Variant Perception

Where We Disagree With the Market

The market has compressed the Safran debate into a single decisive event — the 28 July H1-2026 receivables print — and treats the answer as binary. We disagree on the framing more than on the facts. The durable variable is the cash-conversion ratio of the aftermarket annuity across a multi-year window, not the one-period reversal of a €1.0B working-capital tailwind. The same compression shows up in three other places: consensus increasingly anchors valuation at the peer-median ex-GE (~13× EV/EBITDA), treating the gap to GE Aerospace at 33.9× as a US-listing artefact rather than as the credible cash-flow anchor that the contractual JV identity implies; consensus celebrates +26% Recurring ROCI growth as the compounding rate when the metric flowing to owners (adjusted profit attributable) grew just +3.5% in FY25; and consensus prices neither outcome of the CFM RISE / next-narrowbody architecture decision in 2027–2030, the single binary that sets the wide-moat rating beyond 2035. None of these are contrarian for performance — each is a place where the report's own evidence contradicts a specific market belief and where the disagreement resolves on observable signals.

Variant Perception Scorecard

A score of 55 reads as a genuine but nuanced edge, not a screaming mispricing. The consensus is clear (sell-side mean target €343 from 22 analysts on Moderate-Buy, the stock at 20.3× FY25 EV/EBITDA, debate framed around the late-July H1 print). The evidence supporting the variant views is strong but interpretive — the upstream tabs document the data, but the variant claims rest on how one weighs Recurring ROCI versus Adjusted EPS, the GE multiple versus the ex-GE median, and one print versus a 3-year window. Resolution mostly runs on a 12–24 month clock through the FY26 cumulative cash-conversion print, sell-side estimate revisions, and the architecture-decision flow on RISE.

Consensus Map

Consensus is most clearly observable on issues 1–3 — each has a direct sell-side, management, or price-level signal behind it. Issues 4 and 5 are inferences from price action and sell-side framing rather than explicit consensus statements, and the variant edge is correspondingly thinner. Issue 6 is sentiment-level, not analytical.

The Disagreement Ledger

Disagreement 1 — H1-26 is necessary but not sufficient. A consensus analyst would say: the late-July H1 print resolves the working-capital question, the multiple re-rates, the stock either trades to €343 (consensus target) on a clean reversal or compresses to €235-€245 on a miss. Our evidence disagrees because the upstream forensic tab is explicit that the conversion-ratio question requires a multi-year window: the 3-year FCF-after-M&A / Net Income ratio of 0.94 already trails the 1.17 ex-M&A read, and management's own 2028 ambition is built on cumulative 4-year conversion. If we are right, the market would have to concede that the H1 print is a noisy first reading, not the answer, and that the multiple should be set against a 3-year cumulative window. The cleanest disconfirming signal is a clean H1-26 reversal followed by normalized supplier-finance disclosure and a 2026 FCF print at the upper end of guide, two readings in the same direction — at that point the variant view ages out.

Disagreement 2 — GE is the right anchor. A consensus analyst would say: GE trades at 33.9× because of cleaner US reporting, no IFRS hedge mark-to-market noise, and structural US-listing premium; the right Safran multiple is the European peer median (~13–17×), and at 20.3× the stock is already at the top of that range. Our evidence disagrees because Safran and GE own 50% of the same CFM cash-flow stream by contract through 2050, and the underlying property right is identical — type-certificate-protected aftermarket on the LEAP and CFM56 installed base. The accounting and listing differences justify a multi-turn discount, not a 14-turn gap. If we are right, the market would have to concede that the JV cash-flow identity is the credible anchor and the peer-median ex-GE is the wrong comp set, which closes roughly half the gap (Safran toward 27× rather than peer-median 13×). The cleanest disconfirming signal is GE-Safran segment cash diverging through FY26-27 — if GE Aerospace cash grows materially faster than Safran Propulsion cash on the same JV, the property-right identity claim is weaker than we think.

Disagreement 3 — +3.5% is the rate flowing to owners, not +26%. A consensus analyst would say: the +26% Recurring ROCI growth is the compounding rate, the +3.5% adjusted profit number is depressed by one-time portfolio cleanup (SPI, Interiors), and the right normalized rate sits somewhere in between but closer to the recurring metric. Our evidence disagrees because the €423M Interiors impairment and €(244)M SPI capital loss represent real capital that was deployed (Zodiac 2018, €8.5B) and is now being written down — and there may be more (up to €1.5B Interiors review still open). The pattern of recurring "non-recurring" items above 1.5% of revenue is itself the forensic signal. If we are right, the market would have to concede that the rate at which Safran compounds owner capital is closer to +3.5–8% than to +26%, which implies a sustainable P/E closer to 17–20× than 30–35×. The cleanest disconfirming signal is a clean FY26 (and FY27) with no further excluded-loss items and Recurring ROCI / Adjusted EPS growth converging within 500bps.

Disagreement 4 — RISE is binary and existential, not deferrable. A consensus analyst would say: the architecture decision is 4–6 years out, the demonstrator is still in development, and the market values the franchise on the current LEAP installed base and the 2028 ambition. Our evidence disagrees because every upstream tab (Long-Term Thesis, Moat, Business) names RISE as the single existential moat variable beyond 2035. Pricing the franchise as if RISE is automatic (or as if a RISE loss has no impact) is internally inconsistent — both bull and bear cases implicitly assume RISE wins. If we are right, the market would have to concede that a probability-weighted RISE outcome implies a 5–15% expected-value haircut to fair value today, even at a high P(RISE wins). The cleanest disconfirming signal is an Airbus / Boeing public commitment to open-fan architecture or a CFM-named launch propulsion partner announcement before 2028, which collapses the binary to ~certainty.

Evidence That Changes the Odds

The most decision-relevant rows are #1 (the working-capital question is structurally multi-year, not single-print), #2 (the GE/Safran multiple gap is the cleanest mispricing signal in the entire deck), and #3 (the wedge between Recurring ROCI and Adjusted EPS is what the multiple is ignoring). Rows #4 and #6 are the live pipeline of evidence that strengthens or weakens disagreements #3 and the structural-drag thesis. Row #5 is the binary the market is not pricing; row #7 sets a fair boundary on disagreement #3 by acknowledging that the adjusted-EPS frame correctly excludes the non-cash hedge MTM.

How This Gets Resolved

Five of these six signals run on a 12–36 month clock — the variant framing is genuinely multi-year, not a next-quarter call. Signal #1 (cash conversion) provides interim reads on a 6-month cadence starting with the late-July 2026 H1 print, but the decisive cumulative read is FY28. Signal #2 (GE-Safran gap) reads in real time but the durable test is whether the gap compresses on consistent estimates over 12+ months, not on any single quarter. Signal #4 (RISE) is the longest-dated and the most binary; signal #5 (surtax) is the cleanest near-term test of the structural-cost-drag thesis.

What Would Make Us Wrong

The strongest disconfirming case is straightforward: a clean H1-2026 receivables print with DSO retracing below 145 days, supplier-finance disclosed under €500M, FY26 FCF landing in the upper half of guide, and the wedge between Recurring ROCI and Adjusted EPS narrowing to under 500bps as no further Interiors impairment lands — all in the same 18 months. In that world the market's compressed framing turns out to be correct: the H1 print did resolve the question, the FY25 working-capital stretch was timing, and the +26% Recurring ROCI was the right compounding rate because there is no further excluded-loss residue. Disagreement #1 ages out cleanly; disagreement #3 weakens to a footnote. The variant view that the market's framing is too compressed loses both its empirical and its conceptual edge.

The GE-anchor variant (disagreement #2) is the most vulnerable to a specific kind of fact pattern. If through FY26-27 GE Aerospace's segment cash grows materially faster than Safran's Propulsion segment cash, the contractual JV identity claim is weaker than we are stating — the same property right is producing materially different cash for the two partners, which would imply the gap reflects real economic substance rather than listing-and-accounting penalty. The data exists to test this; both companies disclose the relevant segment numbers quarterly. We should be honest that we have only the FY25 snapshot today and the through-cycle test is ahead of us.

Disagreement #4 (RISE) is the hardest to "be wrong about" in the near term because resolution is 2027–2030, but the symmetric risk is real. If the architecture-decision question itself never gets sharper — if Airbus and Boeing remain non-committal into the late 2020s — then a probability-weighted RISE adjustment to fair value today is largely academic for any underwriting horizon under 5 years. The variant view holds intellectually but stops being decision-useful, which is a soft form of being wrong.

The other place we could be wrong is more general: the consensus is not as compressed as we read it. Stan's verdict itself, in a sense, is already a variant view — "lean long, wait for confirmation" is not the consensus framing of "buy on the print." If institutional consensus is actually more nuanced than the sell-side preview anchoring suggests, then several of our disagreements are differences of degree, not of kind, and the variant strength score of 55 is itself too high.

The first thing to watch is the late-July 2026 H1-2026 receivables and contract-asset disclosure — not because it resolves the variant view (it does not), but because it is the next observable update to whichever side of disagreement #1 holds, and because the way the market reacts to it will reveal how compressed the consensus framing actually is.