Variant Perception

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

Variant Perception — Where We Disagree With the Market

The honest opening: there is no directional edge here, and pretending otherwise would be the mistake. On direction the report and the market agree — AUTO1 is a genuinely good business, the wholesale Merchant engine is profitable and moated, and Autohero's margin is the swing factor. That view is the consensus: 11 Buys against 2 Holds and 1 Sell, a mean target of $35.43 versus a $27.98 spot (~27% upside), after the stock already round-tripped +62% off its spring low into the June Capital Markets Event. Being bullish on AUTO1 is not a variant view; it is the crowd.

The edge is narrower, more specific, and mostly leans the other way. It lives in two measurable gaps between what the price now pays for and what the evidence supports:

1. Pace, not direction. The crowded long has underwritten the 2027 Autohero margin inflection as an on-schedule, on-magnitude event — a ~46% step in group EBITDA from the FY2026 guide to roughly $438m. The report's evidence says management deliberately made 2026 a "hold gross-profit-per-unit flat, buy volume" year and pushed the margin step into 2027, an inflection that has slipped before [1]. We sit ~8–11% below the Street on the decisive FY2027 number.
2. Denominator, not the multiple. The market pays a debt-free platform multiple for what is increasingly a securitization-funded balance-sheet lender whose reported profit it cannot yet bank — $92m of net income against −$544m of IFRS operating cash flow, with net debt at $1.60bn behind a "no corporate debt" headline.

Both disagreements are bearish-leaning into a name where the good news is banked and the down-move is asymmetric — the −20% single-session drop on the February 2026 margin guide is the base rate. The single signal that resolves it is the next print's Autohero adjusted-EBITDA-per-unit and group GPU — not the unit headline, which is already strong.

The variant, scored

Source: analyst scoring. Consensus clarity is high because the market belief is unusually observable (14-broker coverage, a dated target distribution, a clean estimate feed, and a violent price-reaction history); variant strength is held to the low-60s because the disagreement is about pace and quality rather than direction, and direction is where the crowd and the report agree. "Weeks to next resolving print" is the estimated window to Q2/H1 2026 results (late Jul-early Aug, from the prior-year cadence); the decisive read on the 2027 setup is the Q3 print (~early Nov 2026).

The score says: a real, monetizable disagreement of medium strength, with unusually clear consensus to push against, strong evidence, and a fast clock. It is not a deep-value "the market is wrong about the company" call — those are banned and would be false here. It is a "the market is wrong about when and about what kind of asset this is" call.

What the market believes — mapped to its signals

Every row below is a market belief nailed to at least one concrete consensus signal, then translated into the testable underwriting assumption the price is actually making. Where I disagree, the disagreement is with the assumption, not the company.

Sources: ratings, targets and the FY2026/FY2027 estimate path per the analyst-estimates feed (data/estimates/analyst_estimates.json; mean target $35.43, range $23.45-$43.57) and the Web Research tab; valuation multiples per the Web Research tab; the +62% rally and price-reaction history per the Current Setup & Catalysts tab; the "no corporate debt" framing cross-read to the Q1 FY2026 results presentation [2].

One nuance the table surfaces and the price ignores: the consensus EPS the multiple rests on is already drifting down, not up — FY2026 EPS has been trimmed to $0.73 from $0.76 ninety days ago, and FY2027 from $1.11 to $1.09 (analyst-estimates feed). The market holds the "compounding justifies the multiple" belief even as the numbers it depends on are being cut. That is the seam the variant view pulls on.

The disagreement ledger

Three disagreements survived all five tests (what a consensus analyst would say; what the evidence contradicts; whether it is material; whether an observable signal resolves it on the right horizon; what would prove us wrong). They are ranked by how much each would change a PM's underwriting. Each is classified against the high-quality buckets — wrong time horizon, wrong quality of earnings/denominator, wrong liquidity/implementation — and none is a banned "high quality but undervalued" or "market too pessimistic" form.

Sources: FY2027 modelling gap and the volume-over-margin framing per the Current Setup & Catalysts tab and the Q4/FY2025 call [1]; cash-flow and leverage evidence per the Forensics tab and the cash-flow statement [3]; float, ADV and borrow-data gaps per the Short-Interest tab.

Disagreement 1 — the FY2027 step is a deferred event the price treats as a dated one

What consensus says. The FY2021-to-2025 Retail per-unit glide — from −$4,644 to −$482 — is nearly complete, so FY2027 brings the crossover, the group steps to ~$438m EBITDA (EPS ~$1.09), and the platform re-rates. Mean target $35.43 embeds it.

Where the evidence disagrees. Management itself reframed 2026 as a deliberate pause: gross-profit-per-unit held "broadly flat compared to 2025," with the $57–86m EBITDA improvement "fueled by unit growth" rather than margin [1]. That moves the entire re-rating burden onto a 2027 step that the same per-unit data shows is the hardest, last mile — Retail is still −$482 per unit against a $1,663–$2,764 segment-EBITDA-per-unit target, and the loss is sensitive to the 20–40%-growth volume mix that dilutes margin [4]. The Catalysts work models FY2027 nearer $390–407m if the inflection lands late — ~8–11% below the Street — and the estimate feed already shows FY2027 EPS being cut, not raised.

What the market must concede if we are right. That the FY2027 ~46% EBITDA step is a forecast of an undated event, not a scheduled one — and that a GPU-flat year plus a soft European cycle can push the crossover past 2027, forcing out-year EPS down rather than up.

The cleanest disconfirming signal. Two consecutive quarters of Retail adjusted EBITDA per unit stepping toward or through zero while group GPU holds. That validates the Street's schedule and kills this disagreement. Two quarters of the per-unit loss widening while GPU stays flat confirms it.

Source: FY2025 actual $232m and FY2026 guidance $287-315m (midpoint shown) per the FY2025 trading update [5]; FY2027 Street ~$438m per the analyst-estimates feed; the FY2027 "our model" $399m is the analyst's variant estimate carried from the Current Setup & Catalysts tab.

Disagreement 2 — a lender priced as a platform; profit the market cannot yet bank

What consensus says. AUTO1 is a debt-free, self-funding platform; the ABS facilities are matched, non-recourse working-capital plumbing, so the deeply negative IFRS operating cash flow is an accounting quirk and the platform multiple is fair.

Where the evidence disagrees. The audited record shows reported profit and cash moving in opposite directions: $92m of net income against −$544m of IFRS operating cash flow in FY2025, with the cash position held only because +$560m of financing draws refilled the burn — a pattern structural since FY2022 [3]. Net debt is $1.60bn (D/E 2.26x, from 0.92x in FY2022), and the "No Corporate Debt" slide that anchors the bull narrative excludes $1.05bn of inventory-ABS liabilities that were already 87% drawn at Q1 2026 [2]. The Forensics tab grades the gap Elevated (57/100) precisely because the company's headline "AUTO1 Cash Flow" of +$232m sits $777m above the −$544m IFRS figure. And the company's own FY2022 report names liquidity "the most relevant potential financial risk," conceding it needs continued bank and capital-market access until it reaches positive operating cash flow [6]. Layered on top, the captive-finance book — receivables up ~50% in 2025 — already produced a $13.9m credit charge management called a one-off, on a portfolio young enough that a "one-off" is hard to distinguish from an early read on reserve adequacy [7].

What the market must concede if we are right. That a meaningful and growing slice of reported profit is credit-derived and funding-dependent, that the right denominator is $1.60bn of net debt rather than zero, and that the multiple should carry a discount for reliance on continuously open, cheap ABS.

The cleanest disconfirming signal. Group IFRS operating cash flow turning positive without fresh ABS draws for two or more quarters — the company's own "self-funding" claim made real. That, plus a captive-finance book that seasons without a second charge, refutes this disagreement.

Source: reported net income and IFRS operating cash flow, FY2022-FY2025, per the Forensics tab and the cash-flow statement [3].

Disagreement 3 — analytically right, institutionally hard

What consensus says. A $6.1bn cap covered by 14 brokers is a liquid mid-cap that can be sized and exited at will.

Where the evidence disagrees. Only 23.2% of the share capital was placed at IPO; founders (~21%) and SoftBank (~14.8%) sit on a concentrated register, the true free float is thin, and average daily turnover is only ~$16m (~0.75m shares). The Short-Interest tab is explicit that no aggregate short-interest, short-sale-volume or borrow data exists for this Xetra name — so a PM cannot currently size the cost or feasibility of the short leg, and a sponsor block sale into a thin float is an asymmetric-down technical the analytics ignore. This is the one disagreement that is not about being more bearish — it is about a thesis being correct and still difficult to express at institutional size.

What the market must concede if we are right. That the implementation constraint — float, borrow, overhang — is a real input to expected return for this name, not a footnote.

The cleanest disconfirming signal. BaFin/Bundesanzeiger net-short publications above 0.5%, an announced placement or block, or realized impact-cost on the tape that shows the float absorbs size cleanly.

The evidence layer — what a PM can audit fast

These are the items that actually move the probability of the variant views, each paired with the consensus read, our read, and — critically — its fragility, the thing that could make the evidence misleading.

Sources: estimate path and revisions per the analyst-estimates feed and Web Research tab; the 2026 GPU framing cross-read to the Q4/FY2025 call [1]; cash-flow and ABS-draw evidence per the Forensics tab, the cash-flow statement [3] and the Q1 FY2026 presentation [2]; the $13.9m credit charge per the Q1 FY2026 call [7]; float and borrow gaps per the Short-Interest tab.

How this resolves — observable signals only

Each signal below is observable in a filing, an earnings call, the tape, an estimate revision, or a regulatory docket — never "better execution" or "time will tell." The top two resolve durable thesis variables; the rest inform.

Sources: the resolving-evidence definitions track the long-term thesis spine - Retail per-unit economics [4], the cash-flow/ABS framing [3] [2], and the captive-credit charge [7] - per the Long-Term Thesis, Forensics and Current Setup & Catalysts tabs; estimate path per the analyst-estimates feed.

Red team — what would kill these views before the market does

This is written to break the variant, not to protect it.

The funding disagreement has its own kill switch. The capex intensity is genuinely ~0.3% of revenue, the trading (car) business is substantially self-funded by matched non-recourse ABS, and a fresh consumer-loan securitization (FinanceHero-2, $278m, Sep 2025) proves the ABS market is open and pricing AUTO1's paper. If FY2026 operating cash flow turns positive on its own as inventory growth normalizes, the "lender priced as a platform" framing loses most of its force — and the company's "self-funding" claim becomes the more accurate description than ours. The audited statements are unqualified, with no restatement, probe, or auditor flag in the record; this is a quality-of-earnings discount, not a fraud flag, and a clean season of cash conversion removes it. Finally, the tradability disagreement is the most easily neutralized of the three: if no block sale materializes and the float absorbs flow, it stays a footnote rather than a cost.

The intellectually honest read is that all three disagreements are bearish-leaning into a name whose business has genuinely inflected — which is exactly the setup that punishes a stubborn short. The variant view is not "AUTO1 is a bad business." It is "the price has underwritten the best version of the next 18 months, on schedule, on a platform multiple, and the evidence says the schedule and the denominator are both more uncertain than the multiple admits."

The one signal to watch first