Variant Perception

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

Where We Disagree With the Market

The sharpest disagreement: the market is paying a marketplace multiple for a business whose economic engine is migrating into a balance-sheet-leveraged consumer/dealer financier, and the right denominator is not Adj EBITDA but cash earnings net of the $607M of net new debt it now takes to fund a year of growth. Consensus anchors on FY26 Adj EBITDA of $291-320M and walks to PTs near $38-43 (~45-62% upside vs $26.54), with overwhelmingly Buy ratings and a quant short-pile (~4.22% disclosed) that is factor-driven rather than thesis-driven. The evidence in the report says three specific things consensus is not pricing: (i) FY25 produced $92M of net income against -$570M of FCF — a 14-year unbroken pattern that just got worse in the "profitable" year, (ii) the AUTO1.com network-effect headline of 60,000+ partner dealers is actually 36,200 active per the first Q1 26 disclosure (a ~60% activation rate), and (iii) the Autohero retail leg — the segment the market is paying a D2C re-rating premium for — has no durable moat in a region where Mobile.de/AutoScout24 own buyer attention and EU consumer-credit rules cap the finance attach that saved Carvana. The 17 June 2026 Capital Markets Day is the venue where the first two of those gaps either widen or get partially closed — if Fintech is broken out and Autohero stand-alone economics print clean, our variant view weakens; if either is left consolidated and opaque, it widens.

Variant Perception Scorecard

The 62/100 variant strength reflects an honest tension. Consensus is unusually clear (13 analysts, JPM $43 anchor, Berenberg lone Hold, factor-only quant short) and the evidence behind our disagreement is documented in the upstream tabs (forensic CFO/NI -5.94x, active-dealer disclosure, Autohero structural caps). What knocks the score back from 75 is dating: the strongest disconfirming evidence — segment economics and Fintech disclosure — sits inside a 12-day window (CMD 17 June 2026) and the next cash-flow print is 89 days out (H1 26 report 2 Sep). The variant view is right or wrong in a hurry; this is not a slow-burn thesis.

Consensus Map

The map is consistent enough to underwrite three specific bets against it. Rows 2 (balance-sheet framing), 3 (Autohero re-rating logic), and 4 (network-effect headline) are the three implied assumptions where the upstream evidence ledger pushes hardest in the other direction; row 1 (operating cadence) and row 5 (cycle position) are the bull/bear tension Stan already documented and where we have nothing additive to say.

The Disagreement Ledger

#1 — Wrong denominator. Consensus models a marketplace re-rating off Adj EBITDA: FY26 $305M midpoint × 17-22x forward EV/EBITDA gets you to PTs in the $38-43 cluster. The evidence says the business that earns the EBITDA is increasingly a leveraged financier — $1.70B of non-recourse ABS plus a $375M dealer-financing book, with $607M of net new corporate debt funding the FY25 working-capital cash gap. Credit businesses do not trade at 17-22x earnings; specialty finance and securitisation issuers trade at 10-12x earnings or 1.0-1.5x tangible book. If the market has to concede that even a third of the business deserves the credit-cohort multiple, the marketplace re-rating closes against a credit-cohort fade; the cleanest disconfirming signal is H1 26 OCF (2 Sep 2026) printing better than -$291M alongside FY guide reaffirmation, which would mean the FY25 burn was a one-off inventory build rather than a structural feature.

#2 — Wrong segment for the multiple. The moat tab is unambiguous: AUTO1.com Merchant has a narrow but real moat (high-proof on pan-European wholesale liquidity, medium on the pricing engine and ABS architecture), and Autohero is rated not yet a moat. The bull/sell-side underwrite Autohero as the next leg of the re-rating — Carvana retail GPU runway, Fintech attach. But Autohero faces three structural caps the upstream evidence is explicit about: Mobile.de / AutoScout24 own the DACH buyer entry point (no proprietary traffic source), aided brand awareness is 35% versus CVNA's 70% built over 12 years, and EU consumer-credit rules cap the finance attach that earned Carvana ~50% of GP. If we are right, what the market would have to concede is that the segment paying the high multiple has no moat and the segment with the moat trades at the marketplace floor. The CMD on 17 June 2026 forces the question by disclosing historic segment economics for the first time — if Fintech is left as a footnote and Autohero stand-alone economics print thin after marketing allocation, the variant view widens; if Fintech is broken out as ≥10% of GP with credit metrics and Autohero GPU after CAC is ≥$2,794, we are wrong.

#3 — Headline network-effect overstated by ~60%. The Q1 26 disclosure of 36,200 active buying partners against a 60,000+ registered base is the single piece of new information from the past quarter that has not propagated into consensus models. Sell-side notes still cite 60k. The moat tab flags this directly as a partial refutation of the network-effect claim; the bull/bear shared tension table names $1,094 Merchant GPU as the tripwire on the same moat. What we would have to see to be wrong: the active-dealer count climbing toward 40k by FY26 with Merchant GPU holding $1,106+ through two more quarters and CMD disclosing supplier mix stable. What confirms us: two consecutive quarters of active count below 35k, or CMD supplier-mix disclosure tilting away from off-lease toward lower-margin C2B as OEM-direct remarketing eats the highest-quality stock.

#4 — Technical asymmetry not priced. This is the lowest-conviction variant in the sense that it does not change long-term underwriting — but it materially changes implementation. An ~87-session cover-clock on a $13M-per-day tape, with hard-dated CMD inside 12 days and consensus PTs 45% above spot, means the move size in either direction is likely to exceed the fundamental information content of the print. The signal worth watching is direction of Bundesanzeiger ≥0.5% disclosures inside the 14 days surrounding CMD; either-way trims read as the smart-money tell ahead of the print.

Evidence That Changes the Odds

The seven items are not independent. The cash-quality lens (rows 1 and 6) tells one story; the moat-quality lens (rows 2, 3, and 7) tells the second; the management-signalling lens (row 4) confirms both; row 5 is the inventory mechanic that links them. The honest read is that an institutional reader can dismantle any single item — but the pattern across them is the variant view, not any individual fact. If H1 26 OCF improves AND active dealers climb AND Merchant GPU stabilises AND CMD breaks out Fintech with clean credit metrics, all four disagreements collapse together. None of those individually refute us. All four together do.

How This Gets Resolved

The resolution path is densest in the next 89 days — the same window the catalysts tab flagged for the bull/bear debate, but read through a different lens. CMD answers disagreements #2 and #3; H1 26 OCF answers #1; the Bundesanzeiger trajectory answers #4. Nothing in the variant view requires us to wait a year. The FinanceHero spread tell (row 4) and the Mobile.de / AutoScout24 threat (row 6) are the slower 6-12 month resolutions that update the long-term version of the call.

What Would Make Us Wrong

The cleanest disconfirming case is straightforward and we should price it honestly. If the 17 June Capital Markets Day discloses Fintech as a third pillar with $117M+ of stand-alone GP and a credit-quality framework, and the H1 26 cash-flow print on 2 September shows operating cash flow narrowing to better than -$291M without any off-balance-sheet ABS reclassification, two of our four disagreements collapse inside 90 days. That is a genuine plausible outcome — Wallentin's CFO profile, the +10.7% QoQ growth in consumer-finance ABS, and the directional improvement in DSO (60.3 → 35.8 days FY24 to FY25) all point in that direction. We could be wrong on cash quality and wrong on the segment-mix story at the same time, and the variant view would have to fold back to "the technical asymmetry was the only real disagreement."

We should also take seriously the possibility that we have over-weighted the 36,200 active-dealer figure. It is a single quarter's disclosure with no industry-comparable activation-rate benchmark. Used-car wholesale auction platforms typically have long-tailed registered bases with most volume concentrated in 30-40% of accounts — the AG1 number may sit inside the normal range for the asset class, not below it. If Q2 26 prints 38,000+ active without any other tells deteriorating, our network-effect compression argument loses force. The moat-tab evidence ledger explicitly calls this out: "new disclosure; baseline could be seasonal; needs four quarters to read trend." We are betting that the first signal is the signal.

A third honest concession: the multiple framework on disagreement #1 may overstate the variant. Even if a quarter of the business deserves a credit-cohort multiple, blending 75% marketplace × 17-22x and 25% credit × 10-12x lands closer to 15-18x rather than the 12-15x we used as anchor for the downside case. The mathematics of mix change favour us less than the narrative of mix change implies — the consensus multiple is wrong, but it may be wrong by 2-4x, not 6-10x. That tightens the materiality without removing it.

Finally, we should respect the alternative narrative consensus implicitly underwrites: that AG1 is one of three companies on the planet — alongside CarMax and a repaired Carvana — that have proven the integrated B2B/D2C/Fintech used-car model at scale, and that the right multi-year outcome is a re-rating off cash-conversion proof rather than current-cash quality. If CMD on 17 June presents a credible 4-6 year framework with a bridge from 2.4% to 4%+ margin and Fintech as a third pillar, the burden of proof flips back to us — and our variant view becomes the marginal-PM position rather than the central one. Six months from now, the variant could read as right on direction, wrong on magnitude.

The first thing to watch is the 17 June 2026 Capital Markets Day Fintech disclosure — if Fintech is broken out with quantified GP contribution and credit-quality metrics, our cash-denominator argument weakens and the platform re-rating logic survives; if Fintech is left bundled or footnoted, the variant view widens by the day.