The critical minerals bottleneck isn't the permit — it's the years of studies before it

In the US, a mine can take more than 30 years from discovery to production, and much of that is consumed by sequential engineering studies, not permitting red tape. A startup is betting on orchestrating these disciplines with AI to compress that timeline.

In an interview with MINING.com published in September 2026, Ari Rostami, CEO of mining technology company AiMinr, described a mismatch putting pressure on US critical minerals policy: while US national security operates on timelines of a few years, taking a mineral deposit from discovery to commercial production can take more than 30 years, adding up engineering studies, technical validation, financing and permitting. The deadline gets even tighter with two regulatory changes already confirmed: updated DFARS requirements take effect in January 2027, requiring materials used in certain defense applications to come from countries not aligned with China, and the US Department of War will ban, starting January 1, Chinese-origin rare earth magnets in defense systems covered by the rule. According to Rostami, US critical production infrastructure today has no way of meeting that deadline.

The core of Rostami's argument is that permitting reform, on its own, tackles only part of the problem. He estimates that between 30% and 40% of a mine's total development time is consumed by technical studies — preliminary economic assessment, pre-feasibility study, feasibility and detailed engineering — and not by regulatory red tape itself. Two structural problems explain this delay, according to him. The first is the reliance on engineering that is intensive in specialized labor, at a time when the mining industry faces a shrinking pool of experienced technical professionals. The second, even more structural, is that these studies usually run in a linear, siloed flow: geologists hand data to mine planners, who hand designs to process engineers, who hand them to finance teams — and when an important assumption changes midway, the effect cascades, forcing weeks or even years of rework in disciplines that had already moved ahead. Rostami cites cobalt as an example of the cost of this delay: the market expected growing demand for the metal driven by electric vehicles, which spurred interest in new supply, but battery makers increasingly shifted to cobalt-free chemistries such as lithium iron phosphate — by the time domestic cobalt mines finally reach production, the technology they were planned to serve may have already changed direction, or the market may have already been captured by competitors. The answer AiMinr is testing is not to speed up each discipline in isolation, but to connect all of them — geology, geotechnics, hydrology, ventilation, metallurgy and financial modeling — within a single platform, orchestrated by more than 50 specialized artificial intelligence agents, so that a change in an assumption, such as a commodity price or a mine design, propagates automatically across the connected disciplines instead of requiring each model to be manually rebuilt. One technical detail matters here: AiMinr separates the AI orchestration layer from the calculation engine. The agents do not generate engineering numbers on their own — they manage deterministic algorithms used in processes such as mass balance and mine planning, which makes the results traceable and allows them to be validated against the project's real data, instead of relying on a generative AI that merely 'looks' right. It is a distinction the mining industry, historically cautious about new technology, has been demanding: the company already works with a brownfield operation in Alaska using the platform to rework short- and long-term planning and resource and reserve estimation.

For those planning or evaluating critical minerals projects, the practical lesson is to redirect where acceleration efforts are focused: permitting reform remains relevant, but it does not on its own resolve the mismatch between national security timelines and mine development timelines, because a huge slice of the schedule lies in the technical studies that come before — and after — regulatory approval. For project managers, the case reinforces a lesson that applies beyond mining: a linear, siloed workflow across disciplines is, in itself, a source of delay as significant as any external barrier, because it turns any change in assumptions into cascading rework. And for those evaluating artificial intelligence adoption in the industry, AiMinr's model points to a more mature use of AI than generating text or images: orchestrating already validated deterministic engines, with traceable results — an application that may gain ground first in engineering and planning offices, before reaching the mine operation itself.