SpaceKnow.

Here is a genuinely strange idea that turns out to be a good business. Suppose you want to know how China's factories are doing. You could wait for the government to publish a number, and then wonder how much you trust the number. Or you could point a fleet of satellites at roughly 6,000 industrial sites, photograph them, and have a machine count the buildings, the roads, the construction, the cars, the changes. Do that three times a week and you have an economic indicator that owes nothing to anyone's press office. This is more or less what SpaceKnow does, and it is the kind of thing that sounds like science fiction until you notice it is running on Bloomberg Terminals.

SpaceKnow was founded in 2013 by Pavel Machalek, who had previously analyzed data from NASA's Kepler space telescope, and Jerry Javornicky, who had previously built a popular email client called eM Client. That is a slightly odd founding pair - one person who knows how to stare at faint signals from space, one person who knows how to turn messy inputs into a product people actually use. In hindsight the combination is the whole company. The hard part of satellite analytics is not getting the pictures. Anyone can buy pictures. The hard part is making the pictures mean something to a person who will never look at a pixel.

The company's stated mission is to "index the physical world," which is the sort of phrase that could mean nothing, except that here it means something quite specific. Google indexed the web. SpaceKnow wants to do the equivalent for the parts of the economy you can see from above: ships in ports, aircraft on tarmacs, cars in lots, roofs going up on factories, lights coming on at night. The world is constantly broadcasting its own activity in the form of visible change. Someone just has to be watching all of it, all the time, and be able to tell you what changed.

Clouds cover the Earth about two-thirds of the time. Most competitors wait for a clear day. SpaceKnow uses radar and refuses to.

The flagship product is called SpaceKnow Guardian, and its central trick is integration. There is no single satellite that sees everything; there is an ecosystem of them. Guardian ingests feeds from many providers - optical imagery from companies like Planet Labs and Satellogic, synthetic-aperture radar from ICEYE, data programs with the European Space Agency - and runs them through proprietary machine-learning algorithms that detect objects across hundreds of categories and flag change automatically. The radar part matters more than it sounds. Optical satellites are blind through clouds and darkness, which is a problem given that clouds cover most of the Earth most of the time. Radar sees straight through weather. So while a purely optical competitor is waiting for the sky to clear over a port, SpaceKnow can keep counting.

What comes out the other end is deliberately boring, in the best way. The customer does not get a data-science project. They get an answer, a dashboard, an API response, an alert, an index value. That is the quiet product decision underneath everything: SpaceKnow does not ask its customers to become remote-sensing experts. It absorbs the complexity and hands back something legible. In 2024 Fortune named the company one of its Most Innovative Companies, which for a firm like this is less about any single flashy demo and more about that act of translation - taking intelligence-community-grade capability and making it usable by a hedge-fund analyst or a construction manager.

And those are, in fact, the customers - which is the second strange thing about the business. The same core engine serves defense and intelligence agencies that need situational awareness, financial firms that want to nowcast the economy before the reports come out, construction and real-estate outfits tracking site progress, and environmental teams monitoring wildfires, floods, deforestation, and carbon. On paper these have nothing in common. In practice they all need the identical primitive: know what is physically happening somewhere, faster and more independently than the other guy. SpaceKnow built that primitive once and sells the answer many times, which is a nice position to be in.

The economic indices are where the whole thesis becomes concrete. The China Satellite Manufacturing Index is published multiple times a week and has been distributed on Bloomberg Terminals, where it functions as an independent read on Chinese industrial activity. The company has done similar work elsewhere - using night-light intensity, for instance, to produce an independent view of African economies where conventional statistics are thin. The pitch to a trader is simple and a little subversive: here is a number about the economy that no ministry produced, assembled from photographs of the economy actually happening. As the co-founder once put it, economists should look to the skies.

None of this requires a large company, which is perhaps the most quietly impressive part. SpaceKnow runs lean - on the order of 38 people, split between New York and Prague - while measuring the industrial output of the world's second-largest economy on a rolling basis. That is the leverage of building the right instrument. The funding history is modest by space-industry standards: early seed rounds followed by a $4 million Series A in 2017 led by Reflex Capital and BlueYard, the same year co-founder Javornicky landed on the Forbes 30 Under 30 Europe list. This is not a company that raised a fortune and pointed it at hardware. It is a software company that decided the satellites already in orbit were enough, if you were clever about reading them.

The interesting tension in a business like this is that a general-purpose sensing platform is a kind of moral instrument as much as a commercial one. The tools that price a trade are the tools that watch a forest burn or a flood spread. What you choose to point them at is a decision. SpaceKnow's answer, so far, has been to point them at nearly everything and let the customer decide - defense here, finance there, the climate over there - which is either admirably neutral or slightly vertiginous, depending on your mood. Either way, the machine keeps looking down, three times a week, and the physical world keeps answering.