Exclusive: Earth AI is vertically integrating the search for critical minerals | TechCrunch

A model is only as good as its data, and for Roman Teslyuk, the data wasn’t coming fast enough.

“I hate delays,” Teslyuk, founder and CEO of Earth AI, told TechCrunch.

For the last few years, Earth AI has been searching for critical minerals like copper, platinum, and palladium in parts of Australia where no one thought there would be any. The startup’s AI models suggested a few spots that have proven themselves promising, but locating rock with the highest concentration of minerals has been slower than Teslyuk would like.

The problem, he said, was the labs.

“Since we ramped up the drilling capacity, we started getting these massive delays,” he said. Typically, labs that process rock samples for evidence of critical minerals have backlogs of around two months, Teslyuk said. But lately, as interest in developing new sources has jumped, the delays have more than doubled. “We’re 7 km behind — 7,000 meters of samples we don’t have data about.”

So Earth AI is setting up its own labs instead, the startup exclusively told TechCrunch, hoping to bring the time down from five months to five days.

Earth AI’s models have been good at highlighting areas with the potential to develop into a mine, Teslyuk said, but once those have been identified, the startup still needs to drill to confirm what minerals lie below and how they’re distributed. Subsurface exploration has come a long way, but there’s still no replacement for drilling.

Once the drill cores have been extracted, they need to be processed by a lab. “We don’t know whether we hit gold or not. We can’t see it with our eyes,” he said.

For final decisions on the economic value of a mine, including those that might influence a sale, Earth AI will still use third parties to validate its discovery. But during the exploration process, a speedy in-house lab has the potential to significantly reduce costs by ensuring the drill is sent to the right spots to obtain the best data for the model.

“If you don’t have the answers in time, you have to wait for five months for the answer, the next question [of where to drill] is not as good as it could be,” Teslyuk said. “To minimize drilling, you want to effectively ask the right questions, to get the information in time so you can narrow down exactly where to go.”