Date: 2026-05-22 Position status: 10 shares @ EUR 62.25, filled 2026-05-18 (T1 of 2) Companion docs: Inflection thesis · Execution log · Hormuz explainer
Yes, Arkema is a good buy at EUR 62 — but it is not a no-brainer, and the way it can disappoint is more interesting than the way it can fail.
The thesis is structurally sound: four independent value drivers, two of them (assets + dividend) that work even if the other two never show up. Probability-weighted expected return is +40 cents per EUR deployed over three years — better than VWCE's ~+26c, but not dramatically so, and the bear case (30% probability) returns roughly zero excluding dividends.
The position size (EUR 627.40, ~28% of the active Graham satellite) is right-sized for this conviction level. The kill-condition T2 plan at EUR 52 is the right correction to the RI.PA mistake. The honest worry is not that we lose money — the floor holds — it's that AKE compounds at ~7% while VWCE compounds at ~8% and we spent four years being early.
That is a defensible outcome to risk. Read on for why.
The screen tells most of the story before any narrative:
| Metric | AKE.PA | What it says |
|---|---|---|
| Price / Book | 0.77x | Paying 77 cents for every EUR of book value |
| Forward P/E | 10.5x | Market pricing in negative real growth |
| Dividend yield | 5.67% | Income while waiting, 35% payout — well covered |
| Net debt / EBITDA | 2.13x | Declining YoY, no covenant pressure |
| Margin of safety | ~23% | Discount to book before any growth assumption |
Forward P/E of 10.5x against a Graham "no-growth" anchor of 8.5x implies the market expects only ~1% earnings growth in perpetuity. For a EUR 4.6B specialty chemicals leader with US PVDF and Singapore PA11 plants about to come online, that is a low bar. Anything that doesn't catastrophically miss expectations is upside.
Q4 2025 EBITDA hit a EUR 74M trough — the worst quarter since 2020. Then Q1 2026 snapped to EUR 283M (+283% sequential). Even if you assume half of that snap-back was the Hormuz acrylic windfall (more on that below), the other half is broad-based recovery across Bostik (adhesives), Coatings, and Advanced Materials — segments with no exposure to a Persian Gulf chokepoint.
The pattern fits the chemicals destocking cycle: customers over-ordered in 2022, sat on inventory through 2023-25, and are now reordering. Once the cycle turns, it tends to run for 18-24 months.
This is the part most analyst notes miss. Arkema isn't a single-thesis stock. It has four largely independent value drivers:
The bear case requires all four to fail simultaneously. The base case requires only Pillar 1 and Pillar 4. The bull case is the upside if Pillars 2 and 3 also work.
Thierry Le Hénaff has run Arkema since the 2006 TotalEnergies spin-off — 20 years. He executed the Bostik acquisition (2015) and Ashland adhesives (2022) without blowing up the balance sheet, and his "specialty over commodity" pivot is what turned a refinery byproduct business into one Europe can plausibly defend against Chinese overcapacity. 20-year tenures in European industrials are rare; he's the closest thing AKE has to a Buffett-style "the CEO is the moat" figure.
The three-scenario expected value with conservative weighting:
| Scenario | Probability | 3yr Return | Contribution |
|---|---|---|---|
| Bull | 45% | +75% | +33.75c/EUR |
| Base | 25% | +28% | +7.00c/EUR |
| Bear | 30% | 0% | 0c/EUR |
| Total EV | +40.75c/EUR |
Compare to VWCE's ~+26c/EUR over three years. AKE beats the index by ~14 percentage points in expectation — but with materially more variance, and most of that edge comes from the 45% bull-case weight. Push bull probability down to 35% and EV drops to +30c/EUR, barely beating the index.
This is not a fat-pitch trade. It's a positive-edge trade.
The most-thesis-aware analyst is Berenberg, who downgraded to Hold at EUR 62 on April 20, 2026 — exactly at our entry price. Their argument is precise:
This is not a wave-of-hand bear case. It's a specific, falsifiable model with a clear catalyst (Hormuz vessel traffic data) and a near-term timeline. Barclays is even more bearish at EUR 47 target.
Per [[shr-002]], systematic insider buying is one of the strongest bullish signals available. AKE has the opposite: zero open-market purchases by C-suite or board in the last 12 months, and the only confirmed transaction is a deputy GM selling 7,602 shares at EUR 52 in November 2025.
Absence of buying is not contrary evidence — Le Hénaff may simply have full conviction without needing to add — but it is a missing confirming signal at a moment when management's pay-grade peers (a French insurance CEO, the AIG CFO whose buy at $86.54 we tracked into AIG) are buying. The execution log treats this as an active RED, and that's right.
Arkema doesn't have an ATM ([[shr-025]]), bond-for-equity ([[shr-037]]), or convertible debt overhang ([[shr-039]]) — share count is roughly flat. That's a real strength.
But it does have EUR 600M/year of growth capex against a EUR 4.6B market cap — roughly 13% of equity value reinvested every year into PVDF, PA11, and Bostik integration. If the end-market demand for those products underperforms (Northvolt bankruptcy already cost them a customer; Chinese cell makers are vertically integrating PVDF in-house), that capex becomes a stranded asset and FCF stays compressed for years longer than the bull case assumes.
This is the same "capex into a structurally weak end-market" risk that destroyed European chemical commodity producers in the 2010s. It's lower-probability for Arkema (PVDF demand is real, even if pricing isn't great), but it's not zero.
The thesis treats the four pillars as independent. They're not, fully:
Treating four correlated pillars as independent is the same mistake [[shr-008]] flagged in biotech DDs: convergence of evidence that shares an underlying model is not the same as true independence.
The bear case isn't "lose money." Even in the 30%-probability bear scenario, dividends pay roughly 17% over three years and the stock holds EUR 50-55 — a flat to slightly positive total return.
The real risk is opportunity cost. If we hold AKE for three years and it returns 0%, VWCE returns ~26%. That's a 26-point underperformance on EUR 627 = ~EUR 163 of foregone diversified return. For a satellite position in a EUR ~2,200 portfolio, that's material but not catastrophic.
This frames the question correctly: AKE is not a "could lose 50%" trade. It's a "could underperform VWCE by ~25%" trade. That's the actual downside being underwritten.
The inflection thesis confidently assumes 2027 EPS at EUR 7.50 (bull case). If Berenberg's 10.3% downward revision is right, that's closer to EUR 6.70 — still above the EUR 5.94 consensus, but multiple-compression to 11x puts the stock at EUR 74, not EUR 90. The bull case stock target is materially sensitive to whether the acrylic windfall lasts through 2027 or fades in H2 2026.
The bearish analyst notes focus heavily on the acrylic Hormuz risk and the construction demand weakness. They under-discuss the PVDF US and PA11 Singapore ramps — capex that is already spent and just needs to produce. If even one of those two plants ramps on schedule with reasonable pricing, it adds EUR 50-100M of incremental EBITDA in 2027 that the bears aren't modeling.
This is the asymmetry that makes the position defensible: the bear case requires multiple specific negative outcomes to print; the base case requires only that the cyclical recovery continues at all.
EUR 627 (T1) is ~28% of the active Graham satellite portfolio, with EUR 200-300 (T2) reserved for monthly contribution deployment if the price falls to EUR 52 with no kill conditions triggered. That's a maximum exposure of ~EUR 900, or ~40% of the satellite.
If the bear case prints, EUR 900 → ~EUR 750 (after dividends, before T2 averaging). If the bull case prints, EUR 900 → ~EUR 1,500. The position sizing matches the conviction level — meaningful but not concentrated, with structural cash flow (dividends) reducing variance.
The shr-034 lesson — "size smaller, don't refuse to enter, when multiple reds are active" — is correctly applied here. T1 wasn't deferred; T2 is conditional on price weakness without thesis break.
Yes, but with three honest caveats.
The 30% bear-case probability is non-trivial. If you're not comfortable with a 3-in-10 chance of zero return (positive only after dividends), this is the wrong position. The bull case being 45% is the assumption doing most of the work — and as noted, dropping it to 35% reduces edge to ~+30c/EUR, barely above VWCE.
No insider buying remains an unresolved flag. Per [[shr-002]], this is the missing confirming signal. The position is justified in spite of it, not because of it. If Le Hénaff or the CFO files an open-market purchase in the next two quarters, the conviction goes from "defensible" to "high." If we get to Q3 2026 with no insider buying and Hormuz has normalized and Q2 results are soft, the thesis is materially weaker than at entry.
The opportunity cost vs. VWCE is the real downside. The bear case isn't "lose money." It's "underperform the index by ~25 percentage points over three years." That is the actual risk being underwritten. For a satellite position with positive expected value and structural diversification benefits (European industrial, low correlation to VWCE's tech-heavy top), that risk is acceptable. For a core position, it wouldn't be.
Upgrade to high conviction (would justify T2 ahead of EUR 52 trigger):
Downgrade to exit (would override the four-pillar floor):
"Do you have a view on whether destocking is ending?"
The Q1 2026 numbers, the management commentary on Bostik and Coatings, and the broader chemicals-sector data say yes — and the bear case (Berenberg) doesn't actually dispute the cyclical recovery, only the acrylic windfall portion. The argument is about how much of the recovery is real, not whether any of it is.
That's enough conviction to justify the position. Not enough to bet the portfolio. Which is exactly what we did: ~28% of the Graham satellite, with structural downside protection, with a sized T2 plan, with the EUR 36 dividend already on the calendar.
Final verdict: Arkema at EUR 62 is a good buy for a Graham satellite position. It is not a good core holding. The position size correctly reflects this distinction.
Monitoring this position over the next six months reduces to a small number of measurable signals:
| Signal | Source | Threshold | What it tells us |
|---|---|---|---|
| Hormuz daily vessel traffic | Kpler / industry | Above ~1,000/month | Spread normalization coming — Pillar 2 fade |
| Acrylic acid spot prices (Asia/Europe) | Platts / IHS | Below $300/t | Pillar 2 over; check if Q2 still beats without it |
| Q2 2026 EBITDA (late July) | Company release | Above EUR 250M = on track; below EUR 200M = kill condition | Direct test of broader recovery |
| Berenberg / Barclays / JPM revisions | Sell-side notes | Multiple upgrades | Sentiment re-rating starting |
| Le Hénaff / CFO insider buys | AMF DOC filings | Any open-market purchase | RED flag flips to GREEN — high-conviction signal |
| PVDF US / PA11 Singapore ramp commentary | Quarterly calls | On-time, on-spec | Pillar 3 execution |
| PFAS Pierre-Benite lawsuit | Court filings | Above EUR 200M judgment | TRIM signal |
Quarterly re-score is sufficient. The position is built to be held through cycles, not traded.
Playbook rules invoked in this analysis: [[shr-002]] (insider signal), [[shr-008]] (correlated evidence streams), [[shr-012]] (Graham IV sensitivity), [[shr-013]] (cyclical tranche entry), [[shr-014]] (PEGY for dividend payers), [[shr-016]] (fundamental stops, no price stops), [[shr-020]] (same-day red flag check), [[shr-025]] / [[shr-037]] / [[shr-039]] (dilution-mechanism trio, not applicable here), [[shr-034]] (smaller T1 with multiple reds — applied as "full T1, structured T2"), [[shr-019]] (never sell VWCE core).