AFMJF — Alphamin Resources Corp.

Alphamin Resources Corp. (AFMJF / TSX-V: AFM / JSE AltX: APH) — deep-value analysis, REFRESH

Date: September 8, 2026 · Reference price: … (OTC, AFMJF) = … (TSX-V, FX rate 1.3834) · Shares: 1,287,774,479 · Market cap: … Regime: REFRESH over the thesis from 17.08.2026 (rapoarte/deep/2026-08-17-deep-AFMJF.md), triggered by a single mechanically unverifiable falsifier. Reporting and pricing currency: USD (no FX gap at the flow level; the primary listing, however, is in CAD — see the price note in the Summary) Primary source: audited consolidated financial statements FY2025 and FY2024 (IFRS, signed 11.03.2026 and 17.04.2025), unaudited interim statements at 06/30/2026 (signed 30.07.2026), MD&A FY2025 and Q2 2026. Alphamin does not file with the SEC (a Canadian issuer on SEDAR+, traded OTC in the US as AFMJF); the SEC-Filings\AFMJF\ folder doesn’t exist and can’t exist — edgar_10k_downloader.py correctly returns “CIK not found.” Re-checked today: the folder is missing, which is the expected state, not a run error.

REFRESH regime note — what I inherit and what I re-derived

The diff package delta-AFMJF-20260907.md declared REFRESH for the reason: “1 unverifiable falsifiers.” The falsifier in question is Pre-mortem 1 (“tin enters a new structural regime and stays above …/t”), whose metric — year-over-year revenue growth over two consecutive quarters — can’t be evaluated because Q3 2026 hasn’t been reported; the next reporting is scheduled for November 3, 2026 (tracker, next_reporting field). The other five falsifiers all hold, without exception: price … (threshold < 0.75), net margin 0.250–0.268 (threshold > 0.40), net debt/EBITDA −0.208 (threshold > 0.50), current ratio 1.68 (threshold < 1), market cap 1.378bn (threshold > 1.5bn).

New filings: zero. The last filing remains the interim set at 06/30/2026, signed 30.07.2026, already fully exploited in the 17.08 report. I haven’t reread the filings folder.

Inherited unchanged, because no new filing touches them: the business structure and grade moat (3.29% Sn); FY2025 unit economics (AISC …/t, on-mine …/t produced, off-mine …/t sold); the cost model calibrated on two real points (off-mine = 3,433 + 0.0909 × price; on-mine …/t going forward; sustaining capex …/t; volume capped at 20,000 t/year); all balance-sheet, receivables, inventory, debt, and accruals tables in the Quality of Earnings chapter; the capital-allocation history; the earnings-quality verdict; and the three pre-mortem scenarios as structure. Adjusted net debt remains +… (… DRC tax payable minus … net cash declared at 06/30/2026): the staggered provisional payments in July and September reduce cash and the obligation simultaneously and by the same amount, so they don’t move the net figure.

I re-derived three things, and for each I state whether it’s a FACT or an OPINION move:

1. The Monte Carlo simulation — mandatorily redone, the frozen artifact was broken. The mc-AFMJF-20260907.json file in the diff package was generated with --oe … and --nd …, i.e. absolute values where mc_dcf.py requires millions (ap.add_argument explicitly documents “base owner earnings (mil, flow currency)”). Result: a median intrinsic value of …/share and a median MOS of ** …**, with … probability of undervaluation. It’s not a usable artifact, it’s a unit error three orders of magnitude off reality. I reran it correctly; the new file is mc-AFMJF-20260908.json and it’s the source of the DEEPMC line at the end of the report.
2. The DRC tax function — structurally re-derived. A FACT move, forced by the tin price. The 17.08 report used a linear interpolation of the effective rate between two observed points: 33.4% at a realized price of …/t (FY2025) and 45.0% at …/t (H1 2026). The spot tin price has since risen to …/t (LME cash settlement, 07.09.2026) — i.e. outside the range the fit had been calibrated on. Extrapolated there, the linear interpolation implies a marginal Congolese-state take of 86% of each incremental EBITDA dollar, which contradicts the mechanism in the law. I replaced it with the structural model (base rate 33.4% on gross profit + a 20pp super-profit tax on the portion of EBE above the price threshold), which reproduces the old table up to …/t with under 0.4% error and diverges only above the threshold, where the linear fit over-taxed.
3. The 10% withholding on intra-group dividends — a correction of an omission in the previous report. An OPINION move, declared as such. Note 8 of the H1 2026 statements identifies … of withholding tax on dividends upstreamed from ABM (DRC) to Alphamin, versus zero in H1 2025. Divided by the actual dividend upstreamed (… total ABM × 84.14% = …), the rate comes out to 10.08% — exactly the dividend withholding rate under the DRC’s 2018 Mining Code. The mid-cycle projection in the previous report was calibrated on the FY2025 effective rate, a year when this withholding was zero, and then applied to a scenario where the company distributes everything — hence it missed a permanent 10% leak on the upstreamed flow. The correction lowers mid-cycle owner earnings from … to … and is the main reason the verdict hardens even as the share price has fallen.

What I did not change, though I could have: the mid-cycle tin price remains …/t, the base discount rate remains ** …**, g1 remains …, gt remains …, mine life remains 9 years (the figure from the audited closure provision). The argument for moving them would have been the spot price; but spot price is exactly the variable I refuse to capitalize as permanent, and raising the assumptions would have meant building the thesis around the price move — precisely what the refresh regime forbids.

Executive summary

The thesis in brief, unchanged in substance and harder in the numbers: Alphamin is an excellent business at a price that leaves nothing on the table, and in the meantime I found a permanent 10% leak on the money’s road from the mine to the shareholder, which the previous valuation missed. Bisie is the world’s richest tin mine (3.29% Sn processed grade), operated at an AISC that was …/t in FY2025 in a market where spot is …/t today. TTM EBITDA of … USD on a market cap of … means EV/EBITDA 2.63x on an adjusted net-debt basis. The problem isn’t the asset — it’s that the market is paying the spot tin price as if it were permanent, over a mine life the company itself declares to be nine years, and as if everything that comes out of the mine reaches the shareholder in Toronto.

What actually changed in the three weeks since the prior thesis — and it matters that the answer is “almost nothing at the company, something at the commodity.” Tin hasn’t gotten cheaper: LME cash rose from ~…/t (mid-August) to …/t on September 7, with a peak of … on the 21st and 24th of August. The share price, by contrast, is flat in the primary listing currency: … on August 17, … today, after an all-time high of … reached on August 10. The … decline noted in the diff package is an artifact of the thin USD OTC quote and CAD depreciation (the rate moved from ~1.373 to 1.3834), not a company re-rating. I don’t build the thesis on this move, nor read it as a signal.

Three figures hold up the whole thesis, the same as in August, recalculated. First: the realized price rose from …/t (FY2024) to … (FY2025) to … (Q2 2026), while volume is completely flat — 5,045, 5,016, 5,014 tonnes sold across three consecutive quarters. Second: the DRC’s super-profit tax triggers above …/t in 2026 and above …/t after, adding 20 percentage points on the portion of EBE above the threshold; at today’s spot, of every incremental EBITDA dollar above …/t, the Congolese state takes 57 cents, ABM minorities another 7, and the Alphamin shareholder keeps 35. Third: the audited closure provision uses 9 years of remaining life — the flows are a short annuity from a depleting asset, not a perpetuity.

The flow bridge, updated. The definition used in the DCF is mid-cycle FCFE = (EBITDA at …/t tin − base tax 33.4% on gross profit − super-profit tax − total capex, interest, and lease principal) × 84.14% (ABM interests) × 90% (the DRC withholding on dividends) = … Mechanical CFO − capex gives … TTM from filings — 57% above. The divergence isn’t at interest (which at Alphamin is correctly within operating, verified line by line: “Interest paid” 3,102,418 in FY2025, 1,328,800 in H1 2026), but in three places no automatic bridge sees: … of ABM dividends to minorities, … of lease principal, and … of accrued and unpaid DRC tax, plus the 10% withholding now discovered.

Estimated value: …–1.23/share, with the median of the five triangulated models at … Monte Carlo over 20,000 scenarios gives a median intrinsic value of …, a median MOS of ** …, and only ** … probability the stock is undervalued (versus …% in August), with a 0.8% chance of finding a 30% safety margin.

Verdict: AVOID at … The interest zone falls from 0.75 to ~… Not because the business is bad — it remains one of the best base-metal mines in the world on unit economics, with clean accounting and frugal management — but because at today’s price you’re buying three simultaneous assumptions: tin at all-time highs for nine years, a Congolese state that never again changes the tax regime, and a reserve that expands without any technical report yet documenting it. Checked today: there is no new NI 43-101 technical report — the most recent remains the one from August 2022, for Mpama North.