KoBold's AI copper bet: Mingomba discovery becomes a $2.3B Zambian mine

KoBold Metals is a California-based metals-exploration company that builds AI models of the Earth's crust — a 'Google Maps' for ore — to find copper, lithium, cobalt and nickel, and is now developing Mingomba, its first mine, in Zambia's Copperbelt.

AI can beat mining's discovery odds and compress the 15-plus-year path to production, so KoBold commits billions to build its AI-found Mingomba copper mine.

TechCrunch (2025-01-02) reported the $537M Series C, closed in December 2024 and led by Durable Capital Partners and T. Rowe Price, at a valuation near $2.96B (per the Financial Times). Mingomba's capital cost is estimated above $2.3B (Mining Indaba, 2026-05-05); Kurt House said at the groundbreaking that about $300M had been invested in Zambia so far, with ~$600M expected by end-2026 (Zambian Business Times).

KoBold Metals was founded in 2018 (TechCrunch) and is led by global CEO Kurt House and president Josh Goldman; CNBC Africa describes it as backed by Bill Gates, Jeff Bezos and others, using AI as 'a Google Maps of the Earth's crust'. The company acquired Zambia's Mingomba project in December 2022 and used its AI tools to delineate a high-grade copper orebody; in February 2024 it announced the discovery, with Goldman calling Mingomba 'Kakula-scale in size and grade' (CNBC Africa).

In December 2024 KoBold closed a $537M Series C led by Durable Capital Partners and T. Rowe Price — funding exploration on five continents and the Zambia mine — at a valuation near $2.96B (TechCrunch, Financial Times). By March 2026 Africa CEO Mfikeyi Makayi told Reuters the company had started mine development and was 'full swing' into permitting, targeting production in the early 2030s. In May 2026 KoBold broke ground on Mingomba with Zambia's president: an 80% KoBold / 20% ZCCM-IH joint venture expecting 300,000 tonnes of copper a year at about 5% grade — ten times the global average — with the orebody about 1,700 metres deep (Mining Indaba; Zambian Business Times).

As of 2026-09-02 Mingomba is under construction — KoBold's first-ever mine, with more than 600 people at the site and shaft sinking underway. KoBold expects about $600M invested in Zambia by end-2026 and a final cost estimate after engineering studies in early 2027, with production targeted for the early 2030s; analysts quoted by Mining Indaba call the move from explorer to operator 'a step-change in risk profile'.

Discovery economics are brutally bad — roughly three in 1,000 potential deposits become viable mines (TechCrunch) — so even a small improvement in targeting has outsized value.

AI can encode decades of scattered geological, geophysical and geochemical data that no human team can hold in one head, ranking ground by orebody probability before drilling.

A Tier-1 find like Mingomba — called Kakula-scale, the benchmark set by the DRC's giant Kamoa-Kakula mine — justified keeping the upside by funding the mine itself instead of selling the discovery.

Energy-transition demand for copper gave investors a reason to fund supply that might otherwise take 15+ years from discovery to production.

De-risk the whole funnel, not one drill hole: better targeting pays only when a Tier-1 find justifies compressing the 15-year mine timeline — depth, water and cost decide if geology cooperates.

As of 2026-09-02 KoBold is building Mingomba, its first mine: groundbreaking took place in May 2026 on the 80/20 joint venture with ZCCM-IH, projected to produce about 300,000 tonnes of copper a year at roughly 5% grade in the early 2030s, with more than 600 people at the site (Zambian Business Times). The orebody lies about 1,700 m deep, capital cost is estimated above $2.3B with a final figure due in early 2027, and ~$600M is expected invested in Zambia by end-2026 (Mining Indaba). KoBold has never operated a producing mine — the execution test analysts call a step-change in risk.