KoBold Prefers Ownership Over Software
KoBold Metals
This model makes KoBold behave less like a software vendor and more like a high risk, high upside mineral owner. Instead of charging a mine operator for maps or drill recommendations, it uses its models to decide where to buy or earn into ground, then keeps a direct stake in the discovery. That means a successful target can turn into a project sale, a farm down, a royalty stream, or decades of mine cash flow, but it also pulls KoBold into the capital intensity and timetable of mine building.
-
The key contrast is with software peers like Terra AI, which sell decision support to operators through enterprise contracts. KoBold uses similar subsurface modeling to influence drilling decisions, but monetizes by owning part of the orebody itself, so the reward is much larger if a deposit works and zero if it does not.
-
Its deal structures are built around ownership, not licenses. In Zambia, KoBold partnered with ZCCM-IH on Mingomba, where public materials describe KoBold holding 80% of the project and advancing it into shaft construction. In other cases, earn in agreements give KoBold equity through staged exploration spending rather than upfront acquisition.
-
That ownership approach changes what scale looks like. A software company can add customers with limited balance sheet strain, while KoBold must keep funding drilling, studies, and eventually mine construction. Mingomba alone has been described publicly as requiring around $2 billion of investment, which shows how quickly exploration upside turns into project finance exposure.
The next phase is a move from proving that AI can find deposits to proving that AI led discoveries can survive the long path to production. If Mingomba reaches the early 2030s as a producing mine, KoBold will have shown that exploration software can be a wedge into owning large scale mineral assets, which is a much bigger economic prize than selling tools to miners.