Variant Perception

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

Where We Disagree With the Market

The market is using EV/EBITDA of 9.6× against a 20-year median of 5.2× to call KGHM "top of cycle" — but consensus FY26 EBITDA of roughly $1.9bn is the wrong denominator. Three mechanical changes already in the record will land in FY26 reported EBITDA without requiring a higher copper price: silver hedges struck at half of current spot are scheduled to roll off; the December 2025 minerals-extraction-tax amendment lets capex offset the cash levy from January 2026; and Sierra Gorda has flipped from a capital sink to a sub-$1/lb C1 producer that has repaid more than $1bn to the parent since 2021. Sized together, those changes add $0.7–1.1bn to FY26 EBITDA versus a consensus print sitting near the FY25 reported figure. Forward EV/EBITDA on the corrected denominator is closer to 6×, near the 20-year median, not 1.85× it. The disagreement resolves between 13 May 2026 (Q1 print, first observable cash-tax read) and end-Q2 2026 (strategy + dividend + AGM), so the underwriting window is short.

Variant Perception Scorecard

The 70/75/72 scores are calibrated to a single fact: the variant view is mechanically testable on three discrete FY26 line items, each with primary-source evidence already disclosed in FY25 financials. Consensus clarity is high because multiple data vendors converge on a flat-to-down FY26 EBITDA print near $1.9bn — a clean foil for the variant. Resolution is fast: the Q1 print on 13 May 2026 is the first observable cash-extraction-tax read, and the end-Q2 strategy/dividend window forces the new board to commit a number to FY26 EBITDA. The disagreement is not whether the stock works on a higher copper price — it is whether the next two earnings prints reveal an EBITDA denominator the consensus model does not yet contain.

Consensus Map

The most testable line is the first one. Vendor consensus FY26 EBITDA at roughly $1.9bn is below FY25's $2.86bn Adjusted EBITDA print — meaning the street is implicitly modelling a contraction, not an expansion, into FY26. That contradicts management's own commentary on hedge roll, the December tax amendment, and Sierra Gorda's continued contribution. The other rows triangulate the same setup: a market that has rerated the stock on copper but has not rerated the operating engine.

The Disagreement Ledger

Disagreement #1 — the wrong denominator. Consensus would say 9.6× EV/EBITDA at 1.85× the 20-year median is a "sell the rally" setup, and the average sell-side PT of $83.91 (only 3% above spot) reflects exactly that read. The evidence disagrees because the multiple's denominator does not contain three FY26 changes that have all been disclosed in FY25 financials: the hedge mark-to-market liability (Forensic tab), the December 2025 capex-deduction amendment (Business + Bull + Catalysts), and Sierra Gorda's structural cost reset (Forensic + Bull). If the variant view is right, the same 9.6× compresses to roughly 6× on FY26 reported EBITDA, and the 20-year median lens flips from a sell signal to a buy signal. The cleanest disconfirming signal is the Q1 cash-extraction-tax line on 13 May — if the line is flat YoY at roughly $314m quarterly, the capex deduction is cosmetic and one of three legs of the variant view collapses.

Disagreement #2 — the wrong time horizon on capex. Consensus would say negative cumulative FCF across FY21–FY25 is a structural disqualifier, and the dividend cut to zero in FY24 and FY25 is the proof. The evidence disagrees because the capex peak is a discrete build window, not a steady-state level. Three ventilation shafts (PUZ) plus the Żelazny Most expansion are fixed-duration projects scheduled to complete in 2027–28; capex/D&A at 15.2% versus D&A at 6.5% confirms the spend is growth, not maintenance. If the variant view is right, FY27 capex steps back to roughly $0.9–1.1bn, OCF rises on hedge roll, and FCF turns durably positive — making the 30%-of-net-profit dividend policy feasible at the policy ceiling rather than as an aspirational gesture. The cleanest disconfirming signal is the FY26 capex envelope in the end-Q2 2026 strategy presentation; envelope above $1.5bn with no completion year for PUZ would invalidate the discrete-window framing.

Disagreement #3 — Sierra Gorda is the segment, not the side-show. Consensus and the bear case both treat the $808m equity-method line as one-time accounting flatter, and both bundle the recurring share with the discrete impairment reversals. The evidence disagrees because the cash trail is real: $1bn+ cumulative parent repayments since 2021, $331m in 2025 alone, FY25 C1 of $0.86/lb beating SCCO, and a 4th-grinding-line FID landing by end-Q2. Strip only the explicit reversals ($70m + $72m) and run-rate JV profit is close to $668m — roughly 44% of pretax, not 53.6%, but still structural rather than ornamental. If the variant view is right, the 4th-line FID approval at the scoped $700M envelope shifts the JV's SOTP value above $4.5bn (versus roughly $3bn implied today). The cleanest disconfirming signal is 1H 2026 equity-method profit ex-reversals dropping below $268m — the bull's own pre-disclosed exit trigger.

Disagreement #4 — political alignment, not political churn (low conviction). Consensus says three CEOs in three months under Minister Balczun proves permanent political risk. Our weaker counter-read: the Paszkiewicz installation is the alignment event, not a step in a churn cycle, because every other major Polish SOE under the same minister cycled once and held. If correct, the governance discount compresses 1–2x EV/EBITDA over 6–12 months. We grade this Low confidence because Polish SOE precedent is short and the KNF investigation is still live; this is a sweetener, not a fulcrum.

Evidence That Changes the Odds

The first three rows are load-bearing. Each one independently flows into FY26 EBITDA in a way the consensus number does not contain. Rows 4 and 5 establish that consensus is both quantitatively below management's adjusted FY25 print AND internally split — a market that hedges its own conviction reprices faster than a unified one. Rows 6 and 7 reinforce the second-tier disagreements; row 8 confirms the resolution timing is binary near-term, not gradual.

How This Gets Resolved

The first three signals — Q1 cash levy, dividend recommendation, end-Q2 strategy — collapse into a single 8-week window between 13 May and end-June 2026. By the time the H1 audited financials publish in late August, the variant view is either largely confirmed or largely broken. The signals are clean because each one is a number management cannot avoid quoting in regulated disclosures, and each is mapped to a specific note or schedule. Resolution does not depend on the LME copper tape; it depends on three FY26 line items that the consensus EBITDA model does not yet contain.

What Would Make Us Wrong

The variant view is a denominator argument, and its biggest failure mode is that the denominator works as we expect but the numerator (enterprise value) compounds faster. A copper correction to Goldman's FY26 base of $4.50/lb would clip roughly $268m of EBITDA — the same order of magnitude as the hedge-roll uplift, and enough to neutralise the variant. Equally, the December 2025 minerals-tax amendment is real, but the implementing rules are unpublished; if Polish authorities define "qualifying capex" narrowly, the cash-tax saving could land at the bottom end of the $268–402m range or below. We would not know this until the Q1 print.

The Sierra Gorda leg has its own specific weakness. The FY25 C1 of $0.86/lb depends heavily on the molybdenum by-product credit, and management cut FY26 molybdenum guidance by ~50% in December 2025 — a fact the bull case underplays. If molybdenum credit shrinks materially while copper softens, JV C1 could drift back above $1.00/lb and the "structural mix shift" framing loses its sharpest edge. We would still expect the JV to remain above $268m ex-reversals, but the SOTP-uplift case would compress.

The capex-window argument is the slowest to resolve. We claim PUZ + Żelazny Most are discrete and capex normalises post-2027–28. But Polish SOE projects have a long history of multi-year overruns, and the new strategy could quietly add KGHM 2.0 (across-the-Odra) or Victoria construction capex that keeps the envelope at $1.5bn through 2030. If the end-Q2 strategy quantifies forward capex without a clear step-down year, the discrete-window framing is wrong and the bear's "permanent capex drain" critique stands.

Finally, the entire variant view assumes the next political shock is at least 24 months away. A KNF finding against the Company itself, a fifth Management Board dismissal under Minister Balczun, or an explicitly capital-allocation-driven Treasury vote at the June 2026 AGM would all reset the governance discount higher and overpower a clean operating print. We have graded political-alignment Low confidence for exactly this reason; the variant view does not depend on it being right, but it does depend on it not getting materially worse.

The first thing to watch is the Q1 cash extraction-tax line on 13 May 2026 — if it does not move materially below the FY25 quarterly run-rate of $326m, the lead leg of our variant view collapses on the first observable data point and the underwriting case becomes much harder.