Constellium (CSTM) reported Q1 2026 Adjusted EBITDA of $359M — a +93% YoY beat that annualizes to ≈$1.4B versus full-year 2025 of $846M. The stock is +209% over the prior twelve months and fell -2.95% on the print. The question is what, if anything, in the print is unpriced.

What the filing says

• Q1 2026 Adj EBITDA $359M vs $186M Q1 2025 (+93%); ex-lag $262M (+78%).
• Average Midwest Premium $2,296/t vs $712/t Q1 2025 (+222%).
• Metal price lag (non-cash tailwind) $97M vs $39M Q1 2025 (+149%) — accelerating, not reversing as the prior 10-K narrative implied.
• P&ARP segment EBITDA/ton $578 vs $223 (+159%), explicitly attributed to "favorable metal costs at Muscle Shoals and Neuf Brisach."
• Three explicit references to "current supply shortages of automotive rolled products in North America" — identical language to Q4 2025.
• New $300M buyback authorized March 12, 2026; three-year tenor to Dec 2028; replaces the $79M remnant program (down from $106M at year-end).
• FPI → US domestic issuer transition complete Jan 1, 2026; this is the first 10-Q under the new status. No covenant violations, no new risk factors.

What the market thinks

EV/EBITDA ≈4.5-5x annualized run rate, in line with peer cohort multiples that have already repriced (CENX +262% 1Y, KALU +172%). The ATM August straddle implies ±22.7% into the Q2 print; risk-neutral P(>$35) = 44%. Reverse-engineered, the current price is consistent with FY 2026 Adj EBITDA of $1.26-1.40B and cycle-aware multiples that compress into peak EBITDA. Probability-weighted FY26 = $1,240M. Roughly matched.

Why the gap is narrower than it looks

The cohort signals are corroborated cleanly. AA Q1 10-Q reported MWP +215% YoY (within 3% of CSTM's +222%). KALU Q1 10-Q showed metal price lag $36M vs $21M (+71%) — same direction. Ford Q1 10-Q: "in the second half of 2026, we expect to partially recover the production lost to date," with $1.5-2.0B FY26 Novelis cost guidance. STLD's Millett confirmed a 1.4Mt structural deficit and 60-70k tons Q2 ramp into the gap. None of this is unpriced.

The unpriced piece is narrower than the headline suggests: P&ARP EBITDA/ton +159% at Muscle Shoals versus KALU's conversion margin +3.8%. CSTM is capturing a scrap-discount-to-MWP arbitrage at scale that peer rollers don't share. Combined with the new $300M buyback (8-9% of float through 2028) and FPI institutional flows, ≈30-35% of the $173M YoY EBITDA gain is CSTM-specific.

But variance decomposition pushes back. Peer-basket regression (CSTM ~ AA + CENX + KALU): 57% idio, KALU β=0.49 dominant. SPY+XLB spec: 64.7% idio. All specs fail the 75% threshold. Long-outright at current prices pays full price for cohort beta to capture an idio kicker that is a minority of variance. Probability-weighted across four MWP scenarios (regime persists 35% / partial normalize 35% / rapid 10% / second leg 20%) at cycle-aware multiples gives 12-month E[r] = -2.6%, σ = 28%.

Risks (ranked)

1. MWP normalization H2 2026. Novelis Oswego hot mill restart end-Q2; Bay Minette ramp H2. If both hit and tariffs ease, MWP compresses to $1,200-1,500/t and the lag tailwind reverses into 2027.
2. Cohort beta dominance. Long-outright is ≈65% bet on already-consensus aluminum cohort.
3. Muscle Shoals one-quarter risk. Need Q2 print to confirm $578/ton P&ARP isn't a windfall.
4. Tape sold the print. -2.95% on a +93% EBITDA beat indicates profit-taking pressure.

Catalysts

• 2026-05-15 (~) — CENX Q1 earnings (MWP cross-validation)
• 2026-05-21 — CSTM AGM, $300M buyback approval
• 2026-Q2 end — Novelis hot mill restart target
• 2026-08-15 (~) — CSTM Q2 print (Muscle Shoals durability test)
• 2026-09-30 — September MWP threshold
• 2027-Q1 — FY 10-K resolution

What would change our mind

• P&ARP EBITDA/ton sustains ≥$500 in Q2 (idio thesis hardens; outright position reconsidered).
• Pullback to $26-27 with RSI <40 (margin of safety widens materially).
• Insider Form 4 Code P buying post-print (currently absent).
• KALU cools from RSI 76 (relative value vs. peer basket improves).
• Bay Minette ramp delay announced (gap extends into 2027, MWP elevation duration extends).

Evidence