BVN — Compania de Minas Buenaventura S.A.A.

Deep-value analysis — BVN (Compañía de Minas Buenaventura S.A.A.) — REDO

Analysis date: September 12, 2026 · Reference price: …/ADS (yfinance, 09/12/2026; tracker … on 09/11; the Monte Carlo simulation used its own reading, …) · Market cap: … · 253,715,190 common shares outstanding + 744,640 investment shares, 21.6 mil. in treasury via Condesa · NYSE: BVN, BVL: BUE.LM · 1 ADS = 1 common share · Reporting and trading currency: USD.

Regime: REFRESH. The reference analysis is from 08/21/2026 (rapoarte/deep/2026-08-21-deep-BVN.md, verdict DO NOT BUY at …). The mechanical deep_delta.py triage decided on a redo based on a single new filing — a 6-K from 08/26/2026 it couldn’t classify.

What I explicitly inherit, since nothing beneath it has moved: the entire O’Glove balance sheet analysis (the source is the 2025 20-F, filed 04/30/2026 — no new annual report has appeared); the FY2025 FCF bridge, which I nonetheless re-verified line by line in the text and confirm; the ownership structure and governance; the accounting red flags list; the valuation assumptions (oe 640, g1 3%, r 10%, gt 1%, nd −…), which I re-tested against the new figures and keep; the CEO’s Outsider score, with two point corrections explained in place.

What I re-derived, and why: (1) the filing that triggered the redo — the 6-K of 08/26/2026 — is an Investor Day announcement with no financial content; I say this upfront, so no one thinks the thesis changed because of it; (2) I read the operational press releases of 04/16/2026 and 07/16/2026, which the August analysis hadn’t used in detail, and found there a 47% cut to San Gabriel’s production guidance, a material fact; (3) I found the … dividend approved on 03/30/2026 and paid on 05/05/2026, absent from the previous report, which changes the reading of the capital allocation chapter and resolves a question left open there; (4) I recalibrated the Model 5 EBITDA to prices realized in the third quarter; (5) I ran Monte Carlo over 20,000 scenarios with the same assumptions.

The price move (35.47 → …, … from the reference analysis on the same source) is context, not the reason for the redo. I build nothing around it. The whole sector fell over the same interval: AngloGold …, FCX …, PAAS …, Hecla ….

Primary sources: the FY2025 20-F (SEC-Filings/BVN/BVN_20F_2026-04-30.txt, CIK 1013131); 6-Ks from 03/30/2026 (AGM resolutions and the dividend), 04/16/2026 (1Q26 production and volumes, 2026 guidance), 04/29–30/2026 (1Q26 results), 07/16/2026 (2Q26 production and volumes, revised guidance), 07/30/2026 (2Q26 results) and 08/26/2026 (Investor Day) — all downloaded from EDGAR during this session and saved as .htm + .txt in the local folder; data pack data-pack-BVN-20260912.md; delta package delta-BVN-20260911.md; simulation mc-BVN-20260912.json; the tracker database (analiza_extra, evaluari, scoruri_criterii).

Executive summary

The redo was triggered by an empty filing. The 6-K of 08/26/2026 announces Buenaventura is holding an Investor Day on November 10, 2026 in New York, to mark thirty years since the NYSE listing, and that management will discuss “how the company continues to strengthen its portfolio.” It contains no figures. I state this explicitly, because the REFRESH regime might otherwise suggest something happened: nothing did, in the filing that triggered the redo. The only usable information in it is a calendar date — 11/10/2026 — which is, however, the first structured communication event the company has organized in the period analyzed and a plausible venue for capital allocation announcements.

What did emerge instead, from reading the operational press releases that the August analysis hadn’t unpacked quarter by quarter, are three facts that matter more than the filing that required the redo.

First: San Gabriel’s production guidance was cut 47% in three months. In the 04/16/2026 release, the company published for the first time an annual 2026 guidance of 48,000–55,000 ounces of gold at San Gabriel. In the 07/16/2026 release, the same guidance became 25,000–30,000 ounces, with the explanation “tailings management constraints and current metallurgical recovery rates.” In the first half the mine produced 4,460 ounces, so to reach even the low end of the revised guidance it must produce 20,540 ounces in the second half — 4.6 times the first-half pace. Total direct-operations guidance fell from 133.0–153.0 thousand ounces to 120.0–139.0 thousand, a smaller cut (…) only because Orcopampa (42–47k → 48–53k) and Tambomayo (5–6k → 9–11k) were revised upward. The old mine is covering for the new mine. August’s pre-mortem gave the “San Gabriel never reaches plateau” scenario a 25% probability; now it’s at 35%, and the reason is a company-published figure, not an intuition.

Second: the company distributed … in May 2026 — a fact absent from the August report. The general meeting of 03/30/2026 approved a dividend of … per common share (and … on the investment share), with a record date of 04/21/2026 and a payment date of 05/05/2026. On 274,889,924 common shares that means … declared, of which … actually leaves the group (the rest returns to treasury shares held via Condesa). Relative to 2025’s attributable profit (782.1 mil.), the payout ratio is 34.8% — 13.6 times the 2024 dividend and 2.3 times everything paid in 2025. This also resolves a question the previous report left open: why the cash position stayed flat between 03/31/2026 (759.9 mil.) and 06/30/2026 (758.9 mil.). The answer isn’t “capex and tax payments,” as I had assumed; it’s a … dividend paid on May 5, fully covered by the quarter’s flow. The second quarter generated, by difference, operating flow on the order of …–400 mil.

Third: metal prices have corrected, and the correction hits exactly the metal that matters most. Third-quarter average to date: silver …/oz versus … in H1 (…), gold … versus … (…), copper …/lb versus … (…). In the 2Q26 sales structure, calculated from the volumes and realized prices published by the company, silver is 46.4% of gross metal sales, copper 31.2%, gold only 18.2%. The net effect at constant volume is … on revenue, i.e. quarterly revenue of ~… instead of … and EBITDA from direct operations of ~… instead of …

The thesis remains the one from August, and the five falsifiers I formulated then all hold: leverage is −0.57x (threshold: above +0.3), FCF yield 4.6% (threshold: above 12%), P/B 1.99x (threshold: below 1.4x), ROE 23.6% (threshold: below 20%), price 33.88 (threshold: below 26). None has broken, so the thesis doesn’t retreat.

Estimated value: … – …/share, median …, from five triangulated models — a range identical at the ends to August’s, since the extreme models don’t depend on the facts that moved, and with Model 5 recalibrated downward (… → 34.63) on third-quarter realized prices. Monte Carlo over 20,000 scenarios gives a median intrinsic value of … and an undervaluation probability of …% — higher than August’s …%, but purely because the price fell 4%, not because value rose.

Verdict: DO NOT BUY at … HOLD if a position exists. The argument is unchanged and has strengthened on one point: today’s price simultaneously assumes a precious metals cycle at first-half-2026 levels (already refuted by the third quarter’s silver price), a San Gabriel that reaches plateau (already refuted for 2026 by the company’s own guidance), and a … tax receivable fully recovered (already lost at every ordinary court). The buy threshold remains …