ERO — Ero Copper Corp.

Deep-value report — Ero Copper Corp. (NYSE/TSX: ERO) — RESUMPTION

Analysis date: 2026-09-04 · Price: … (close 2026-09-03, yfinance) · Shares: 104,302,633 · Market cap: …B · Net debt (company definition, 06/30/2026): …M · EV: ~…B · Primary source: SEC filings from SEC-Filings\ERO\ (CIK 0001853860), including the 6-K from 08/05/2026 (financial statements, MD&A and the Q2 2026 release — downloaded today, unavailable at the prior analysis) and the 6-K from 09/01/2026.

REFRESH regime. The baseline thesis is the report from 2026-07-27 (rapoarte/deep/2026-07-27-deep-ERO.md). What follows explicitly states what’s carried over unchanged and what has been re-derived.

Executive summary

What is carried over unchanged from the 07/27/2026 thesis (verified that no falsifier has broken): the corporate structure (Canadian holding, 100%-owned Brazilian assets — Caraíba/Bahia, Tucumã/Pará, Xavantina/Mato Grosso, plus the 60% earn-in option at Furnas with Vale Base Metals); the absence of any price moat (a price taker on copper and gold); the FY2025 production-guidance miss history (… on copper); the Royal Gold gold stream; the quantitatively undisclosed Brazilian tax contingencies; the impaired Paranapanema receivable; and the debt structure (Senior Notes …M/6.5%/February 2030 + revolving …M/December 2028). Nothing in the Q2 2026 filing contradicts any of these points.

What has been re-derived from scratch, because the new filing moved the facts: (1) the trailing-12-month FCF bridge (Q3 2025 – Q2 2026), absent from the old report, which worked on the FY2025 fiscal year; (2) the owner-earnings base used in the DCF; (3) all five valuation models, at the new price; (4) the Quality of Earnings and CEO-profile chapters, absent from the old report.

An important factual correction. The mechanical delta pack (delta-ERO-20260903.md) reports “price at analysis” = … and a move of …. This is wrong: … is the close from 08/27/2026, not the price at the baseline report’s date. The price at the baseline report’s date was … (close 07/24/2026, exactly the figure cited in the old report; verified in the yfinance series: 07/24 = …, 07/27 = 26.20). The real move since the baseline thesis is thus … (… → …), not …. The trajectory: the stock traded sideways at …-27 through 08/04, rose to … on the day of Q2 results (08/05, published after close), to … on 08/07, hit a high of … on 08/25, and corrected to …–… in early September. The market paid for Q2, it didn’t sell it off — which also resolves the contradiction the research brief had left open (quiverquant “Stock Falls” vs tipranks “…”: the latter is correct).

The updated thesis. Q2 2026 is the company’s best operational quarter in history on nearly every line: revenue …M (… YoY), adjusted EBITDA …M, CFO …M (… QoQ), net debt down …M to …M (0.8x EBITDA), plus another …M repaid after quarter-end. Execution is real and verifiable: Tucumã grew plant throughput 27% QoQ, Xavantina recovered from Q1’s disaster (mined gold AISC from …/oz to …/oz), the OneEro program locked in …-15M/year of savings on renegotiated contracts and over …M on smelting/refining terms in 2026. Copper guidance is fully reaffirmed.

And yet the thesis remains negative, for a reason the old report hadn’t quantified. The FCF bridge was built on a 12-month basis and the result is severe: at a realized copper price of …/lb in Q2 2026 — an all-time high — the company generated, over the trailing four quarters, …M of simple FCF (CFO − capex) and only …M of free flow to shareholders after interest, financing fees and lease payments. That’s an FCFE yield of 2.79% on a …B market cap, at the peak of the price cycle. The cause is sustaining capex: …M over the trailing 12 months, up from …M (FY2024) and …M (FY2025), consuming 33% of adjusted EBITDA. Adding mine exploration (…M) and financing cost (…M net), ~…M of owner earnings remain at spot prices; normalized to a deck of …/lb copper and …/oz gold — still generous versus history — …M remains, i.e. 3.8% of market cap.

Estimated value. Five triangulated models, versus …: from … (EPV at the long-term price deck from the company’s own Furnas PEA, …/lb) to … (a 9x EV/EBITDA multiple on peak-cycle EBITDA, plus risked Furnas), with the median at … (DCF FCFE, Monte Carlo median over 20,000 scenarios, IV …/share). Monte Carlo gives 0.0% probability of undervaluation. Only the multiples models — which implicitly assume the market keeps paying for reserves, resources and the Furnas option — approach or exceed the price.

Verdict: MONITOR, unchanged from 07/27/2026, but with a significantly worse margin of safety: the price rose 36%, and my owner-earnings base fell from …M (old report, normalized NOPAT) to …M (the FCF bridge built now, which deducts real sustaining capex instead of assuming sustaining capex = D&A). The two moves point the same way. ERO is a good company at an excellent point in the cycle, at a price leaving zero protection if the cycle turns. It’s not a thesis to sell (it’s not a position), it’s not a buy — it’s exactly what the verdict says: to be monitored, with a clear trigger (see the last chapter).