THE PROBLEM 

Satellites are now imaging every point on Earth's surface with a frequency and resolution that would have been classified capability a decade ago. Hundreds of commercial satellites are generating petabytes of imagery daily—data that contains signals about agricultural yield, commodity flows, military movements, infrastructure stress, supply chain activity, climate change progression, and economic output across every geography on the planet. The hardware problem is largely solved. The intelligence problem has barely been started. 

THE OPPORTUNITY 

The real opportunity is not in launching more satellites—it's in building the AI layer that turns raw imagery into actionable intelligence. We're looking for startups building vertical-specific Earth observation intelligence platforms: tools that ingest multi-source satellite data and deliver decision-ready insights to commodity traders, sovereign wealth funds, insurance underwriters, agricultural operators, and government analysts. The analogy to financial data terminals is exact—the startups that build the Bloomberg of physical-world intelligence will command the pricing power and switching costs that go with it. For the savvy investor, this is data infrastructure for a world that is increasingly unreadable from the ground. 

Analysis & Implications 

In 2019, a hedge fund generated consistent alpha on retail earnings reports by counting cars in Walmart and Target parking lots across hundreds of stores, observed via satellite, in the days before each quarterly announcement. The fund didn't need better financial models. It needed better satellite data and an analyst who understood what they were looking at. That edge is now table stakes in quantitative finance. The parking lot trade has been commoditized. But the depth of satellite intelligence available to non-defense sectors has barely begun to be exploited—because the intelligence platforms that turn raw imagery into actionable analysis have barely been built. 

Consider the gap between raw satellite imagery and useful agricultural intelligence. A satellite passes over a field. It captures multi-spectral imagery. Somewhere in that imagery is a signal about crop health, soil moisture, canopy density, and estimated yield. Converting that signal into a usable yield estimate requires calibration against local weather data, soil type, historical yield patterns, and plant phenology models. The satellite data provider doesn't do that work. The agricultural company receiving the data doesn't have the models. The intelligence never gets extracted from the imagery it's sitting in. 

Ukraine changed how the world thinks about commercial satellite intelligence. Planet Labs, Maxar, and Airbus imagery was used by open-source investigators and governments to document troop movements, equipment staging, and infrastructure damage in near real time—intelligence that previously would have required state-level reconnaissance capability. The same analytical capacity, applied to commercial questions—commodity flows, shipping activity, construction progress, factory output—is available to anyone who builds the right platform. 

The opportunity is vertical-specific intelligence pipelines: take raw imagery from multiple satellite providers, apply the analytical models appropriate to a specific domain, and deliver a decision-ready output. In agricultural commodities, that's yield estimates by crop and geography, two to three months before harvest. In insurance, it's property condition assessment before and after weather events, automated against claims. In sovereign wealth, it's industrial output signals from satellite-observable economic activity across geographies where ground truth is unavailable. 

Orbital Insight and Descartes Labs pioneered this concept, but both ran into the same problem: they tried to build horizontal analytics platforms serving too many verticals and ended up thin everywhere. The winners will go narrow—one vertical, one buyer type, one decision—and build the intelligence pipeline for that specific use case until it's genuinely better than any alternative. 

The business model is information as a service: subscription, priced on the value of the decision it improves, not on the cost of the data. The margins are enormous when you're selling information that moves the needle on a billion-dollar position. Start with a buyer who makes a decision on a regular cadence that would be materially better with better physical-world intelligence. Build for that decision specifically. Expand from there.