Will Marshall - Indexing the Earth - [Invest Like the Best, EP. 251]
What this covers
The conversation is between the host and Will Marshall, who co-founded Planet in 2010 with a group of NASA scientists. The company has since launched about 450 satellites. It scans all of Earth's land every day and keeps every frame, which gives roughly 1,700 images per point and a time axis that works like Google Maps with history. Marshall explains the mechanics: pixels about three meters wide, spectral bands such as near-infrared and red edge that track biomass and crop type, and the SuperDove generation, which yields about five times more imagery per satellite at similar cost. A smaller fleet of roughly 50-centimeter satellites is tasked to specific sites. He frames the business under Indexing the Earth, with machine learning turning pictures into counts of ships, roads, or buildings.
The range of applications is wide. In agriculture he claims field-level imagery can raise yields 20 to 40 percent while cutting inputs, a claim flagged as contested. Other uses include flood and wildfire response, floodplain modeling for Bangladesh and India, a first map of the world's coral reefs, and a Pulitzer-winning journalist's discovery of about 200 detention camps in western China. The policy discussion covers the legal limit of sovereignty at about 100 kilometers, the Cold War norm of mutual observation, and space debris, which Planet avoids by flying low enough to deorbit within about three years. On privacy, Marshall says orbital imagery cannot identify a person.
The sharper arguments come near the end. Marshall argues that Data as the New Oil is imperfect because data can be sold repeatedly. He says land-use change, not climate change, is the main driver of ecosystem loss. He also argues that Natural Capital Externalities must enter economic accounting.
Marshall argues that falling satellite and launch costs have made daily, high-resolution Earth imaging possible, and that the resulting time-series data, not the hardware, is the durable asset that will underpin agriculture, government, finance, and the measurement of natural capital.
- Satellite cost-performance has improved about 1000X over five to ten years while rocket costs fell about 4X, so the data collected per dollar has changed by orders of magnitude.
- Data can be sold repeatedly at near-zero marginal cost, while algorithms are easily copied and the hardware is hard to replicate, so the data archive is the moat.
- Real-time satellite measurement enables enforcement and integration of environmental externalities into economic decisions.
This asset isn't compiled yet
You're seeing its claims, ranked. Compile it to build the argument threads, weight them, and check each claim against your library — the full view.
Planet captures an image of the entire Earth's land every day and retains all of them, yielding about 1,700 images per land point, which makes a deep, recent time stack that can answer what has changed, like Google maps but with a time axis.
“So it's like Google, but with a time access.”
By using consumer electronics in space instead of traditional government-grade components, Planet aimed to cut satellite costs from billions of dollars per satellite to roughly $100,000 to $1 million, a reduction of about 1,000 to 10,000 times.
“can we even leverage consumer electronics? Like the kind of electronics that's in your laptop or in your phone in space”
A journalist who won a Pulitzer Prize used Planet's imagery to identify about 200 previously unknown detention camps in western China, illustrating how the data supports investigative journalism.
“Pulitzer prize this year was won by some journalist who used our data to discover about 200 potential Uyghur”
International law established early in the space age that national sovereignty extends up to about 100 kilometers altitude and that space beyond is not any state's territory, so satellites may photograph any country without permission.
“up to 100 kilometers altitude, it's your territory, and beyond that it's space.”
The data-is-the-new-oil analogy is imperfect because oil is consumed when used while data can be sold to many customers repeatedly, giving data an inherently better business profile, though data still needs refining and may become commoditized.
“oil can only be used once, any bit of oil, whereas data, you can pass off to lots of users.”
Because a single image can be sold to multiple customers at very low incremental cost, Planet's direct margins are very large, and Marshall argues investors value Planet as a data business rather than a hardware business.
“we're selling data, and we can sell each image multiple times. So the profit margins are really large.”
Rocket costs have fallen roughly 4X, mainly through reusability, and satellite cost-performance has improved about 1000X over five to ten years, so space is undergoing a renaissance driven by new, large Earth-observation data sets rather than by rockets and billionaires.
“We are seeing massive shifts, rockets of decreased cost about four X. That's a big deal. Mainly through reusability”
Planet's main scanning satellites capture eight spectral bands, including near-infrared and red edge, which allow tracking of biomass, crop type, and yield, and a haze band that helps calibrate the other data.
“does that in eight spectra bands. So three that are familiar the eye, the red, green, and blue, but we also have five other spectra bands.”
Early space activities by the Russians and Americans, including exploding rocket bolts and leftover fuel that later ignited, created long-lived debris, and deliberate anti-satellite destruction by the US, China, and India added to the problem and signaled military capability.
“Instead of being one bit, there's now thousands and thousands, and thousands bit.”
Planet's daily imagery supports disaster response and prevention, including mapping floods and their damage, identifying wildfire edges and wind, mapping California's trees to target fuel clearing, and modeling floodplains so governments can avoid building in hazard zones.
“we can see the flood the day before, the day after the flood, and see which bridge is down to help the relief effort”
Developing countries often build in floodplains, so modeling how floodplains are shifting lets governments and financiers identify which assets are at risk and advise where to build, a point Marshall argues matters for insurance and finance as well.
“Especially in developing countries, are often build in the wrong place, so we have a big project with Google”
Daily imagery lets governments see floods the day before and after to find damaged bridges and direct relief, and also to model future floodplains so developing countries like Bangladesh and India avoid building in the wrong places.
“we can see the flood the day before, the day after the flood, and see which bridge is down to help the relief effort.”
Planet keeps its satellites at about 400 kilometers so they naturally deorbit within roughly three years and burn up, avoiding adding to the roughly 100 million debris fragments in orbit, most of which are fragments of old satellites rather than natural asteroids.
“we keep our satellites way down low to make sure they're out of the way of debris, not creating any more debris.”
Machine learning automatically extracts roads, buildings, ships, and planes from imagery, so a user can ask for counts over time, such as ships in the top 10 ports, without looking at any pictures, reducing the need for geospatial expertise.
“you can now do that on our platform. ... we're trying to get better and better so that one can actually not have to have a PhD in geospatial science”
The Cold War norm that any nation could launch satellites to observe any other was adopted deliberately because transparency was judged better, and Marshall argues it reduced tensions, so broad access to Planet's data supports peace and security.
“it reduced tensions in the Cold War. We're not, by the way, ever exclusively giving that to one government or another.”
Human land-use change, mainly deforestation for cattle and ocean trawling, rather than climate change, is the main driver of ecosystem destruction Planet observes, and Marshall calls the rate of loss staggering.
“It's not actually climate change. It's mainly these things like deforestation and illegal fishing.”
The global economy can no longer treat natural capital such as trees, air, and rivers as free, so environmental externalities must be integrated into economic accounting, which creates demand for measurement that Planet's data can supply.
“We cannot any longer presume that natural capital is free.”
Planet operates a wide-area scanning fleet that detects changes across the planet, paired with a smaller high-resolution fleet of about 50 centimeter resolution satellites tasked to specific locations, which can image a site up to 10 to 12 times per day when requested.
“the scan that finds changes around the planet... then you can use the high res system to point at those changes”
Agricultural satellite monitoring has existed for about 40 years, but its low cadence of annual or periodic passes produces end-of-year reports that are useless to farmers during the growing season, which requires intelligence a few times a week.
“It's more for the annual survey of, "How did we do?" Which is useful at the end of the year, but it's not helpful for the grower”
Planet produced the first map of the world's coral reefs, classifying types and detecting early bleaching and illegal fishing, and six countries have already established marine protected areas based on it.
“In fact, there are already six countries have established marine-protected areas around coral reefs that we mapped for them.”
At roughly 400 kilometers altitude and with pixels about three meters wide, satellites cannot identify a person, and seeing through buildings is impossible with current technology, so personal privacy concerns from orbital imaging are far off and more relevant to drones.
“really can't see or identify a person from space, and I don't think anyone can, even with the huge satellites”
Daily field-level imagery enables precision agriculture that can raise crop yields by 20 to 40 percent while reducing fertilizer and other inputs by similar amounts, which matters in a trillion-dollar industry.
“That can improve their crop yield by 20 to 40%, at the same time decrease their use of things like fertilizer and other resources by similar amounts.”
Each pixel in Planet's full-Earth scan covers about three by three meters, and each frame is a 47-megapixel image roughly 20 to 25 kilometers across, with about three million frames per day covering over 300 million square kilometers, more than twice Earth's land area.
“Each pixel from our scan is about three by three meters. We collect them in 47 megapixel images.”
The SuperDove generation added four spectral bands for eight total and produces about five times more imagery per satellite per day than the previous generation at roughly the same cost, in response to customer requests.
“Those satellites produce five times more than the previous generation of satellites per satellite per day for almost exactly the same cost.”
Planet's accumulated archive of daily Earth imagery cannot be reconstructed retroactively, so a competitor entering later lacks years of historical time series, making the archive a durable moat.
“That data archive, it's impossible to go back and get.”
Planet has completed 18 design-build iterations of its satellites, treats satellites like servers that get replaced when outdated, and upgrades sensors and processors as they improve, so satellite capability per day has increased about 10,000 times from the first to the latest model.
“We call it strapping space to Moore's law.”
Planet's strategy is to move up the value stack from selling raw imagery to selling analytics-derived information products, since hedge funds, commodity traders, and other buyers want calibrated time-series answers rather than pictures, which lets the company charge more per bit of refined data.
“selling information products derived from that imagery, what we call "going up the stack."”
Satellites have brought awareness of environmental degradation for about 50 years, and climate scientists warned, but action lagged; Marshall argues real-time measurement and enforcement, rather than periodic awareness, is what enables action.
“This is the data set that enables us to take action.”
Algorithms alone have zero value while data has substantial value, so the power lies with whoever holds the best data, which is why Planet emphasizes proprietary data over the open algorithms Google has released.
“Algorithms alone have zero value. Data alone has a lot of value.”
Marshall cites a roughly 70 percent decline in life on Earth over 40 years, including 82 percent of wild mammals, 75 percent of insects, 70 percent of freshwater fish, over half of coral systems, and more than half of forests, with forested land about double 40 years ago.
“82% of wild mammals have gone, 75% of insects, 70% of the fish in freshwater rivers and lakes, over half the coral systems”
In the long term, a live image of the Earth may be possible, and the ultimate product is a search box where users ask questions like how many houses are in a country or where trees were deforested, receiving answers without viewing imagery.
“Just imagine a search query box on it and you can just say, "Hey, how many houses are there in Pakistan?"”
Because Planet sees crop health and copper mine output across the world before others, financial buyers such as hedge funds could derive commodity-price signals from its data ahead of the market.
“We know how well crops are doing before anyone else, the whole world saw your yield. We know the output from all the world's copper mines before anyone else.”
Rocket costs fell about 4X and satellite cost-performance improved about 1000X, so the combined cost reduction is about 4000X, with satellite miniaturization (from bus-sized craft to loaf-of-bread-sized ones with similar capability) driving most of the gain.
“So we have a 4,000 X in cost reduction overall, if you like, but the thousand X is what dominates.”
Planet first checks an embargo list, then an internal ethics committee decides whether a permitted buyer should actually receive data, and it has often refused requests where it was uneasy about the end use.
“We have an ethics committee that even after that, goes, "Okay, we're allowed to sell to these guys, but should we?"”
Planet sells imagery per unit area, such as square kilometers or hectares, with volume discounts for larger purchases, and Marshall says it has real pricing power because no competitor offers comparable data and building a comparable fleet and archive would take years.
“We basically sell imagery on a per-unit area basis: kilometers, hectares, mainly just area.”
Planet designs its own boards, optics, cameras, and radios while buying commodity chips, and its custom radio achieves 1.6 gigabits per second over a thousand kilometers to 48 ground stations, a vertical integration analogous to Apple's approach.
“It's all our own bespoke boards. It's all our own bespoke designs, of optics, of camera systems, of radio systems.”
Planet began in 2010 when about seven ex-NASA scientists started building satellites in a garage, and building its own hardware creates efficiencies because Planet sells data rather than satellite hardware, as Apple and Tesla do.
“back to the garage days, myself and a small team about seven of us left NASA to start Planet. And we literally started building the satellites in our garage.”
Governments and intelligence agencies task satellites only at known locations, while Planet's full scan finds unknown sites, such as a missile base in eastern Iran, so the wide-area scan delivers intelligence that targeted satellites miss.
“We find new things that they didn't know to look for, a new missile base in Eastern Iran we just recently found”
Because Planet holds a huge, continuously refreshed archive of labeled-like imagery across the land mass, it provides the training data that machine learning practitioners consider most valuable, creating a data advantage that is difficult to replicate.
“You ask any machine learning expert, what's the most exciting thing? It's training data, well, we have gobs of it.”
Indexing the Earth requires running computer vision on each of roughly three million daily images to extract roads, buildings, ships, planes, trains, and trees, building a database of every object over time; Marshall calls this intellectually straightforward but enormous in engineering effort.
“I actually think it's intellectually relatively straightforward. Of course, there's such a huge amount of work behind that.”
Google's mapping services automatically task Planet satellites when they detect outdated map areas, so new roads, train stations, and buildings are extracted and updated, meaning everyday navigation depends on commercial satellite data.
“whenever they find any indication that map is going out date, they automatically task our stuff”
Planet intends to build an SDK and tools so thousands of developers can build apps on its data, charging for usage in small amounts with free trials, rather than giving the data away, similar to how Twilio or Stripe monetize APIs.
“You can imagine an SDK and thousands of apps being built on top of this data.”
Compounding cheaper launch and cheaper satellites means the world is generating and transmitting about 10 times more data about the Earth than before, and that data is relevant to the transition to a sustainable economy.
“We are generating and transporting vastly more data 10 times more than we were before or more about the earth”
Satellite data can count the carbon stored in trees and measure other natural capital, making Planet's dataset foundational to a multi-trillion-dollar transition to a sustainable economy.
“We can count the carbon stock in all those trees. Our data is foundational to underpinning the transition to a sustainable economy, which is a multi-trillion dollar transition”
Planet was founded by about seven former NASA staff who began building satellites in a garage, with the mission of imaging the whole Earth every day, a goal previously thought to require about 100 satellites that no one had launched.
“back to the garage days, myself and a small team about seven of us left NASA to start Planet. And we literally started building the satellites in our garage.”
Debris cannot maneuver because it lacks propulsion, so satellites can dodge one another but removing existing debris requires a different approach, such as the ground-based laser scheme Marshall helped conceive at NASA.
“What do we do about the debris, which don't have propulsion systems on them?”
Satellites fall into roughly five categories: imaging, communications, navigation, early warning, and signals intelligence, with navigation, early warning, and signals intelligence largely government-run, while commercial activity is concentrated in communications and Earth observation.
“There's roughly five types today. There's imaging satellites. There's communication satellites.”
Satellites in orbit roughly doubled in the last five years, but about half of them belong to two companies, Planet and SpaceX, so proliferation is highly concentrated.
“half of the satellites are now of two companies, that's Planet and SpaceX. So although there's a proliferation, it's really quite concentrated”
Planet chose its first markets by land use: about a quarter of Earth's land is agricultural, another quarter is forest, and about 10 percent is urban, so agriculture was the obvious first target due to its sheer area.
“About a quarter of the land mass of the earth, 25%, is agricultural land, so we went into that.”
After about six years, Planet launched just over 100 satellites, creating what Marshall described as the largest constellation of satellites in history, though SpaceX has since passed it in count.
“we achieved that mission of having launched the largest constellation of satellites in history. We had just over 100 satellites in orbit.”
Planet maps all deforestation in 64 tropical countries, with the Government of Norway funding the work, and provides the data to forestry ministries to enable enforcement against illegal clearing.
“All the deforestation in the 64 tropical countries, with the Government of Norway paying for the data”
Space-derived Earth data is relevant to every effort to transition to a sustainable economy, including national climate missions, corporate ESG measurement, and digital transformation in agriculture, transportation, and government, so the data is more important than the spectacle of spaceflight.
“It's about data about the earth that helps us to transition to a sustainable economy.”
Marshall argues the biggest business opportunities come from products people do not yet know they want, citing the iPhone and daily Earth imagery, while Planet also responds to customer requests such as fusing datasets to reduce the customer's workload.
“often the biggest opportunities from a business perspective are doing things that people don't know they want yet. No one asked for an iPad or an iPhone.”
Planet's business works as three interlocking legs: the wide scan that detects change, the high-resolution zoom that inspects it, and the machine learning analytics that interpret what is found.
“Really, I think of three legs, the story of planet, which is the scan, the zoom in, and the analytics”
Planet is going public through a merger with a special purpose vehicle to raise capital for sales and marketing and for building higher-level analytics products, and also to gain visibility, since many potential customers still do not know the company.
“the investors of dMY the group that we are merging with as part of our going public”
Asked whether humans are effectively an invasive species, Marshall answered yes, reflecting the view that human land-use change is driving ecosystem loss.
“So we're basically an invasive species, is one way to think about it?”
Marshall argues investors favor Planet because it combines space, sustainability, and a data business with no competitor, which matches what the market currently wants, and he says the company has over $100 million in revenue last year and is ready for public markets.
“We've got a mature business with over a hundred million in revenue last year. We're ready.”
Investors are attracted to Planet because its data is unique and cannot be obtained elsewhere, it serves many vertical markets beyond agriculture, it is a high-margin data business, and it aligns with the market's emphasis on sustainability.
“investors really like it for a number of reasons. One is that our data is completely unique. No one else can get it.”
A live, continuously updated image of the Earth is possible in the long term but not achievable in the near term, according to Will.
“it's possible in the long term that we would more or less be able to have a live image of the earth. ... It's not where we are now, and it won't be for a long time”
Reaching daily imaging of the whole Earth was much harder than expected: Will thought it would take about three years but it took about six, though Planet achieved the largest satellite constellation in history with just over 100 satellites in orbit.
“I thought we could do it in three years or, and it took six, but we achieved something that was really cool.”
Corteva uses Planet imagery to monitor about a million farmers' fields every day, showing that agricultural customers operate at large scale.
“We do that, for example, with Corteva. They use this to image about a million farmers' fields every day.”
The kindest thing anyone has done for Marshall was mentorship, including a PhD supervisor who won a Nobel Prize in physics and a NASA boss who gave him guidance on space systems and substantial freedom.
“I've had some incredible mentors that helped to guide me to where I am today.”