A model is only as good as its data, and for Roman Teslyuk, founder and CEO of Earth AI, the data was arriving far too slowly. His company uses machine learning to hunt for critical minerals such as copper, platinum, and palladium in underexplored regions of Australia. The algorithms have been adept at flagging promising zones, including areas long dismissed by traditional geologists. But turning those digital predictions into hard geological facts has been bottlenecked by an old-world constraint: lab capacity.

As Earth AI ramped up drilling to test its targets, the volume of rock samples surged. Commercial laboratories, already straining under a global rush for new mineral supplies, began to fall behind. What had once been a roughly two-month turnaround for assay results stretched to five months or more. For a data-driven exploration company, that delay was crippling. Teslyuk describes being “7 km behind” in samples, meaning thousands of meters of drilled core with no chemical analysis to feed back into the company’s models.

Without timely assays, Earth AI’s feedback loop breaks down. The company’s system is designed to learn from each drill hole, updating its understanding of the subsurface and refining the next set of targets. When results arrive months late, every subsequent drilling decision is made with stale or incomplete information, increasing costs and the risk of drilling in the wrong place.

To regain control, Earth AI is building its own in-house laboratories, a move that effectively vertically integrates the entire exploration workflow. Instead of shipping cores to distant facilities and waiting months, the company aims to cut turnaround times to about five days. That speed would allow geologists and data scientists to adjust drill plans in near real time, sending rigs toward the most promising structures and away from barren rock.

Earth AI will still rely on independent labs for final confirmation of discoveries, especially when economic decisions or potential mine sales are on the line. But for day-to-day exploration, internal labs could transform the economics of the search. Faster answers mean fewer holes, better-targeted drilling, and richer datasets to train the company’s models, tightening the loop between prediction, testing, and refinement.

In a sector where delays can quietly erode millions of dollars, Earth AI’s bet is that owning the lab step is not just an operational tweak but a strategic advantage in the race to secure critical minerals.