Carvana — Blitz Index 54 (C)

units +52%, SG&A/unit −41% — against 13–14% coupons and an 83% family vote.

Last re-scored: October 28, 2026.

Score breakdown

Think we scored it wrong? Score it yourself →

Radar axes (0–100): Moat 62 · Speed 80 · Creative 73.

Watch item

D −14, below the FLAG line. Gates: G1 ✓ · G2 ✓. 11 points below the threshold — first score. The business case (68 pre-demerits) and the trust case are two different scores.

Strengths

• 10 straight quarters of growth + profitability
• FY2026 EBITDA guidance $2.7–3.0B (raised)
• Reconditioning capacity 1.5M units (2.5× sales)
• +50 institutional 13F holders

Concerns

• GPU down 3 straight quarters (Q2 −$412) — peak-unit-profit question
• 13–14% coupons + PIK + indenture covenants — extension is not repayment
• Garcia family 83% vote + $1,721M TRA + DriveTime related-party
• 62.2M shares short — 10.23 days to cover

Research notes

Thesis: a genuine turnaround wearing an expensive balance sheet

Carvana is the rare company in this scorecard whose business and whose balance sheet tell two different stories, and whose total score of 54 out of 100, grade C, is best read as the collision between them. Add the three business sections together and the result is 68: speed 20 out of 25, moat 37 out of 60, creative destruction 11 out of 15. That subtotal clears the 65-point line that BlitzLab uses as its threshold. Then deductions of 14 points pull the total back to 54, eleven points short of the line. The deduction total of 14 sits one point below the 15-point FLAG threshold, so no FLAG applies. Nothing about the arithmetic is marginal or ambiguous: 20 plus 37 plus 11 minus 14 equals 54, and every item subscore sums exactly to its section total. The question this body tries to answer is not whether the turnaround is real. On the evidence in the filings, it is. The question is whether a turnaround that is real in its operations can outrun a capital structure that still carries the marks of the crisis it survived.

The operating facts, as of the scoring date of October 16, 2026, are striking. Revenue in fiscal 2025 was $20,322 million, up 49% year over year, and the first half of 2026 accelerated rather than decelerated: revenue of $13,808 million in the first half of 2026 was up 52.2% against the first half of 2025, with quarterly revenue rising from $6,432 million in the first quarter to $7,376 million in the second. Retail units reached 596,641 in fiscal 2025, up 43% year over year, and the company reports ten consecutive quarters of growth combined with profitability, a claim the score file notes is checkable against quarterly filings. Trailing-twelve-month revenue stood at approximately $25.1 billion, which is why the score file classifies Carvana as a growth company in turnaround rather than as an early-stage venture. This is a first scoring, not a rescore, and the latest results available at the scoring date were the second quarter of 2026, filed on July 29, 2026. Third-quarter 2026 results, which had not been published before the scoring date and would ordinarily arrive in late October, are excluded from the score and appear in this body only as a future marker, with no anticipated outcome attached.

The unit economics behind that growth are what separate a recovery in revenue from a recovery in the business itself. Total gross profit per unit, or GPU, rose from $3,022 in 2022 to $7,026 in fiscal 2025, a multiple of roughly 2.3 times. Non-GAAP selling, general and administrative expense per unit fell from $5,862 in 2022 to $3,433 in 2025, a decline of 41%. Adjusted EBITDA margin moved from negative 7.7% to positive 11.0% over the same span. Those three movements together are the substance of the first gate verdict. Gate 1, which asks whether growth continues or a second growth curve exists, passes on confirmed evidence: unit growth of 43%, ten consecutive quarters of growth with profitability as reported by the company, the GPU improvement, the per-unit cost decline, and the margin swing. The score file's own summary of the gate is that a turnaround that looked like a near-death experience in 2022 is now demonstrated in unit economics rather than merely asserted in narrative. Gate 2, which asks whether the ten-year ceiling of the new market is at least ten times current revenue, passes as an estimate: the bar implied by trailing-twelve-month revenue of approximately $25.1 billion is approximately $251 billion, and the estimated United States used-car retail market of approximately $930 billion is 3.7 times that bar. The estimate rests on arithmetic built from Cox Automotive data cited in the annual report, 37 million transactions multiplied by an average transaction price of approximately $25,000, rather than on the company's own market claim alone.

Yet the same score file that records those strengths records a capital structure that has not yet been normalised, and that is where the 14 points of deductions come from. Total debt at December 31, 2025 was $4,976 million on a principal basis. Senior Secured notes carry coupons in the 13 to 14% range and include payment-in-kind features, meaning that principal on the 2031 notes grew by $287 million during 2025 without any cash changing hands. The company's 2023 debt exchange, which produced an $878 million gain, bought time; the score file is explicit that it was an extension, not a repayment. The governance picture compounds the financial picture: the Garcia parties together control approximately 83% of the voting power, and a tax receivable agreement liability of $2,228 million includes $1,721 million attributable to the Garcia parties. A reader who wants to understand the score of 54 should keep both halves in view at once. The engine of the business, measured by units, per-unit profit, and per-unit cost, is demonstrably working. The fuel for that engine remains expensive, the covenants that govern it remain restrictive, and the steering remains concentrated in one family. That tension, more than any single number, is the subject of this body.

Speed: growth at a scale where growth is supposed to slow

Speed scores 20 out of 25, the strongest section in percentage terms at 80.0 on the radar conversion, and it is built from four items: growth 9 out of 10, capital velocity 3 out of 5, hiring and productivity 4 out of 5, and share velocity 4 out of 5. The growth item alone explains why the section scores so highly. Fiscal 2025 revenue growth of 49% year over year, first-half 2026 growth of 52.2%, and ten consecutive quarters of unit growth in a range the score file puts at 38 to 46% describe a company still expanding at a pace more typical of a much smaller business, while operating at a trailing-twelve-month revenue scale of approximately $25.1 billion. The score file withholds the tenth point for a specific reason: a perfect growth score would require demonstrated growth in the 100% or higher range, and Carvana's growth, however strong, is in the 50% range. The discipline of that withholding matters because it shows the score is not grading on enthusiasm. It is grading on a scale where the top mark is reserved for something this company has not shown.

The quarterly shape of the growth also matters. Revenue rose from $6,432 million in the first quarter of 2026 to $7,376 million in the second, and retail units of 596,641 for fiscal 2025, up 43% year over year, confirm that the revenue growth is substantially a volume story rather than a price-only story. The company's fiscal 2026 adjusted EBITDA guidance of $2.7 billion to $3.0 billion, raised from an earlier level and cited in the score file as an official company figure, is consistent with that volume story, but it is guidance, not a result, and this body treats it as such. It is recorded as management's stated expectation, not as evidence of what has already happened. What has already happened is the ten-quarter run of growth with profitability, the 43% unit growth, and the first-half acceleration to 52.2%. Those are the confirmed facts that carry the growth score of 9.

Capital velocity, at 3 out of 5, is the weakest item in the speed section, and the reason is cost and constraint rather than access. Carvana can raise capital; the question is on what terms. Through an at-the-market programme, the company issued 6.826 million shares for net proceeds of $1,264 million in 2024 and 1.5 million shares for net proceeds of $539 million in 2025, leaving a remaining facility of $461 million. In August 2026, the company arranged a $1.66 billion Term Loan B maturing in 2033 with Barclays, intended to repay the 2030 Senior Secured notes, as disclosed in an August 14, 2026 filing. The score file marks the coupon and the completion of that refinancing as unconfirmed, and this body preserves that label exactly: the loan's existence and purpose are disclosed, but its pricing and whether it has closed are not confirmed in the score file, and no figure is supplied here to fill that gap. Around both the equity programme and the debt sit indenture covenants that restrict additional borrowing, liens, dividends and share repurchases, and asset sales. The score of 3 therefore describes a company that can fund itself but cannot fund itself cheaply or freely. In blitzscaling terms, speed that is financed at 13 to 14% coupons and under restrictive covenants is speed with a meter running.

Hiring and productivity score 4 out of 5, and the underlying figures show a company expanding its workforce in step with its volume rather than ahead of it. Headcount rose from more than 17,400 to more than 23,100, an increase of 33% year over year, which the score file characterises as expansion during a blitz phase. Revenue per employee on a trailing-twelve-month basis is approximately $1.08 million, an estimated figure that should be read as an order of magnitude rather than a precise productivity measure. The point of the item is directional: a company adding a third to its workforce while growing revenue by roughly half is not diluting its productivity as it scales, and that is consistent with the per-unit cost improvement recorded elsewhere in the score. The score stops at 4 rather than 5 because the evidence is headcount and an estimated per-employee figure, not a demonstrated productivity breakthrough of the kind the top mark would require.

Share velocity scores 4 out of 5, and here the context is a fragmented market rather than a concentrated one. Carvana's share of United States used-car retail is approximately 1.6% according to the annual report, and approximately 1.5% on the annualised run-rate basis used in the second-quarter 2026 shareholder letter, which annualises around 800,000 units. The difference between those two figures is a difference of basis, not a contradiction, and this body reports both with their bases attached. The structural fact that makes a share of 1.6% interesting is that the top ten retailers combined account for less than 10% of the market, according to Cox Automotive data cited in the annual report. In a market that fragmented, a single online platform at 1.6% is already one of the largest participants, and unit growth of 43% means that share is rising. The score stops at 4 because pricing power has not been demonstrated. Carvana is gaining share by growing faster than the market, not yet by showing that it can set prices within it. That distinction will matter when the moat section is considered, because distribution scale and pricing power are not the same thing.

Moat: a logistics network searching for a network effect

The moat section scores 37 out of 60, or 61.7 on the radar conversion, and its four items split into two very different stories. Distribution scalability scores 13 out of 18 and market size scores 10 out of 12. Network effects score only 6 out of 18 and gross margin scores 8 out of 12. The shape of that split is the analytical core of the Carvana case: the company has built something physically difficult to replicate, in a market large enough to matter, but it has not yet demonstrated the self-reinforcing data or ecosystem loop that the strongest moats in this scorecard possess, and its per-unit profit, while much improved, has begun to slip from its peak.

Distribution is the strongest moat item, and the evidence is physical. Carvana reaches 316 metropolitan statistical areas, covers more than 80% of the United States population with its own logistics, and reports annual reconditioning capacity of approximately 1.5 million vehicles, which is 2.5 times its 2025 sales volume. Sixteen inspection and reconditioning centres have been built out of the 56 former ADESA sites, leaving further headroom for expansion within the existing footprint. Website units, a measure of vehicles listed or handled through the online channel, stood at 75,683. Ten consecutive quarters of expansion back the claim that this is not a plan but a demonstrated capability. The score file calls this a proven case of heavy-equipment distribution scalability, and the phrase is apt: Carvana's distribution advantage is not an algorithm but a network of sites, trucks, and reconditioning lines that converts a national online storefront into a delivered vehicle within days. A competitor that wanted to match it would need to build the same physical network, not merely write the same software. That is why the distribution score of 13 sits in the upper part of its range, even though it stops short of the top band, which would require evidence of a distribution advantage that compounds without further capital spending.

Market size scores 10 out of 12, and the reasoning is unusually independent of company claims. The gate arithmetic described earlier, 37 million transactions at an average price of approximately $25,000 producing a market of approximately $930 billion, is Cox Automotive-based arithmetic reproduced in the score file, not a company-supplied total addressable market figure adopted uncritically. Against a gate bar of approximately $251 billion, the market is 3.7 times the threshold, and the company's long-term goal of 3 million units per year would represent 7.5% of that market. The score of 10 rather than 12 reflects the fact that the market figure remains an estimate, built from transaction counts and average prices, rather than a directly observed revenue pool. But the direction is not in doubt: even if the average transaction price or the transaction count were materially lower, the market would remain several times the gate bar. Market size is not the constraint on this company. The constraints lie elsewhere.

Network effects score 6 out of 18, the weakest moat item, and the score file is candid about why. A genuine network effect has not been quantitatively demonstrated. What exists is a management assertion that as scale grows, the company's models improve, which is a claim about a data loop rather than evidence of one. The score file places the score at the bottom of the 5 to 9 anchor band, signalling that some data and ecosystem effect exists but is weak. Two pieces of evidence support that modest placement rather than a lower one. Monthly unique visitors reached 18.5 million, which is a real audience scale, and Carvana Finance creates a form of lock-in: new loans originated reached $12,807 million and gains on loan sales reached $1,193 million in fiscal 2025. A customer who finances through Carvana, and a company that originates and sells loans at that scale, are tied together more tightly than a customer who merely browses a listing. But neither figure demonstrates the classic network dynamic in which each additional participant makes the platform more valuable to every other participant. Visitors do not obviously make the marketplace better for other visitors, and loan origination is a financing business attached to the marketplace rather than a network effect within it. The score of 6 is therefore an honest middle-low judgement: something is there, but it is not the moat, and a reader should not mistake audience size for network effect.

Gross margin, scored through total GPU, is 8 out of 12, and it is the moat item where the trend matters more than the level. The level is strong: total GPU of $7,026 in fiscal 2025, up $118 year over year. The trend is the concern: GPU fell year over year for three consecutive quarters, with the second quarter of 2026 at $7,014, down $412 year over year, the first quarter at $6,783, down $155, and the fourth quarter of 2025 at $6,427, down $244. The score file raises the possibility that the fiscal 2025 figure was a peak, and this body preserves that framing as a question rather than converting it into a conclusion. Two counterweights are recorded in the same section. Operating leverage is real: non-GAAP selling, general and administrative expense per unit fell 41% from 2022 to 2025, so the cost side of the unit equation continues to improve even as the gross profit side softens. And adjusted EBITDA margin, while down from 11.0% in fiscal 2025 to 10.4% in the first half of 2026, has not collapsed; it has flattened. Free cash flow of approximately $889 million, an arithmetic figure derived in the score file rather than a directly reported line, points in the same direction: the business generates cash, but the per-unit profit engine that generates it is no longer accelerating. The score of 8, with the score file explicitly noting that a 10 was withheld because of the softening trend, captures that balance. It rewards a strong level and penalises a weakening direction, which is exactly what a scoring system should do when a peak is possible but not yet proven.

Creative destruction: replacing the lot with the platform

Creative destruction scores 11 out of 15, or 73.3 on the radar conversion, and its three items describe a displacement that is well under way but not complete. Replacing the existing order scores 4 out of 5. Irreversible behaviour change scores 3 out of 5. Cost-structure change scores 4 out of 5. Together they portray a company that has changed how a meaningful minority of used-car transactions happen, in a market whose structure makes that change unusually consequential.

The order being replaced is the offline franchised-dealer model of used-car retail, in which inventory is local, selection is limited to what is on a particular lot, and the transaction is negotiated in person. Carvana's model replaces that with a national online inventory, centralised reconditioning, and delivery within days, and the annual report describes the company as, in its own words, "the leading e-commerce platform for buying and selling used cars." That self-description is a company claim, and this body reports it as such, but the structural facts behind it are independently observable: a share of approximately 1.6% of national retail, achieved as a single online platform, in a market where the top ten retailers combined hold less than 10%. In a concentrated market, a 1.6% share would be a footnote. In a market this fragmented, it is evidence that a new model can aggregate demand nationally in a way the incumbent structure, built lot by lot, could not. The score of 4 rather than 5 reflects the incompleteness of the displacement. The offline dealer remains the dominant way Americans transact in used cars, and Carvana's share, while rising, remains a small fraction of the total. Replacement is in progress, not accomplished.

Behaviour change scores 3 out of 5, and the evidence here carries an explicit verification caveat that this body preserves. According to the fourth-quarter 2025 shareholder letter, 30% of retail customers and 60% of customers selling a vehicle completed their transactions without any human contact. Those figures are company-reported and independently unverified, and the score file marks them as a confirmed and unconfirmed mixture. The direction they point to is nonetheless significant: a transaction as large, as infrequent, and as traditionally tactile as a car purchase is being completed, by a substantial minority of this company's customers, without speaking to anyone. That is the kind of behaviour change that, if it generalises, does not easily reverse, because it rewires what customers expect a car transaction to feel like. The score of 3 rather than higher reflects the status of the evidence. The figures come from one company's shareholder letter, they have not been independently verified, and the majority of retail customers still do interact with a human at some point. The behaviour is becoming mainstream within Carvana's customer base; it has not yet been shown to be mainstream across the market.

Cost-structure change scores 4 out of 5, and here the evidence is the strongest in the creative destruction section because it is arithmetic rather than assertion. Non-GAAP selling, general and administrative expense per unit fell from $5,862 in 2022 to $3,433 in 2025, a decline of 41%, and GPU rose from $3,022 to $7,026 over the same period. Those two movements describe a structural change in how a used car reaches a customer: the fixed costs of logistics and reconditioning infrastructure, spread over a rapidly growing unit volume, produce a lower cost per transaction than the variable, lot-by-lot cost structure of the traditional model, while the centralised reconditioning process supports a higher gross profit per unit than the company achieved when it was smaller. The score of 4 acknowledges that the cost structure has demonstrably changed while stopping short of claiming that the change is permanent or complete. Cost advantages built on volume can narrow if volume growth slows, and the GPU softening discussed in the moat section is a reminder that a changed cost structure is not the same as an impregnable one. But as of the scoring date, the direction and the magnitude of the change are confirmed in the filings, and they are the most concrete evidence in the score that Carvana is not merely selling cars online but selling them through a different cost structure.

Deductions and risks: why 68 becomes 53

The deductions section is where this scorecard departs most sharply from a conventional growth narrative, and it is the section a reader must understand to understand the total. Deductions total 15 points: debt 5, revenue quality 3, governance 4, and regulatory and social costs 3. Each item is grounded in the filings, and together they describe risks that are structural rather than cyclical. They will not disappear with one good quarter, and several of them would require years of deliberate action to reduce.

Debt, at negative 5 out of a possible negative 10, is the largest single deduction, and the figures explain why it is not larger and why it is not smaller. Total debt at December 31, 2025 was $4,976 million on a principal basis, composed of asset-based debt of $940 million, Senior Secured debt of $3,929 million, and Unsecured debt of $107 million. Senior Secured coupons in the 13 to 14% range are exceptionally high for a company of this scale, and the payment-in-kind feature means that the 2031 notes added $287 million of principal during 2025 without any cash payment. Indenture covenants restrict additional debt, liens, dividends and share repurchases, and asset sales, which limits management's freedom to manoeuvre. The August 2026 Term Loan B of $1.66 billion, intended to repay the 2030 Senior Secured notes, is a step toward refinancing, but its coupon and completion are unconfirmed in the score file, and this body does not treat an intended refinancing as a completed one. Against those negatives stand real offsets, and the score of negative 5 rather than a deeper deduction reflects them. Net debt to adjusted EBITDA is 1.18 times, interest coverage is 4.4 times, cash stood at $2,630 million, and a floor plan facility of $1.5 billion was barely drawn, with only $58 million used. Those ratios describe a company that can service its debt from its current earnings and cash, even though the debt itself is expensive and restrictive. The score file's summary phrase captures the balance precisely: the refinancing was achieved, but repayment has not been. The 2023 exchange bought time. It did not extinguish the obligation, and the maturities of the 2030 and 2031 notes remain ahead. The central risk in this section, which the score file names as the single largest blitzscaling risk for this company, is the combination of falling GPU and rising rates arriving together, in which case a turnaround narrative could become a second crisis. That is a scenario, not a prediction, and this body records it as the score file frames it: as the structural reason the debt deduction exists.

Revenue quality, at negative 3, is subtler than the debt item but important for reading the profit figures correctly. There has been no accounting restatement, the auditor is Grant Thornton, and internal control over financial reporting was effective as of December 31, 2025 with no material weakness reported. Those are clean facts, and they limit the deduction. The concern is the composition of reported profit. Of fiscal 2025 GAAP net income of $1,895 million, approximately $685 million consisted of non-cash tax effects, comprising a valuation allowance release and tax receivable agreement liability recognition. Quarterly net income is also moved by fair value swings on Root warrants of more than $100 million in either direction. And fiscal 2023 net income of $150 million was entirely attributable to the $878 million debt extinguishment gain, meaning that the headline profit of the turnaround's early year was an accounting consequence of the debt exchange rather than an operating result. None of this implies misstatement; the score file records no restatement and no control failure. It implies that GAAP net income is a noisy guide to operating performance for this company, and that adjusted EBITDA, GPU, and per-unit costs are more reliable lenses. A reader who cites the $1,895 million net income figure without noting that approximately $685 million of it is non-cash tax effects would be overstating the cash substance of the result, and this body therefore keeps the two figures together whenever the net income figure appears.

Governance, at negative 4 out of a possible negative 5, is the deepest deduction in percentage terms, and it is the item most likely to surprise a reader who has focused only on operations. The Garcia parties together control approximately 83% of the voting power, with Garcia II at 50% and Garcia III at 33%, through Class B shares carrying ten votes per share, making Carvana a New York Stock Exchange controlled company. The tax receivable agreement liability of $2,228 million includes $1,721 million attributable to the Garcia parties, which means that a substantial future cash obligation of the company runs to its controlling shareholders. Related-party transactions with DriveTime, a Garcia-associated business, are extensive: vehicle service contract sales commissions of $338 million in 2025, representing approximately 20% of other revenue, warranty administration fees of $24 million, leases, and servicing fees of $6 million. The board of ten directors was judged entirely independent, which is a genuine counterweight, but independence of judgement and concentration of voting power are different things, and the score reflects the latter. When one family controls 83% of the vote, receives the majority of a $2,228 million tax receivable liability, and transacts with the company at a scale of hundreds of millions of dollars a year, the governance risk is a single point of failure in the score file's own terms. It is not a finding of wrongdoing; it is a structural observation about where power and cash flows sit.

Regulatory and social costs, at negative 3, complete the deduction picture with a set of contingent liabilities that are difficult to quantify, which is itself part of the risk. A securities class action, In re Carvana Co. Securities Litigation, case number CV-22-2126 in Arizona, survived a partial dismissal on December 16, 2024 and is in discovery, with no loss estimate possible. State attorneys general inquiries covering titling, registration, lending practices, and after-sales service continue, with litigation possible if matters are not resolved amicably, in the annual report's phrase. Patent and consumer protection class actions proceed in parallel. The legal reserve moved from $10 million to $14 million, a figure the score file characterises as thin against that docket. The current status of the dealer licence suspensions imposed in 2022 and 2023 is not separately stated in the current annual report, and this body marks that as unconfirmed rather than assuming either that the issue is resolved or that it persists. The deduction of 3 reflects a docket that is active, multi-front, and unquantifiable, against a reserve that is small relative to the matters it would need to cover. It is not the largest deduction, but it is the one whose eventual cost is least predictable from the filings.

Two auxiliary signals, explicitly not reflected in the score, are worth recording because they show how the market is reading the same facts. Institutional holders numbered 831 in the second quarter of 2026, up by 50, holding 737.5 million shares after a five-for-one split, worth $50.8 billion, according to local filing data cited in the score file. Short interest reported by FINRA stood at 62.2 million shares as of September 15, 2026, down 5.6% but still substantial, representing 10.23 days to cover. The score file's reading of that pairing is that discovery and doubt coexist: institutions are accumulating a position in the turnaround while short sellers maintain a thick position against its risks. Neither signal changes the score, and this body does not use either to adjust it. They are included because they corroborate the central tension of the scorecard from outside the company's own filings: the business evidence attracts capital, and the capital-structure and governance evidence sustains scepticism, at the same time.

What to watch: the three tests that will move this score

The score file identifies three key drivers, and they translate directly into what a reader should watch in future quarters. The first is the simplest to state and the hardest to resolve: whether the business can outgrow its capital structure. The business subtotal of 68 already exceeds the 65-point threshold, and the only reason the total is 53 is the 15 points of deductions. That arithmetic defines the path to a higher score with unusual clarity. It does not require the business to become dramatically better; it requires the deductions to become smaller. Debt normalisation, which would mean refinancing expensive secured notes on confirmed cheaper terms rather than merely intending to do so, a reduction in the payment-in-kind growth of principal, and progress toward repayment rather than extension, would address the largest deduction. Governance change, which would mean any dilution of the 83% voting concentration or reduction in related-party scale, would address the deepest deduction in percentage terms. Either would move the total more than any plausible improvement in an already strong speed score. The score file also notes three model questions for a future version of the methodology, none of which changes the current score: whether payment-in-kind and ultra-high-coupon debt deserve a separate debt anchor distinguishing refinancing from repayment, whether large non-cash tax effects in a turnaround deserve a dedicated revenue-quality anchor, and whether related-party economics of the tax receivable agreement and DriveTime type deserve a quantitative governance anchor. Those are open questions for the methodology, recorded here so that a reader understands the current score's limits.

The second test is GPU. Fiscal 2025 total GPU of $7,026 may have been a peak, with three consecutive year-over-year declines culminating in the second quarter of 2026 at $7,014, down $412 year over year. The score file names the third-quarter 2026 results, expected around late October and excluded from this score because they had not been published before the scoring date of October 16, 2026, as the first judgement point on that question. This body attaches no anticipated outcome to that marker. What matters structurally is the interaction the score file identifies: per-unit cost leverage, with selling, general and administrative expense per unit down 41% since 2022, has so far offset GPU softening, and adjusted EBITDA margin has flattened at 10.4% in the first half of 2026 rather than falling. The moment to watch is the moment those two lines stop offsetting each other, if they do. If GPU stabilises while per-unit costs continue to fall, the unit equation improves. If both soften together, the turnaround narrative faces its first serious operating test since 2022. Either way, the judgement will come from reported GPU and per-unit cost figures in future filings, not from expectations formed in advance.

The third test is the market's own verdict, as expressed in the short interest and institutional data. Short interest of 62.2 million shares and 10.23 days to cover as of September 15, 2026 is thick by any standard, and it means that scepticism about the capital structure and governance is not a fringe view but a substantial positioned view. Institutional accumulation, up 50 holders to 831 in the second quarter of 2026, means that the operating evidence is attracting capital at the same time. Those two positions cannot both be fully right about the same future, but they can both be reasonable readings of the same present, because the present genuinely contains both a working business and an unnormalised balance sheet. The score of 54, grade C, with no FLAG marker, is BlitzLab's way of holding that contradiction in a single number: a business that scores like a blitzscaler, deductions that score like a company still carrying its crisis, and a total that belongs to neither story alone. The next chapters of this story will be written in filings: in GPU and per-unit cost lines, in the confirmed terms of any refinancing, in the trajectory of the Garcia voting concentration and related-party scale, and in the resolution or escalation of the litigation docket. Until those filings arrive, the honest summary is the score itself. Carvana has proven that it can sell cars at national scale, profitably, through a different cost structure, and it has not yet proven that it can do so on a capital structure and governance structure that no longer carry the marks of 2022.

More Platforms companies

Other Platforms names on the Blitz Index, highest score first: Alphabet (83/A), Meta (78/B), Airbnb (76/B).

Next earnings

Q4 2026 · Expected

Coverage

Carvana Co. · Q3 2026 due Oct 28, 2026 (AMC) · Score 54 · Grade C

Embed this score

Copy the HTML below to show this score on your site. The badge links back to this page and updates automatically when the score changes.

<a href="https://blitzlab.app/companies/cvna/"><img src="https://blitzlab.app/badge/cvna.svg" alt="Carvana Blitz Index 54/C — BlitzLab"></a> Copy HTML

Preview:

BlitzLab scores businesses against a scholar’s standard — not investment advice. No price targets, no buy/sell recommendations. Figures are from SEC filings (XBRL) as labeled in the source scorecard.