The Business of Geological Intelligence: How Satellites Replaced Boots on the Ground

A junior exploration company in Western Australia spent $4.2 million last year flying helicopters over a copper prospect. Three field seasons. Forty-seven drill holes. They found nothing commercial.

Meanwhile, a competitor working the same belt ran spectral analysis on Sentinel-2 imagery, narrowed the target zone to about 380 hectares, and hit mineralization on their fourth hole.

That's the shift. That's the whole story, really.

I've been watching mineral exploration budgets get rewritten over the past three years, and the math is starting to make traditional prospecting look almost irresponsible. Not because boots on the ground don't matter — they absolutely still do — but because of what you're asking those boots to do. Walking unmapped terrain hoping to spot an outcrop is a lottery ticket. Walking a 200-meter anomaly your satellite stack already flagged? That's actual work.

Why this shift took so long

Honestly, I thought satellite-based exploration would've eaten this market by 2018. I got that timing wrong.

The technology was ready. Hyperion launched in 2000. ASTER had been mapping mineral signatures for years. The bottleneck wasn't sensors — it was the workflow gap between a geologist who knows what chlorite alteration looks like in a hand sample and a data scientist who can pull spectral indices out of raster bands. Those two people rarely sat in the same room. When they did, they spoke different languages.

What changed isn't the satellites. It's the software layer sitting on top of them.

Platforms like GeoMine AI are doing something specific that the old remote sensing software never quite cracked: they translate spectral outputs into the vocabulary geologists actually use. Argillic alteration. Iron oxide ratios. Propylitic halos. Not just "NDVI band 4 minus band 3." That sounds small but it's the entire reason adoption stalled for two decades. Geologists wouldn't trust tools that didn't speak geology.

And cost matters. A full-spectrum analysis over a 10,000 km² license area used to mean $200K minimum once you priced in airborne hyperspectral surveys. The same area, processed through modern satellite stacks with cloud compute, runs closer to $15,000. Sometimes less.

What miners are actually buying

Here's the thing most outsiders miss about mineral intelligence platforms — the customer isn't really paying for data. Sentinel and Landsat data is free. ASTER archives are free. The customer is paying to not waste a field season.

A single drill program in a remote part of the DRC or Mongolia can burn through $8 million before you've logged a meter of core. Camp logistics. Helicopter time. Geologists at $1,800 a day. Permitting delays. If a software subscription costing $40K a year cuts your target generation phase from 18 months to 5, that's not a SaaS purchase. That's an insurance policy on your exploration budget.

The major mining houses figured this out first. Rio Tinto's been running internal spectral teams for years. BHP too. What's new is that juniors — companies with market caps under $50 million — can now access roughly the same analytical horsepower without building a remote sensing department from scratch.

That's the democratization story. It's also why I think this category gets a lot bigger before it consolidates.

The data nobody talks about

Look, there's an unglamorous truth underneath all this. Most mineral exploration still fails. Even with perfect satellite intelligence, the discovery rate for economic deposits hovers somewhere around 1 in 1,000 prospects evaluated. Spectral analysis doesn't change geology. It just changes which 1,000 prospects you bother evaluating.

What it does brilliantly is kill bad targets fast. And killing bad targets is, financially, almost as valuable as finding good ones. Every dry hole you don't drill is capital you can redirect.

I talked to a geologist in Perth last month who put it this way: "Twenty years ago I'd spend a week in the field to rule out a prospect. Now I rule it out before lunch." He wasn't bragging. He sounded a little sad about it, actually. There's a craft being lost. But the economics aren't sentimental.

Where the business model is heading

A few things I'm watching:

The pure data plays will struggle. Selling raw imagery or basic indices is becoming a commodity, and Planet Labs and Maxar already own that shelf. The defensible business is the interpretation layer — taking spectral outputs and turning them into ranked drill targets with confidence scores. That's where the moat lives.

Governments are about to become huge customers. Critical minerals policy in the US, EU, Japan, Australia, and India is pushing public money into domestic exploration. National geological surveys are buying satellite intelligence subscriptions the way they used to commission airborne surveys. That's a procurement cycle that hasn't fully landed yet.

And the agricultural overlap is real. The same spectral techniques that flag iron oxide alteration also map soil salinity, crop stress, and water content. Companies built for mining are quietly finding that ag insurers and commodity traders want the same pipes pointed at different problems.

The broader point is that geological intelligence stopped being a niche service for senior miners and started becoming infrastructure. Like accounting software. Like CRM. Boring and essential.

I keep coming back to that Western Australia example. $4.2 million versus a software subscription and four drill holes. If you're an investor reading exploration prospectuses in 2025 and the company isn't running spectral targeting before they spud a hole, you should probably ask why.

What are they hiding from — the data, or themselves?