January 7, 2014
Silicon Valley’s New Spy Satellites
Robinson Meyer
The Atlantic
January 7, 2014
Imagine an energy company which manages a pipeline through Canada’s taiga. The company’s charged with maintaining that pipeline, with making sure it isn’t leaking and hasn’t been compromised. So, every day, the company pays a local to get in a plane and fly over the otherwise inert, massive metal tube, looking for objects, organic or otherwise, that shouldn’t be there.
“[Imagery from] five years ago is great, but how about from last year, last month, last week, yesterday?”
Or that’s what they’ve done for many years. Five years from now, that pilot might be out of a job. Tiny satellites, whizzing over head in low Earth orbit, could photograph every meter of the pipeline. It won’t seem like anyone’s nearby, but, should a truck or stain appear on the ice, a system administrator in Houston would get a text message warning of a problem.
Humans began photographing their home planet from space in a scientifically useful way about a half-century ago. Now the images are ubiquitous: On a web search, in a phone app, on the news, we see the browns and blues that denote pictures taken from the sky. They have rollicked around the culture, spawning both the techno-hippie Whole Earth Catalog and the $3 billion military contractor Digital Globe.
The first picture of Earth taken from space, above, was photographed by a V-2 missile 65 miles in the air on October 24, 1946. Below, a recent image of farms taken by one of Planet Labs’s doves in Washington.
“Google Earth whetted consumers’s appetites for pictures of Earth from space,” Scott Larsen told me. But the pictures in our browsers, he said, have now become old and out of date.
“[Imagery from] five years ago is great, but how about from last year, last month, last week, yesterday?’”
Larsen leads Urthecast. It’s one of a cadre of startups—three are now out of stealth mode—tossing cameras out of the atmosphere and trying to turn them into a business. Each of the three is choosing different methods, different kinds of devices, and different orbits. Each is selling something a little different. They areUrthecast, Planet Labs, and Skybox.
Urthecast, for instance, plans to install two cameras—one still and one video—on the International Space Station, then beam video down using the Russian Space Agency’s antennae. Planet Labs, another, hopes to send 28 satellites, each about the size of a garden gnome, into low orbit. It will immediately control the largest private Earth-observing fleet of satellites ever created. SkyBox, finally, only hopes to operate two satellites in the next year—but its business plan seems most promising, and borrows the most from the modern startup playbook.
The capital and efficiency engines of Silicon Valley, having transformed markets and interactions both public and private on Earth, now look skyward.
Silicon Valley is making what, in any other decade, we’d call spy satellites.
Orbiting Cameras, Operated by Behemoths
Four government entities and two companies contributed imagery to this Google Maps image. (Google)
Like the geography they depict, it can be easy to feel like the pictures in Google Earth have always been there, like they don’t have a history or a source.
In fact, a set of accreted structures dictate what and how satellite imagery reaches American consumers. Some of these limits are governmental: The U.S. government restricts the resolution of imagery that can be sold to the public, for example. Today, if the side of one pixel of Earth imagery is less than 50 centimeters long, it can’t be sold in America. (Though regulators may soon loosen those limits.)
The marketplace limits buyers still further, though. If you go looking for medium-resolution imagery, you’ll find it a hard task. If you want a picture of your house—or your pipeline—who do you buy it from?
You have a few options. You might download it for free from weather satellites, like those in NASA’s MODIS program. MODIS imagery is quite low-resolution, though: It will give you good (sometimes beautiful) photos of the clouds over your home state.
MODIS imagery of the northeast United States after last week’s snowstorm (NASA)
Other options are prohibitively expensive. The pan-European consortium that owns the airplane-manufacturer Airbus also operates a satellite company calledAstrium, which sells imagery commercially. The German company Blackbridge and the Israeli ImageSat do the same. A few governments—Taiwan, India, South Korea—will sell imagery from their spy satellites to you.
Digital Globe’s WorldView-2 satellite
(Illustration from Digital Globe)
There’s also data from some other U.S. government-run projects, like the long-running Landsat program. That imagery’s free, and sometimes recent—but it’s not very high-resolution. Really, there’s only one American company that will sell you imagery in the medium- to high-range: Digital Globe.
Digital Globe is a public corporation, worth about $3 billion and based in Colorado. Most of its business flows from the Department of Defense: Last fiscal year, a full 60 percent of its revenue came from just one contract with the National Geointelligence Agency. It also supplies a substantial amount of imagery to Google Earth.
Digital Globe specializes, in other words, not just in big institutional buyers, but in the biggest of the big. It makes sense: In the modern Earth observing business, most of the players are enormous. Even in the list I just gave, most of the institutions either maintain a standing army or manufacture warplanes.
The wrecked Italian cruise vessel Costa Concordia, as captured by Digital Globe’s WorldView-2 satellite
And it makes economic sense, too, for these satellites operators to be behemoth. Not only is operating a satellite expensive, but sending it to space is maybe moreso. You can’t buy rocket fuel from a gas station.
Let’s say you want to compete with one of these leviathans. You have a few options to make it cheaper. Here’s one: Find a satellite that one of those huge institutions regularly services. Strap your cameras on to it. Get one of those institutions to pay for it all.
Replace “satellite” with “International Space Station,” and you have Urthecast.
Like a Floating Geyser of Free Money
The Russian section of the International Space Station (NASA)
Of all the satellite startups, Urthecast’s business is probably the easiest to understand. Soon after the publication of this article, perhaps in a matter of days, astronauts will venture outside the International Space Station (ISS). They will bolt two cameras, each about the size of a two-liter soda bottle, on to the Russian part of the structure. They’ll connect some wires, presumably, push some buttons, and then the two devices will start sending both still and moving imagery below, to the same planet that they watch.
It’s a lovely idea. It is also ludicrously inexpensive.
“I tend to think we’ll get pictures no one’s ever seen before.”
Four years ago, Scott Larsen, now Urthecast’s CEO, was working at Merchant Bank in Vancouver, when—after a series of introductions—an inquiry from the Russian space agency (the RKA) “came across [his] desk.”
“They were interested in seeing what was possible,” he said, in putting a camera or two on the ISS. They encouraged him to run a feasibility study for half a million dollars, then stopped corresponding. That was, until:
“In the fall of 2010, they called up and said, hey: If you wanna do it, we wanna do it,” said Larsen.
What was that “it”? An incredible deal. If Urthecast paid to build 2 cameras, the RKA would handle not only getting the devices to space but also the cost of their installation and their downlink. Downlinking data from satellites is incredibly costly, often a major limiting expense for Earth-observation companies. For Urthecast, it would be free.
Or nearly free: Once installed, the RKA would keep all the imagery Urthecast’s cameras capture of Russia. Urthecast can sell the imagery of anywhere else on Earth.
Urthecast, in other words, has a monopoly of sorts, on high-quality pictures and video from the ISS. The Russian space agency has created a giant geyser of free money, in low-Earth orbit, for the benefit of a private company.
The Soyuz rocket bearing Urthecast’s cameras launching on November 25 (Urthecast)
Urthecast incorporated in Canada on Christmas 2010. It went public before it put anything in space, in June of this year, raising about $9.3 million U.S. dollars. It’s now a publicly-traded company worth about $120 million.
At the end of December, cosmonauts attempted to install Urthecast’s cameras onto the ISS. It didn’t work.
The ISS is an unusual satellite to strap a camera to—it flies an orbit lower and more elliptical than other Earth observing satellites.
“I tend to think we’ll get pictures no one’s ever seen before,” said Larsen. Once that imagery is online, he told me, “everyone will come to the website at least once.”
Because of that—and because of the company’s virtual monopoly—Larsen said the company anticipated all manner of monetization strategies: ads on the website, broadcast deals, and a full range of content partnerships
“We are a media company,” he said, “there’s no doubt about it. We’ve done something no one’s done before.”
If they do do it. At the end of December, cosmonauts attempted to install Urthecast’s cameras onto the ISS. It didn’t work—the cameras failed to relay signal to the ground. Now the devices are back inside the station, being examined for problems.
If they get the cameras working, Urthecast will find buyers. TV news agencies, entertainment companies, app developers are in talks with Urthecast: According to Larsen, the company’s already signed an agreement with the United Nations.
A Flock of Doves
Planet Labs already has two nanosatellites in orbit—one captured this image of forests near Gales Creek, Oregon in April. (Planet Labs)
Befriending the RKA and bolting something to the world’s most famous space-laboratory is one path to profitability. Planet Labs, based in San Fransisco, is trying another. In January, it hopes to send 28 micro-satellites in space—the largest constellation of private Earth-observing satellites, ever. (The incredibly awesome collective noun for a group of satellites is a constellation.)
Planet Labs can send that many into space because the satellites are designed for extensibility. 10 by 10 by 30 centimenters, Planet Labs’s “doves” are derived from Cubesat technology—the set of tools and standards developed at Stanford in the early 2000s that let universities affordably fly small, low-orbit satellites.
More than 100 have been put in orbit since the idea’s inception. When high schoolers in Virginia built a satellite and launched it, as happened last month,they did it with Cubesats.
One of Planet Labs’s doves rests, above. Another, below, deploys its solar panels. (Planet Labs)
So Planet Labs’s doves are cheap. They’re so cheap, in fact, that Planet Labs can anticipate some portion of them never working.
“People say, if you never miss a flight, you’re spending too much time in the airport,” the company’s CEO, William Marshall, told me. “Well, if you never lose a satellite, you’re not pushing the envelope.”
Businesses in space have historically erred very conservative, and for good reason. Every machine, every orbiting device, cost hundreds of millions of dollars; every major technology investment was a bet-the-company adventure.
Marshall told me that 20 percent of the doves can fail in orbit—can never work at all—without the company losing a meaningful amount of imaging capacity. There’s a theory of programming—worse is better—that holds that cheap, simple software can be faster and easier to use than the expensive, complicated fare. Planet Labs applies this to space.
Planet Labs’s complete constellation—“flock”—of 28 doves (Planet Labs)
It’s an application with huge potential benefits. With its horde of doves, Planet Labs can image the entire planet much more frequently than is now possible. It can refresh its satellites much more frequently: It can almost always just throw more satellites up there, to the point of diminishing returns.
It also has downsides. The imagery Planet Labs’s doves capture will never be great—current imagery technology and the size of Cubesats doesn’t create pictures below a resolution of one meter. Urthecast’s video camera will be better.
But Planet Labs might be able to succeed on other fronts: Marshall has broad, humanitarian ambitions for his company’s data. He said that, often, subsistence farmers in the developing world could use imagery of their farms, but currently can’t afford it. The usual non-military buyers for Earth imagery are banks and large financial firms.
Farmland and a national highway in western Poland (Planet Labs)
Marshall thinks Planet Labs can sell to both financial firms and farmers because the two have different needs, because they need different kinds of imagery at different times. As imagery ages, he said, its price might decrease.
He also that Planet Labs might also be able to lend its imagery to humanitarian causes, such as the crowdsourced mapping that helped victims of Typhoon Haiyan.
But is that a viable business? Maybe.
Pixels or Intel?
There are two ways to sell satellite imagery. The first is to sell the pictures themselves: The actual pixels that depict oceans, mountains, and cities. When a satellite company sells its imagery to Google Earth, this is what it’s doing.
“Here is money.” “Thanks! Here are pixels.” (Or, that’s how the transaction occurs in theory.)
Shizuoka Prefecture in Japan, as photographed by the second of Planet Labs’s doves. Note Shizuoka stadium in the lower left, in the city of Kakegawa. (Planet Labs)
But there’s another way to make money off satellite imagery. A company could take pictures and analyze them for certain objects or trends, then sell the information it gathers.
For instance (and this is something of a canonical example in imagery analytics): With good-enough satellite imagery, you could photograph every Wal-Mart parking lot in America on Black Friday, then analyze the images in total to detect the number of cars. Take that car-count, guess at the typical income bracket associated with each lot, throw in some other variables and voila: You have a good sense of how much money the retail behemoth made on one of the most important shopping days of the year.
“Combining weather data with satellite imagery with other things… That’s really the future of our industry.”
There are other applications. Over the course of season, a keen firm could take a picture of every cornfield in the midwest and estimate how much corn the U.S. produces in total, then trade futures against that knowledge.
It’s a great business, this selling intelligence on the global supply chain to the people betting about it. It’s also, for now, a somewhat theoretical one. Interpreting images with algorithms, especially at the specificity needed, remains incredibly difficult. Facebook bought a facial recognition company earlier this year for an undisclosed price in part so it could apply its algorithms to recognizing all sorts of photos that look nothing like faces. That’s how treasured the algorithmic-image detection technology is.
(Think: With really good algorithms, your spambots could defeat any image-based CAPTCHA.)
The first of Skybox’s satellites captures Perth, Australia—and parking lots full of cars!—in early December. (Skybox Imaging)
And without algorithms to read imagery, the costs of image interpretation stay stubbornly high. Meanwhile, other applications—the kind requiring looking simultaneously at millions of pixels—remain impossible. So if you want to analyze images at any scale, you might be able to do it with an army of interpreters. Might.
When I asked Marshall about whether Planet Labs would sell pixels or intel, he demurred.
“We’re certainly entertaining both,” he said—maybe selling “raw imagery, and analytics on top of it.” He called the company in “the early stage of defining our products.”
Perhaps it will. But maybe the dichotomy is a red herring. You don’t need to actually analyze images successfully in order to build a business around the possibility of doing it.
Cloud Computing, Except for Literal Clouds
The coastline of Somalia, as captured by Skybox’s first satellite (Skybox)
Enter Skybox. Based in Mountain View, California, it put its first satellite in spacein November. But when I talked to Dan Berkenstock, one of its co-founders, he didn’t mention satellites at first.
“There’s never been a good sandbox in this industry,” Berkenstock told me, referring to imagery analytics.
What did he mean? Something like this: In the past half decade, companies have sprung up around Amazon’s hosted cloud services. Amazon operates massive web servers, tools so adaptable you can build a company around them without having to actually pay to operate a data center. Amazon lets its customers pay for what they need, or can use, and no more.
Skybox wants to make something like Amazon’s cloud products, but for the Earth. To do that, Berkenstock told me, it has combined historic data from US government satellites, aerial photography, and weather centers. It’s created a recent history of the Earth in data. It will then augment that record with data and imagery from its own satellites.
Infrared imagery, from SkySat-1, of a gold mine in Western Turkey. Infrared imagery helps experts deduce the health of plants—what “should” be green in the image above appears red. (Skybox)
Eventually, companies will be able to run their own algorithms in the massive Skybox database (which I’ll resist calling Skynet). They can build applications and programs within and with it, using its historic records. They can, crucially, test their image-detecting algorithms within a massive sandbox.
“Combining weather data with satellite imagery with other things,” Berkenstock told me. “That’s really the future of our industry.”
Skybox’s first two satellites in its clean room (Skybox)
Perhaps it is. Regardless, it’s savvy. “Earth-data-as-service” lets Skybox sell the potential of analytics without actually… having to do a lot of analytics.
It also lets them slowly build up a fleet. Skybox has constructed its satellites—each about the size of a dorm room fridge—out of parts you’d find in a hardware store. It uses cheap and available supplies, in other words, rather than costly space-tested materials preferred by military contractors. It runs its mission control in a Google Chrome app that uses the web development kit Bootstrap.
Skybox’s first satellite, SkySat-1 is already functioning and in the air. It’s sending down pictures—and video—now.
Skybox seems, to me, to most apply the lessons of the past decade of web technologies to space. But it also operates the most expensive satellites.
Pictures Aren’t Knowledge
The marketplace coming into view already bends across many variables. Urthecast seems safe: As long as its technology works, people will always want pictures from the ISS. The company is camped securely, empty water bottles i hand, next to its floating free-money geyser.
Planet Labs seems unsure of its business strategy, at least in public, but its satellites are so inexpensive that it can always chuck more at its problems. And SkyBox walks a middle path, armed with a brilliant business idea and a couple satellites neither costly nor cheap.
Whatever happens, these three companies are likely to create a market beneath the current leviathans and not supplant them.
“There will always be a market for Digital Globe,” Larsen told me. “There are still gonna be people who want that.”
In the final days of 2013, Skybox became the first company ever to release HD video from space.
But for many customers, he said, “good enough is good enough. It doesn’t need to be calibrated to 45 cm. It doesn’t need to be as accurate as what they’re getting.”
Who will buy this new, good enough imagery? Why would they buy it? When I talked to Skybox’s founder, Dan Berkenstock, he said many companies wanted satellite imagery to answer one question:
“‘Is my stuff okay?’ That’s the central question satellite companies can answer. All these things around the world that are supposed to be happening or not supposed to be happening that affect me—are they?”
“When you talk to astronauts, they all say going to space changes them.”
So perhaps many people will buy the picture or intel, and all three of these companies will flourish. Or perhaps one will absorb the others, or a popping technology bubble cut off funding to all three. Business plans don’t anticipate history—but sometimes they gesture at it. Very soon, it’s likely that it will be substantially easier to buy recent, high-resolution imagery of the Earth’s surface. Large companies will have access to the same global quasi-omniscience as the world’s powerful governments. If successful, these startups will consolidate power.
Nanosatellites, like those used by Planet Labs, over the Earth (NASA)
Cheap satellites will also grant humanity pictures of the world it has made and is making—a world wracked by people. Every satellite CEO I talked to mentioned being able to track melting glaciers, changing rivers, or dying forests. Satellite technology has proven useful in humanitarian crises; some research indicates it can detect a genocide in progress.
How do we—as citizens, as consumers, as possible customers—handle that, the tension between the power of new knowledge and the power of consolidated knowledge? Over the next few years, we, and our governments, will find out.
At the end of our call, Larsen spoke to me about Urthecast’s aim.
“When you talk to astronauts, they all say going to space changes them,” he said. This was especially true of repeat astronauts—“people who go up twice—go up once, then go back 4 to 5 years later, and see how things have changed.”
They all spoke, he said, of wonder.
“When people are continually reminded how small Earth is, how we’re all living together, people come back with a sense of planet stewardship.”
Planet stewardship: a beautiful idea, something the future hungers for, and something these startups may or may not provide.
No comments:
Post a Comment