, about 23.5x FY26 adjusted EBITDA, with the growth pillars intact. What would take us to Underperform: unit growth falling below 30% while per-unit profit is still declining, which would mean the company is spending margin and no longer getting volume for it.

Net: The best volume quarter in Carvana's history, a higher guide and a 15.7% rally. It also brought a fourth straight decline in the per-unit profit that the long-term target depends on, a management statement that this will continue by choice, and the end of guidance. We took the bull side on growth at Q2 and were right. We stated in advance what would make us step back on margins, and it happened. We move to Hold.