A composed read of the Afrimintel bankability layer at build v2.136.15. The counts and distributions below are recomputed from the structured feeds on every deploy, so they cannot go stale relative to the data layer; each carries its origin surface as a data-source deep-link and propagates that surface's three-state badge — Sourced, Derived, or Absent — unchanged, with no Absent filled and no Derived promoted in aggregation. One figure — the Correction-Velocity SLA — is a documented policy value, not a feed measurement, and links to where it is set. Feed vintages: bankability as of 2026-06-25 · capital provenance 2026-06-28 · country & EITI dataset v2.28.9.
What this report is for. A provenance-tracked map of the African mining assets that carry published, bankability-grade economics — built to be a defensible starting point for diligence: where the bankable universe sits, who controls it, and how trustworthy each figure is. It is deliberately a landscape, not a buy/sell recommendation, an asset ranking, or a substitute for an independent technical review. Its claim is narrow and testable: every figure here traces to a named source and carries its confidence state — follow the links and check.
How to read it. A figure's badge is the badge its source surface already carries — nothing is upgraded in aggregation and no Absent is filled. Sourced = named primary document. Derived = published Afrimintel method over Sourced inputs. Absent = named, never invented. Follow any figure's data-source link to its origin surface for the full per-asset basis.
1 · The bankability landscape
Composed live from the cross-asset screener. The screener segregates three hard-labelled confidence tiers: after-tax (DFS/PFS-grade, carried on a single 8% real post-tax basis and ranked only where a price-normalised value is published — 7 Sep 2026 ruling), pre-tax, and PEA/scoping-grade — pre-tax and PEA are never ranked against after-tax, never fed into any comparator. The 83-asset headline spans all three tiers; only the 68 after-tax are DFS/PFS-grade — the 6 pre-tax and 9 PEA-grade are lower-confidence and shown segregated below the headline.
Scope & selection — read first. This is a map of the assets with published, bankability-grade economics Afrimintel has verified — not a census of African mining. Coverage is therefore selected toward projects whose owners chose to publish a DFS/PFS/PEA (and toward listed issuers with a disclosure obligation); marginal, early-stage, privately-held and non-disclosing projects are under-represented by construction. The roster is a documented sample of the published-economics universe — it grows only as new African DFS/PFS filings clear the bar, not toward a fixed target.
83bankability-complete assets (verified, published economics)68after-tax tier (DFS/PFS-grade, npv8-common)6pre-tax tier (segregated)9PEA / scoping-grade tier (segregated)159assets on the roster33countries represented
Counts are computed from /api/v1/bankability.json using the screener's own tier predicates, reproducing the screener tally exactly. The bankability-complete count is a live feed measurement, not a target — it grows only as new African DFS/PFS filings clear the ≥2-source / three-state bar; there is no fixed denominator. Full filter/rank view: cross-asset screener ›
After-tax cohort — the ten largest after-tax NPVs on a common 8% basis
Asset
Country · commodity
Study (vintage)
NPV @ common 8% (npv8-common)
Goulamina
Mali · Li · production
Leo Lithium / Firefinch DFS Update (Dec 2021);
US$2,946mSourced
Kola
Republic of Congo · Potash (MOP) · development
Kore Potash Kola Optimised DFS update (Feb 202
US$1,985mDerived
Toliara (Vara Mada)
Madagascar · Mineral sands (Ti/Zr/REE) · development
Energy Fuels Toliara/Vara Mada FS (Jan 2026)
US$1,720mDerived
Assafou
Côte d'Ivoire · Au · development
Endeavour Mining Assafou-Dibibango DFS (23 Apr
US$1,715mDerived
Mahenge
Tanzania · graphite · development
Black Rock Mining eDFS update (FEED 2024; reaf
US$1,652mDerived
Kabanga
Tanzania · Ni-Cu-Co · pre-FID
Lifezone Metals FS TRS 2025
US$1,579mSourced
Ewoyaa
Ghana · Li · pre-construction
Atlantic Lithium DFS (Jun 2023) + DFS Update (
US$1,500mSourced
Kasiya
Malawi · rutile+graphite · pre-licence
Sovereign Metals Kasiya DFS (Apr 2026); supers
≤US$1,448mSourced
Orom-Cross
Uganda · graphite · development
Blencowe Orom-Cross optimised DFS (May 2026; s
US$1,411mDerived
Platreef
South Africa · PGM · operating
Ivanhoe Platreef 2025 Integrated Development P
US$1,400mSourced
This is a magnitude list, not a quality ranking — and an older study date is not, by itself, a stale-high signal.npv8-common normalises the discount rate only — not the commodity-price deck, study stage, or fiscal vintage. Each NPV is struck at its own study's commodity-price deck and is not re-cut to current spot, so whether a figure overstates or understates today turns on that study's price deck versus current prices, not on the study's age. Long-term decks are frequently set conservatively: the two lithium assets here use spodumene decks of US$978/t SC6 (Goulamina) and US$1,587/t SC6 (Ewoyaa) — long-run assumptions, not their 2021/2023 study-date spot — so their size is not an artefact of stale-high pricing. Assets span different commodities at different decks and are not cross-comparable. Each value carries the badge its source row carries — Sourced where the issuer reported at 8%, Derived where re-discounted by published method; ranking a Sourced figure against a Derived one does not equalise their confidence. Per-asset price deck, stage and basis live in the feed.
Operating vs. financeable — stage is shown on every row. Of the 68 after-tax assets, 22 are already operating or in production: for those the NPV is the issuer's pre-construction study figure, a historical record, not a forward case a financier could underwrite today. The remainder are pre-production (development / pre-FID / pre-construction stage) — there the NPV is closer to the case a DFI or lender would actually finance. A "bankability-complete" flag means the economics are present and provenance-named; it is not a statement that the asset is seeking, or open to, construction finance.
2 · Country & transparency layer
The Country Risk composite scores 40 jurisdictions (the 159-asset roster above spans 33 of them; the composite covers the wider set). Weighting: Fraser IAI 40 / TI CPI 30 / NRGI RGI 30, core-renormalised over present pillars; EITI membership a +5 bonus, not a weighted pillar. The composite re-normalises when a pillar is missing rather than penalising the gap — so a score built from fewer pillars sits on the same scale as a fully-populated one. Read the basis column.
Composite — top cohort
Country
Composite
Basis — pillars present
Botswana
Kalahari Platform
75
Fraser · TI · NRGI
Zambia
Lufilian Arc + Kalahari
65
Fraser · TI · NRGI
Ghana
West Africa Birimian
62
Fraser · TI · NRGI
Morocco
Saharan Metacraton
62
Fraser · TI
Senegal
West Africa Birimian
60
TI · NRGI no Fraser pillar
Rwanda
Karagwe-Ankole Belt (Kibara) + East African Rift flank
58low-confidence · single pillar
TI no Fraser pillar
South Africa
Kaapvaal Craton + Cape Fold Belt
57
Fraser · TI · NRGI
Tanzania
E. African Rift + Mozambique Belt
55
TI · NRGI no Fraser pillar
Namibia
Kalahari Platform
52
Fraser · TI
Ethiopia
E. African Rift + Afar
49
TI · NRGI no Fraser pillar
Comparability caveat (the basis column is load-bearing). Several top-cohort countries are scored without the Fraser Investment Attractiveness Index — the one mining-specific pillar — and one is a single-pillar score the feed flags low-confidence. A country scored on corruption-perception and governance indices alone is not evidentially equivalent to one with the full mining-attractiveness pillar, even at the same composite number. Treat the cohort as differing in evidential basis, not just in score.
Source: /api/v1/countries.json · full 40-country table and per-pillar inputs at /country/ · composite vintage and per-province freshness at /freshness/. Composite values are Derived under the published weighting; pillar inputs (Fraser, TI, NRGI) are Sourced to their primary third-party releases. Independence: Afrimintel does not set country scores — they are computed mechanically from named third-party indices under published weights; the editorial-leadership independence firewall is published at /methodology/independence/.
A structured EITI layer covers 28 African implementing countries (validation outcomes and disclosure detail, three-state) — a transparency-context layer, not a bankability score. Detail: /eiti/.
3 · The correction record
Afrimintel publishes its corrections. The track record logs 12 material corrections to date — including self-caught hallucinations, source-citation imprecisions, and fabrications caught on the hallucination-check layer — each with its drift class and the audit-log entry that closed it.
12logged corrections in the record7 daysCorrection Velocity SLA — material errors (business days)30 daysSLA — non-material (business days)
Full record with per-correction drift class and closing audit-log entry: /track-record/. The Correction Velocity SLA is defined in the governing Quality Standard · For journalists.
4 · Control & Capital Provenance
A neutral, factual map of controlling equity sphere across the bankability roster — the sphere of the largest controlling equity bloc and/or the operator. It is a record of who controls capital, not a judgment about it. Classifications are Derived from multi-source verification; nothing is invented.
Control sphere
Assets
Western
75Derived
Chinese
10Derived
Co-controlled
1Derived
African
5Derived
91classified, of 94 on the roster3held Absent for cause (data-layer only)
Caveat (propagated from the source surface). The Western control-sphere count INCLUDES 2 asset(s) subject to announced-but-unclosed Chinese acquisitions (ewoyaa, kurmuk) — currently Western-controlled, but classify by deal_status not control-sphere alone. If these close, the Western count falls and Chinese rises. Western share reflects public-disclosure culture, not the universe. Control and offtake are independent; offtake is tracked in the data layer and is not surfaced as a flagship finding while coverage is thin.
Why 3 assets are held Absent — and why this is disclosed, not hidden. Assets that fall under the platform's published independence firewall — where the editorial leadership holds a declared external affiliation — are recorded as ownership-fact only, with no editorial interpretation, and are held Absent on this judgment axis rather than scored. That is a deliberate conservative treatment to keep the analytic value while holding the conflict-of-interest line; it is not a data gap. The firewall and the affiliation it covers are published in full at /methodology/independence/. Per-asset ownership facts (including the held assets) remain on the data-layer axis surface; this report composes the distribution only.