FY25 Closes at the High End But Q4 Recon Cost Stumble + No FY26 Framework Drive an 8% Sell-Off — Maintaining Hold

Initial Read. The eighth consecutive industry-leading quarter, but the first in which an operational mistake (reconditioning costs) was visible in the print — and management chose to skip a quantified FY26 framework, leaving investors to underwrite the long-term thesis without near-term anchors.

Results vs. Consensus

Top-line and FY25 anchors landed where we modeled or better; the Q4 margin and the missing FY26 framework are the binding signals.

Quality of Beat / Miss

• Revenue: The +58% Y/Y print again magnified by gross-revenue accounting on marketplace-partner inventory (Q3’s same dynamic, larger this quarter). Underlying organic growth is closer to the +43% retail-unit pace.
• Margins: Adj EBITDA margin compressed 100bps Y/Y (9.1% vs. 10.1%) for two reasons: (1) the gross-revenue accounting optic mathematically dilutes the percentage even when dollars grow; (2) genuinely elevated recon costs in Q4. Garcia and Jenkins both put the dollars rather than percentage as the more useful read — on $/unit Adj EBITDA, the Y/Y figure is approximately flat to slightly down (Garcia cited “down by about $14 year-over-year”).
• EPS: $4.22 GAAP is meaningless on its own. Strip out the $618M non-cash tax benefit (and the $67M warrant fair-value reduction offset within it), and the run-rate operating EPS is closer to $1.50, which itself is held up by the gross-revenue accounting / mix; on an apples-to-apples basis the Q4 operating earnings power is meaningfully softer than Q3 even though Q4 is seasonally a similar volume quarter.

Operational Performance

Key Topics & Management Commentary

Overall Management Tone: Visibly more candid than prior quarters about an operational miss while still confidently anchoring on the long-term framework. Garcia spent the most prepared-remarks airtime to date on a single underperforming line item (reconditioning), framing the stumble as predictable for any business growing 40%+ and pre-committing to a 3-to-6-month recovery timeline. Notably, the 10-K-tied related-party allegations got a firm, factual response and no defensive flailing.

The Reconditioning Stumble: What Happened, How They’re Fixing It

Sharon Zackfia (William Blair) opened Q&A by going straight to the issue. Garcia’s response was the most candid moment of any CVNA call this year:

“In Q4, I think there’s no question that our expenses were a little higher than we would have liked there. And I think that is partially the result of these additional sites kind of having a single line instead of multiple lines and there being some extra costs that flow through there as a result. I think it’s also partially a result of as we kind of spread out, we had some newer managers. ... I think this is one of those things where I think sometimes if you take a little step backwards, it kind of fires you up, and my strong guess is we’ll be in a better spot in 3 to 6 months than we would have been otherwise.”
— Ernest Garcia, CEO

The mechanics: 10 new ADESA reconditioning sites integrated in 2025 (15 entering the year, 25 by year-end, 34 capable centers in total), each starting on a single production line at sub-half utilization. Cost per car runs higher in this ramp phase than at mature multi-line legacy IRCs. Add newer site managers learning the playbook, and the cost line gets pressured exactly when overall volume is growing fastest.

Brian Nagel (Oppenheimer) followed up on what specifically changed in Q4 vs. Q3. Garcia’s answer reinforced that this was a cumulative-pressure issue, not a discrete event:

“The most important answer, honestly, there’s no unique dynamic that instantaneously changed. ... We grew quickly. We’re growing inventory quickly in the fourth quarter. We’re hiring new managers and kind of moving around some management layers to put us in a position to continue to grow quickly. And so I think there are moving pieces and sometimes that leads to a little backsliding.”
— Ernest Garcia, CEO

Read-through: Q1 2026 will see retail GPU under further pressure (Jenkins explicitly guided non-vehicle costs higher Y/Y in Q1) before improvement. The structural takeaway is that integrated ADESA sites take longer to reach legacy-IRC efficiency than the Q2/Q3 narrative implied. Not a thesis-breaker; an unhappy data-point on the speed of the cost-takeout flywheel.

FY26 Guidance: “Significant Growth” Without Numbers

Multiple analysts probed the missing FY26 framework. Rajat Gupta (JPMorgan) asked whether “profitable growth” specifically meant Y/Y EBITDA-per-unit expansion in 2026. Garcia’s answer was the most carefully hedged of the call:

“We had those 3 priorities from last year. This year, we’re leading a touch into growth. The other 2 priorities remain the same. ... We don’t think there — these are necessarily trade-offs. The trade-off is in our focus and where our priority is more than anything else.”
— Ernest Garcia, CEO

The operative phrase is “leading a touch into growth.” Combined with the explicit 2026 plan to begin full ADESA site build-outs (a step-up in capex intensity), the implication is FY26 is a re-investment year — growth-rate sustained, margin expansion paused. That’s analytically defensible but materially different from the trajectory the post-Q2 multiple was pricing.

Assessment: The non-disclosure is itself the disclosure. Management would have given a quantified range if it would have looked attractive. The fact that they didn’t implies FY26 adj EBITDA margin sits flat or modestly down vs. FY25’s 11.0% — consistent with a year of capex-intensive ADESA build-outs and continued advertising re-investment.

The Tax Benefit: Real Long-Term Value, Distortive Headline

Joe Spak (UBS) asked Jenkins to walk through the tax mechanics. The substantive answer:

“We have an UP-C corporate structure, the UP-C corporate structure generates significant tax assets when LLC units are exchanged into common shares, and we’ve had those changes happening over a number of years. ... Up until the fourth quarter, we’ve had a full valuation allowance against those tax assets. But with the realization of sustained profitability, we’ve now released that valuation allowance leading to the significant deferred tax benefit in Q4.”
— Mark Jenkins, CFO

Net of the tax shared with pre-IPO LLC unitholders (the tax receivable agreement liability), Carvana common shareholders captured more than $600M of value via the DTA release. The economic substance is that future cash taxes will be meaningfully reduced — a real, durable benefit. The accounting mechanics distort Q4 GAAP EPS and net income margin, but the cash benefit is genuine.

Read-through: Watch FY26 cash tax payments — they should be meaningfully below the implied GAAP rate, with the differential being the realized DTA. This is one piece of the print that genuinely strengthens the bull case once the headline distortion fades.

Related-Party Loan Sale Allegations: Firm Denial, 10-K Backstop

Daniela Haigian (Morgan Stanley) raised the short-report allegations in the most direct exchange of the call. Jenkins’ response left no rhetorical room:

“The answer there is very simple. All of our related party transactions are disclosed in our financial statements. As a specific matter, we do not sell loans to related parties and have not done so for all of the years from 2017 through 2025. Recent short reports that suggest otherwise, are inaccurate. ... Those reports are 100% inaccurate.”
— Mark Jenkins, CFO

Garcia’s follow-on (“if we have another short report during a quiet period at the end of the year, just maybe think back the last couple of years to recognize the pattern”) was characteristically blunt. Combined with the 10-K disclosures, this should not be a thesis-driving event absent new evidence.

Assessment: Watch item rather than thesis variable. The allegations and the responses both go in the same bucket as any short-report-driven price action — they create volatility but rarely create durable value. We’d need a regulatory inquiry or restated financials to elevate this to a thesis input.

Capital Allocation: Toward Investment-Grade and Toward ADESA Build-Outs

Two related capital-allocation signals: (1) net debt / TTM adj EBITDA at 1.3x and improving, with the company explicitly committed to investment-grade credit ratios over time; (2) full ADESA build-outs (vs. the prior capex-light integrations) starting in 2026, with the original $1B+ multi-year framework still the order of magnitude. The two pull in slightly different directions — deleveraging is FCF-positive, build-outs are FCF-negative — but combined they signal management is comfortable using FCF for organic growth rather than for share repurchase or distributions.

Customer Sharing: A Strategic Choice, Not a Margin Failure

The -$60/unit shipping-fee reduction (a deliberate pass-through of the inbound-transport savings) and the ~1% interest-rate reduction (a deliberate Other-GPU give-back) were both framed as fundamental gains shared with customers. Garcia’s framing — that high contribution margins permit continual customer-side reinvestment that compounds the offering advantage — is consistent with the long-arc playbook. It is also a useful pre-explanation for why GPU pressure may persist into 2026 even with operational improvements.

Guidance & Outlook

Implied math: A “significant” growth read on $2.20B FY25 adj EBITDA and 596,641 retail units puts our FY26 framework at $2.6–3.0B and 740–810K units — consistent with 28–35% unit growth and adj EBITDA dollar growth roughly tracking units. Margin flat-to-down by 50–100bps is the implicit base case absent guidance.

Guidance style: Carvana’s pattern is conservative-to-realistic Q-on-Q, with FY guidance typically raised through the year. The choice not to provide a quantified FY26 framework is unusual and signals that even management isn’t yet ready to anchor a number — either because the recon recovery timeline is uncertain or because the build-out capex profile is still being calibrated. Either way, “trust us” is a tougher sell at the current multiple than it was at Q2 levels.

Analyst Q&A — Notable Exchanges

Q&A skewed toward the recon stumble, the missing FY26 framework, and the related-party allegations. Notable threads:

• Sharon Zackfia (William Blair) opened directly on the recon dynamics and probed the “top-quartile” benchmark cited in the shareholder letter (~$220/unit benefit if all sites reached top-quartile efficiency). Garcia’s response acknowledged the gap and committed to a 3-to-6-month recovery framework.
• Jeff Lick (Stephens) probed the depreciation-environment reversal in Q1; Jenkins guided sequential retail GPU increase in Q1 despite continued non-vehicle cost pressure.
• Daniela Haigian (Morgan Stanley) raised the short-report related-party allegations — the most-watched exchange of the call. Jenkins’ firm denial and Garcia’s “another short report” quip were both well-handled. Her follow-up on variable adj EBITDA margin trajectory drew the cleanest restatement of the 13.5% long-term path.
• Brian Nagel (Oppenheimer) drew the most candid framing yet on what changed in Q4 (no single event, just cumulative pressure from rapid integration cadence + new managers).
• Rajat Gupta (JPMorgan) pressed on whether “profitable growth” meant EBITDA-per-unit expansion in 2026; Garcia’s carefully hedged “leading a touch into growth” framing was the most strategically meaningful guidance signal of the call. His follow-up on post-sale vertical integration drew Garcia’s clearest defense of the vertical-integration thesis broadly.
• Joe Spak (UBS) walked Jenkins through the affordability-and-rates question and the UP-C tax-asset mechanics. Jenkins’ explanation of the DTA valuation-allowance release was the cleanest disclosure of the call.
• Chris Pierce (Needham) got the non-vehicle cost framing (recon + inbound transport, not vehicle acquisition) and the title-and-registration progress narrative (~99% of packets completed by deadline, claimed best-in-class).
• Ron Josey (Citi) raised conversion rates and the e-commerce experience improvements; Jenkins on guidance philosophy — the most useful articulation of the “significant growth in units and adj EBITDA” framing without committing to a number. The operating ROA >20% framing was the new disclosure.
• Marvin Fong (BTIG) explored the customer pass-through math (rate cut to customers vs. flat Other GPU). Garcia’s cost-of-funds and attach-rate decomposition was the cleanest articulation of the trade-off.
• Lee Horowitz (Deutsche Bank) probed whether 2026 production growth meets demand (Garcia’s “build the operational chain and demand follows” framing) and whether competitor pricing is an issue (no notable change).
• John Babcock (Barclays) closed with two on volume/expansion: the implied 1Q sequential growth and the GPU-vs.-shipping-fee math.

What They’re NOT Saying

1. No quantified FY26 unit or EBITDA range. The single most-watched item from the print and the single biggest tell. Carvana’s pattern is to provide directional ranges; the choice not to here implies meaningful uncertainty about either recon recovery timing or capex pacing for ADESA build-outs.
2. No quantified FY26 capex framework. Full ADESA build-outs are described qualitatively (“multi-year, ~$1B order of magnitude with some inflation”) but no FY26-specific dollar number was provided. Modeling assumption: $300–500M of incremental capex tied to build-outs in 2026 alone. Could meaningfully pressure FCF.
3. No update on the franchise-dealership acquisitions. Same as Q3 — Garcia explicitly declined to comment on what is now ~6–7 Chrysler dealership acquisitions. Pattern continues. The non-engagement signals deliberate strategic optionality being kept off the model.
4. No specific commentary on share count trajectory or further ATM issuance. Q3’s ATM (clean deleveraging tool) was a positive trade. Whether it stays active in 2026 (and if so, for what use of proceeds) was not addressed. Watch item.
5. No quantified target for the “top-quartile” recon-cost benchmark. The shareholder letter referenced a ~$220/unit benefit if all sites reached the top quartile; on the call, Garcia confirmed the spread but didn’t commit to a timeline for closing it. Reasonable analytical framing puts this as a 4–6 quarter realization, not 1–2.
6. No detail on legacy-cohort unit economics. The same conspicuous non-disclosure as Q2 and Q3. By Q4 of the third year of profitable scale, the company should plausibly have shown some Atlanta or older-cohort decomposition; its continued absence is increasingly notable.

Market Reaction

• Pre-print setup: Stock had partly recovered from the Q3 -9.7% reaction, helped by the year-end rally and incremental news flow on ADESA integrations and balance-sheet retirement. Sentiment was mixed-to-positive heading into the print, with elevated short interest as a partial offset.
• Initial reaction: Print landed after the close on Feb 18. Shares fell roughly 8% in the next session as analysts trimmed price targets on the weaker-than-expected Q4 EBITDA, the recon stumble, and the missing FY26 quantified framework. Pre-market action saw moves into double-digit-percent territory before stabilizing.

The reaction is consistent with the second consecutive quarter of negative tape on a strong-headline print — confirming the rating gate we identified at Q2 and Q3 remains intact. The market is no longer giving CVNA the benefit of the doubt on margin expansion at the post-rally multiple; specific operational improvements need to print before the multiple expands further.

Street Perspective

Debate: Is the recon stumble a 1-quarter event or a longer-arc cost-takeout slowdown?

Bull view: Garcia committed publicly to a 3-to-6-month recovery; the underlying issue (new ADESA sites starting on single production lines + new managers) is fundamentally a ramp-curve issue that resolves with scale. Q1 2026 will already be visibly better even if not all the way back. The structural cost-takeout story is intact — this is a normal hiccup, not a re-rating event.

Bear view: The recon stumble is the first concrete evidence that the cost-takeout flywheel doesn’t work as smoothly as the Q2/Q3 narrative implied. If new ADESA sites take 2–3 years rather than 4–6 quarters to reach legacy-IRC efficiency, the path to 13.5% adj EBITDA margin extends meaningfully. Combined with deliberate customer pass-throughs on shipping and rates, FY26 margin could be flat-to-modestly-down, not expanding.

Our take: Closer to bull. The 3-to-6-month framing is operationally credible and Garcia has consistently delivered against pre-committed timelines historically. But the broader point that the cost-takeout is non-linear is real — we underwrite a slower-than-Q2-implied glide path to 13.5% margins, with 11.5–12.5% as the realistic FY27 range.

Debate: Does the missing FY26 framework signal management caution or just communication preference?

Bull view: Carvana has historically given multi-year frameworks (3M units / 13.5% margin) rather than tight FY ranges. The qualitative “significant growth” framing is consistent with that approach and shouldn’t be over-interpreted. The Q3-to-Q4 unit guide setup created a natural reluctance to anchor specific numbers given the recon recovery uncertainty.

Bear view: Most growth companies at this stage of maturity provide quantified FY ranges. The conspicuous absence implies management isn’t confident in the recon recovery timeline + capex pacing combo enough to anchor a number. That should compress the multiple until visibility improves.

Our take: Closer to bear. We don’t think the omission is a smoking gun, but it materially elevates the value of the Q1 2026 print as a confidence-restorer. If Q1 prints with visible recon improvement and an FY26 framework, the multiple deserves to recover. If Q1 prints with continued recon pressure and still no framework, downward earnings revisions are likely.

Debate: How much should the related-party short report matter?

Bull view: The 10-K disclosures are clear; Jenkins’ on-call denial was unambiguous; the underlying loan-sale partner expansion ($18B in committed agreements through 2027) is independently validated by external counterparties. There is no there there.

Bear view: Even spurious short reports create overhang and investor friction. The Garcia family’s ownership structure and DriveTime relationships have generated periodic governance questions for years and will continue to do so.

Our take: Bull. The substantive defense is solid. We treat this as a periodic noise-event that creates entry-point optionality but is not a thesis variable absent new evidence (e.g., regulatory inquiry, financial restatement, board-level change).

Model Implications

• FY26 retail units: 740–810K, implying 25–35% growth. We anchor at 775K (~30% growth).
• FY26 adj EBITDA: $2.6–2.9B; we anchor at $2.75B. Implied margin ~10.5–11.0% — flat to slightly down vs. FY25’s 11.0%.
• Q1 2026 retail GPU: Continued Y/Y pressure on non-vehicle costs; expect another quarter of negative GPU year-over-year before Q2/Q3 inflection.
• FY26 capex: Step-up vs. FY25 on the ADESA full-build-out program; modeling $400–500M of incremental capex over the historical baseline.
• FY26 cash taxes: Meaningfully below GAAP rate as the DTA realizes; meaningful FCF tailwind that partially offsets the capex step-up.
• Net debt: 1.3x at year-end; trending toward sub-1.0x by mid-FY26 even with the capex step-up. Investment-grade aspiration is now plausibly a 2027 event.
• Share count: Stable as ATM activity slows post-2028-notes retirement. Watch for new issuances if they appear.

Thesis Scorecard

Overall: Operating thesis intact, with three bull pillars (deleveraging, finance platform, DTA realization) modestly strengthened, two bear pillars (recon, FY26 framework) modestly elevated. The aggregate is a thesis that’s slightly more nuanced than at Q2/Q3 but not directionally different.

Action: Maintaining Hold (initiated at Hold at Q2 2025; maintained at Q3 2025). Q4 reinforces the structural pieces of the bull case while making the rating gate more visible. The reconditioning stumble is recoverable but the lack of FY26 framework + capex step-up + customer-side reinvestment combine to put the FY26 margin trajectory below where the multiple was pricing post-Q2. We continue to wait for either (a) a Q1 2026 print where recon costs visibly improve and an FY26 framework is articulated, or (b) a meaningful multiple compression that re-creates margin of safety. Either path could shift us to Outperform; this print does not.

Net: The franchise is real; the scale of operational complexity at 40% growth is real too. Q1 2026 is the most-watched print of the year.