Variant Perception

Variant Perception — Where We Disagree With the Market

Figures converted from EUR at historical FX rates — see data/company.json.fx_rates. Ratios, margins, and multiples are unitless and unchanged.

The single sharpest disagreement, stated plainly. The Street is Buy-rated (avg target $35.43, +27% to a $27.98 spot, no Sells) and underwrites AUTO1's FY2026 adjusted-EBITDA guide of $293.8–323.1m — a +33% YoY jump — as proof that operating leverage is reaccelerating. Our read of the evidence is that a large H2 YoY EBITDA headline this year is mechanically a comp-lapping artifact, not widening per-unit drop-through: Q1 2026 grew adjusted EBITDA only +3.0% on +22% units because it lapped the strong $68.3m Q1'25, while the easy comps ($49.7m / $61.0m / $53.1m) arrive in Q2–Q4. Adjusted EBITDA per unit is the tell, and it fell −16% YoY ($328 → $276) in Q1'26 and ran flat-to-down all of 2025. The variant is not "the stock is expensive" — it is that consensus is extrapolating the wrong number (a YoY headline off easy comps) to justify a 38× EV/EBITDA multiple that only holds if per-unit economics genuinely inflect. If the Street stops extrapolating the 2.4%→5% margin path, the multiple de-rates from ~38× toward 20–25× regardless of how good the unit and gross-profit headlines look.

This is a measurable gap with a near-dated resolution path: three hard prints (Q2 on 29 Jul 2026, H1 with first balance-sheet detail on 2 Sep, Q3 + FY-guide action on 4 Nov) each test whether the drop-through is real or comp-lapped. We are below consensus on EBITDA quality, not on volume — the volume and share story genuinely works.

Source: spot $27.98 (XETRA, 19 Jun 2026); consensus target $35.43 and Buy rating (MarketScreener / StocksGuide, ~14–15 analysts, no Sells); EV/EBITDA ~38× (Web Research tab); FY2026 guide $293.8–323.1m, midpoint $308.4m vs $232.1m FY2025 (company guidance, 25 Feb 2026).

The variant scorecard

Source: analyst scoring. Variant strength reflects materiality (a multiple de-rate, not just a 5% estimate miss) tempered by a genuinely credible bull bridge; consensus clarity is high (hard targets, ratings, guide numbers) but dented by the absence of a published Street quarterly EBITDA split; resolution clusters around the next three prints (29 Jul / 2 Sep / 4 Nov 2026).

Mapping consensus before we disagree

Every "the market believes X" below is nailed to a concrete consensus signal and converted into a testable implied underwriting assumption.

Source: consensus signals from Web Research tab (analyst targets, PT revisions, PEG), Financials tab (multiples), Moat tab (EV/GP bracket), and company guidance. Implied assumptions and confidence are the analyst's.

The disagreement ledger

Three candidate disagreements survived all five tests (a consensus analyst's view; contradicting report evidence; materiality to valuation/risk/timing; an observable resolution signal over the right horizon; and a clean refutation). They are ranked by expected value to a PM — the multiple-moving one leads.

Source: synthesised from the Catalysts, Financials, Forensics, Moat and Long-Term Thesis tabs. Rankings, bucket classifications and confidence are the analyst's.

Disagreement #1 — the comp-lapping illusion (the monetizable one)

What consensus would say: "AUTO1 guided FY2026 adjusted EBITDA to $293.8–323.1m, up 33%, and raised guidance twice during 2025. Q1 was just a noisy quarter of Autohero growth investment; the operating leverage is structural and the next prints will show EBITDA reaccelerating faster than gross profit."

Why our evidence disagrees: The +33% is arithmetically achievable without any improvement in per-unit drop-through, because the back-half of 2025 was weak. Q1'26 grew only +3.0% precisely because it lapped the strong $68.3m Q1'25. Q2–Q4 2026 lap $49.7m / $61.0m / $53.1m — so even per-unit EBITDA staying flat at ~$276 would produce a 30–45% YoY headline in H2 — a reacceleration headline with no movement in the per-unit number underneath it. The decisive metric — adjusted EBITDA per unit — fell −16% YoY in Q1'26 and never exceeded its Q1'25 high anywhere in 2025.

What the market must concede if we are right: that the FY2026 EBITDA "growth" it is extrapolating into a 5%+ margin path is a base-effect, and that 38× trailing EBITDA was paid for an inflection the unit economics do not yet show. Cleanest disconfirming signal: adjusted EBITDA/unit rising YoY at both Q2 and Q3 toward the $328 mark — that would mean the H2 beat is real drop-through, not comp math, and the variant is wrong.

Source: derived from AUTO1 quarterly adj. EBITDA ($68.3/49.7/61.0/53.1/70.3m) over units (204k/200k/219k/219k/249k); Catalysts tab. Adj. EBITDA/unit Q1'25 $328 vs Q1'26 $276 = −16% YoY; GPU roughly flat ($1,320 → $1,341).

The bull case is "each incremental car drops a widening slice of gross profit to EBITDA." The chart shows the opposite at the unit level: GPU is flat while EBITDA/unit faded over 2025, because Autohero capacity and brand-marketing reinvestment is absorbing the gross-profit growth. The honest counter (red-teamed below): the $328 Q1'25 comp was a high-water mark, so the YoY comparison flatters the bear here just as the easy comps will flatter the bull in H2 — which is exactly why the per-unit level, not the YoY headline, is the number to read.

Source: AUTO1 quarterly adjusted EBITDA, FY2025 (Catalysts / Financials tabs). Q1'26 lapped the strong $68.3m Q1'25 and grew only +3%; Q2–Q4'26 lap $49.7/$61.0/$53.1m, so a large H2 YoY headline does not require any per-unit improvement.

Disagreement #2 — "record profit, debt-free" hides a captive lender

What consensus would say: "AUTO1 turned its first profit ($92m), carries no corporate debt, sits on $700m+ of cash, and the negative free cash flow is just working capital funding growth — it flips positive whenever they choose."

Why our evidence disagrees: Two facts the headline narrative omits. First, $71.8m of FY2025 revenue is interest income (+77% YoY) off a $1.0bn instalment-purchase and merchant-financing book — finance income that is only "earned" if the loans perform, and whose ECL coverage, arrears and charge-offs are not disclosed. Second, the cash gap is structural and widening: CFO/NI is −6.9×, the accrual ratio rose from ~11% to ~21%, and $421m of the $710m cash is pledged (free cash ~$289m). The model is a used-car retailer bolted to a fast-growing, untested captive lender, refinanced on non-recourse ABS that begins amortising in early 2027.

What the market must concede if we are right: that a growing share of "profit" is credit spread of unknown quality, and that "debt-free" understates a real refinancing dependency (the April 2027 consumer-loan ABS roll). Cleanest disconfirming signal: the 2 Sep H1 report disclosing adequate, stable ECL coverage and an operating-cash-flow inflection — that would confirm the profit is real, reserved, and self-funding, and the concern shrinks to a footnote.

This is genuinely variant rather than a restatement of the forensics tab's "elevated" grade: the forensic work concludes the accounting is honestly presented. The variant claim is narrower and aimed at consensus — the market is pricing the credit book at zero risk and zero opacity discount while it remains the single least-disclosed driver of the just-achieved profit.

Disagreement #3 — volume is not pricing power

What consensus would say: "AUTO1 is the pan-European winner; share rose 2.5%→3.1% on +22% units, the field consolidated (Cazoo dead), and the moat is narrowing toward Carvana-like economics — worth 4.85× gross profit."

Why our evidence disagrees: A moat that conferred pricing power would show segment GPU rising as share rises. It does the opposite: Merchant GPU fell $1,148→$1,124 and Retail GPU $3,090→$3,002 YoY. Blended GPU rises only on mix (higher-GPU Retail becoming a bigger share of the count), not on power. With 2.4% EBITDA margin and ~7% ROIC (≈ cost of capital), the numbers are prima facie evidence of a narrow, not widening, moat — and the closest structural twin, Aramis, has sat near 1% operating margin for two decades under the identical vertically-integrated model.

What the market must concede if we are right: that 4.85× gross profit pays for a "narrowing-toward-wide" moat the segment numbers have not delivered, and that the fair anchor may be Aramis (~0.6× GP), not Carvana (~12×). Cleanest disconfirming signal: Merchant GPU bending upward YoY for two quarters while share keeps rising — the moat finally converting to price. This is the slowest-resolving of the three (multi-quarter to multi-year) and the most philosophical, hence ranked third; it is the structural why beneath #1.

The evidence layer — auditable fast

The best report-wide items that actually move the probability of the variant, each with its consensus read, our read, and its fragility (what could make the evidence misleading).

Source: Catalysts, Financials, Forensics, Moat, Long-Term Thesis and Web Research tabs. Consensus/variant reads and fragility assessments are the analyst's.

How this resolves — observable signals to put on a watchlist today

Every signal is observable in a trading update, half-year report, segment disclosure, or broker wire — no "better execution" or "time will tell."

Source: AUTO1 IR financial calendar (Q2 29 Jul, H1 2 Sep, Q3 4 Nov 2026); metrics disclosed in quarterly updates, segment reporting and the half-year report. Mapping is the analyst's.

Red team — the evidence that would break this view before the market does

A fair attempt to kill the variant, not protect it. Each of these is a real path to being wrong:

The $328 comp cuts both ways. Because Q2–Q4 2025 per-unit EBITDA was only $237–273, the YoY per-unit comparison is easy in H2 2026. If Autohero's reconditioning automation (CAT AI auto-detecting ~90% of damage) and finance attach ramp even modestly, adjusted EBITDA/unit could rise YoY in Q2 and climb in absolute terms — which would be genuine drop-through, not comp math, and would validate consensus. The variant explicitly loses if per-unit EBITDA prints above $328 at Q2 or Q3.

The cash burn is, mechanically, growth investment. Carvana ran the identical own-inventory-plus-captive-finance model, burned cash for years, then swung to +$1.04bn operating cash flow at scale. If AUTO1 ever throttles growth, working capital unwinds and FCF turns sharply positive — meaning the "profit isn't cash" concern is a timing artifact, not a quality defect. Disagreement #2 weakens materially the day OCF inflects.

The finance-book opacity may resolve benign. FinanceHero ABS placements were 3.6× oversubscribed at ~87bps — the market is funding this book cheaply and willingly. If the 2 Sep H1 disclosure shows adequate, stable ECL coverage, the credit-risk leg of the variant collapses to a footnote.

The moat is real and company-specific. Unlike Aramis, AUTO1 owns a genuine pan-European data-and-liquidity flywheel that took share through the 2023 price war (gross profit compounded while revenue fell 16%). "Value-first" can be a deliberate, harvestable land-grab — exactly the Carvana playbook — rather than a permanent commoditisation. If segment GPU bends up even once as share rises, disagreement #3 is on the back foot.

Consensus already half-concedes. Targets are being cut (JPM $48→42, RBC $34→29) and SWS DCF (~$31.5) sits below the $35.43 Street average — the market is not uniformly euphoric, so part of our "edge" may already be in the tape. The variant requires the further de-rate that revisions have only begun.

The single most important resolving signal

Watch adjusted EBITDA per unit, YoY, at the 29 July Q2 print — read it underneath the headline, not the headline itself. The units and gross-profit numbers will be strong and are largely priced; the YoY EBITDA growth will look big off the easy $49.7m Q2'25 comp. The one number that resolves the central disagreement is whether per-unit EBITDA rises above its prior-year level toward $328 (consensus is right — the drop-through is real) or stays stuck near $276–299 while volumes boom (we are right — the multiple is paying for an inflection that is not happening). Everything else on this page is downstream of that single observable.