We Built Our Own Lab

When I started Earth AI nine years ago, I had one conviction: the problem with mineral exploration isn't that we need to go more remote. It's the technology we use. Geologists have been bashing rocks on mountains for decades, waiting for something to show up. That's not an efficient search technology. I wanted to change that.

So we built an AI platform trained on 400 million geological data points, representing 50 years of exploration history across an entire continent. We taught it to look for hydrothermal systems, the geological conditions that create mineral deposits, rather than scanning for the metals directly. Think of it like this: a mineral deposit inside a hydrothermal system is like a seed in an apple. When the seeds get deeper and smaller, they become really hard to find. But apples? Apples are much bigger, much more recognizable.

It worked. Our discovery success rate hit 75%. The industry average is about half a percent. We were finding deposits that didn't fit the textbooks, things academics insisted couldn't exist in certain places. We drilled anyway and proved them wrong.

But then we hit a wall. Not with our predictions. With everything that comes after.

The Bottleneck We Didn't See Coming

Traditional exploration is brutal logistically. You need six types of contractors organized. You're negotiating with landowners, managing equipment schedules, dealing with people who don't show up when they're supposed to. Early on, we learned that even when we gave exploration companies twenty promising sites, they didn't have the muscle to go test them. They wanted us to narrow it to one or two. But we couldn't know which of those twenty was better without drilling.

And nobody was drilling.

So we brought drilling in-house. We found an abandoned drill rig on a coal mining site, called them up, and asked if we could have it. Thirty thousand dollars. We thought building our own drilling capability would be straightforward.

It was not.

It took years. We prototyped, broke things, went back to the drawing board. You can't model geology on a computer; you have to go out and test it. Eventually we developed a mobile, low-disturbance drilling platform that changed how we operate. Today we run multiple rigs around the clock, 24 hours a day, seven days a week. We can mobilize to any site without ground preparation. We contain our own waste management. We have our own drillers. When we finish a hole, we move to the next one immediately. The farmers are happy because all that's left is a tiny hole in the ground that we fill up. Within a week, there's no sign we were ever there.

We've made four discoveries. We're drilling out one of them into a defined resource right now.

The system works. But success created the next constraint.

Seven Kilometers Behind

We scaled drilling faster than the rest of the industry could absorb. We're now more than seven kilometers (7,000 meters of drilled core) behind on assay results. The commercial labs that analyze mineral samples have stretched their turnaround to four or five months.

The reason is structural. Labs have been through enough boom-bust cycles that they refuse to scale capacity even when demand spikes. They know the bust will come eventually. They don't want to be caught with excess infrastructure when it does. Critical minerals demand is currently in boom mode. The labs aren't keeping up.

So our rigs were waiting. Our geological models were going stale. And the thing that makes us different, our iteration speed, was grinding to a halt.

Drill Holes Are Questions

Every drill hole is a question you ask the earth.

I think about it this way: every drill hole is a question you ask the earth. The quality of your next question depends on how quickly you get the answer to your last one.

What We Built

Our in-house assay lab is purpose-built for speed.

Commercial labs handle any sample type, any analysis, any customer. That flexibility comes with overhead. Our lab does one thing, and it does it fast. A single all-encompassing spectrometer that flags anomalous readings. When we strike something interesting in the field, we can confirm results in days and redirect our rigs immediately.

Target turnaround: under a week. For urgent samples, when we hit something that looks genuinely promising, one person can take a single drill hole's sample directly to the lab, crush it, pulverize it, and analyze it in a single day.

We built the facility in about two months. It's already processed more than 31 kilometers of drill core.

The Lab Makes the Rigs Smarter

The lab doesn't just unblock downstream processing. It changes what happens upstream.

With in-house assay turnaround, every drill hole feeds back into our geological model within days. That model gets better. A better model targets the next drill hole more precisely. Fewer holes per discovery. Faster progression from resource-grade resolution to reserve-grade resolution. Fewer rigs sitting on suboptimal targets.

The lab makes the rigs smarter. Smarter rigs justify more rigs. More rigs produce more samples. The lab scales to absorb them.

Each loop tightens the next. That's what compresses the discovery timeline.

The Full Stack

When I look at what we've built over nine years, every piece exists because we hit a bottleneck and decided to solve it ourselves rather than work around it.

Our AI finds targets faster than traditional geology. Our in-house teams handle mineral rights and permitting so we're not waiting on external timelines. Our proprietary drilling fleet operates day and night with minimal environmental impact, no contractor dependencies, and the ability to mobilize at short notice.

And now, our in-house lab ensures that data flows continuously from drill site to analysis to decision.

We're the only exploration company in the world that controls every step from AI-powered target identification through drilling and sample analysis. No handoffs. No waiting. No dependencies.

Why This Matters

I've been watching the critical minerals conversation intensify over the past few years. Governments are waking up to the supply chain problem. There's talk of strategic reserves, reshoring refinery capacity, securing offtake agreements.

But here's the thing: you can build all the refineries you want, but refineries need feedstock. You can't refine deposits that haven't been discovered yet.

The industry is finding 40 new deposits per year globally. Down from 150 a decade ago. Hundreds of mines are projected to close in the next five years. The demand trajectory is only accelerating. We need to find 100 new deposits in the next decade just to keep pace.

Traditional exploration timelines won't get us there. Decades from prediction to production isn't fast enough.

What's Next

I'll be honest: the lab will probably never be "correctly sized." By the time we reach capacity, we'll have outgrown it. The plan is to keep building. The next facility, and the next one after that. You prototype, you test in the real world, you get feedback, you iterate. Software or hardware, it's the same process.

Nine years ago, people told me mining would never be a trend. AI wasn't proven. Hardware startups were too hard. Nobody believed critical minerals would matter.

Now we're sitting at the intersection of three things that have all become undeniably important: AI that actually works, industrial technology that investors take seriously, and a critical minerals supply chain that governments treat as a national security priority.

We're not waiting for the industry to catch up. We're building the infrastructure to find the mines the world needs, faster than anyone thought possible.

The ore bodies are out there. They've been stuck in those rocks for 400 million years. Now we have the tools to find them.