A boom in space-based intelligence is coming. Can ground networks keep up?

ASPEN, Colorado—A flood of space-based intelligence is heading toward U.S. networks as satellite constellations grow and new sensors come online—not just photos, but radar, thermal, and radio data. But to properly exploit it will take new tools, new tech, and even new ways of working with contractors, the head of the National Geospatial-Intelligence Agency says.

“As I look to the future, as we now move in a situation where you're gonna have a constellation that's even larger—yeah, I think exponential is too strong—but where you have a rapid increase in the number of terabytes coming from space over the next eight to 10 years,” Vice Adm. Frank Whitworth, the head of NGA, told Defense One at the Aspen Security Forum in July. “I don't anticipate the number of eyes, the number of brains will increase. I will need something that provides an advantage, and that will be AI … And if you don't have a good enough compute, it's gonna slow down.”

NGA now runs Project Maven, the Pentagon’s seven-year-old effort to use artificial intelligence to filter oceans of data for things the military needs to see and sense. But Whitworth said NGA will lean more and more on the satellite companies that gather the data to help make sense of it.

“Right now it's in our authority actually to purchase commercial analytics for imagery, and not necessarily the raw pixels themselves. We've announced a ceiling of almost $500 million…for analytics,” Whitworth said. 

He said NGA, established to provide intelligence to the rest of the military, is increasingly a multi-source integrator.

“I want commercial. I want international and [I] want national technical means. I've got to find a way to integrate it all, ensure that it is findable,” he said.

The CEO of Maxar, a leading provider of space imagery and data, confirmed that NGA was looking for help analyzing data. AI is part of the solution, Dan Smoot said, but many of the commercially available tools just don’t deliver insights fast enough.

“If you think about analytics companies that are doing advanced AI, they're subscribing to information and they're 48 hours behind,” Smoot said. “But to anybody who's really looking at—you talked about Adm. Whitworth or any of the intel communities or DOD—they're really trying to understand how they can get into a near-term capability to solve an issue on the ground.”

That means two things: being highly receptive to urgent government needs for satellite data in key areas, such as the Second Thomas Shoal, but also proactively providing analysis of emerging situations in other areas as they begin to show up in Maxar’s images and other data.

Smoot called it a “push, pull” model. The pull is the more traditional way that government gets satellite imagery from a commercial provider. 

“They can come to us and sit there and say, ‘We're interested in INDOPACOM, a very specific [area]. What's happened over the last couple days and what's happening right now?” he said.

The push function is newer. It involves pushing intelligence up to the government customer if a new event is emerging in an area that isn’t getting much attention, Smoot said. 

“If there's a change on the ground, how do you actually give that analyst the information quickly, so they can actually disseminate what the change is, versus trying to parse through all information to find the change? And the way that we're approaching it, the way that our insights are gonna work is, we're gonna actually be able to push that change so that they'll [meaning the government customer] understands the information that they're looking for, and be able to subscribe to something,” meaning request more images and data collection in that area.

Smoot said that means Maxar must build partnerships with analytics companies and other providers of sensed data. 

“For tipping and cueing and other things that you want to be able to do, you can actually bring in other sensors. And if you think about what we've been doing with the U.S. government for a very long time, on our [government-to-government] platform stuff that has been providing some virtual capability, not just ours, but others. Now we're just going to put it into more of an intelligence form factor,” he said.

Relying more on commercial imagery, rather than just imagery that comes from government-owned satellites, is also facilitating much faster intelligence- and information sharing with partners and allies, since those commercial entities face lower and different barriers to sharing. 

But all of those activities—collecting more high-definition satellite images through more visits, bringing in more types of data, running analysis quickly and trying to share it more broadly—is creating new challenges elsewhere. They require increasing the speed at which satellites communicate with one another and with the ground (via relatively new technologies like space-based optical communications) and, of course, increasing computing power on the ground as well. 

That’s why Maxar is looking to build new satellites that can process and communicate more data, but also leaning on companies like Amazon Web Services.

“We're going to be working more with the AWSs of the world to try and do the computation faster to the ground…As we're working on our next-generation constellation right now. We're in a build mode [to increase] onboard compute as well as an increase” in optical communications to get data to ground stations faster, he said.  “What you're seeing companies like AWS do, and even KSAT, is they're bringing compute closer to the antenna. … That way, you can actually get faster processing.”  

It’s similar to the way big stock brokerages in the 2000s began to move their operations closer to the stock exchanges’ actual servers to be milliseconds faster on trades.

AWS, meanwhile, is stretching to meet quickly rising compute needs for space data by building more connection points, including connections at the edge, in potential combat locations, for the  U.S. and other allied militaries so operators have more options to give and receive data, Liz Martin, the director of Pentagon sales at Amazon Web Services, said in Aspen. “We think about what are the optionality of connection points, and how frequently does your edge talk back to the bigger cloud? And how much processing can you do there versus how much processing do you need to do back in the cloud?”

AWS is playing a growing role in U.S. military exercises, which are also bringing in more satellite data, so much so that Martin late last year established a team of AWS techs to participate in exercises—and not only with the United States but NATO allies as well. 

“We were being asked to participate in global exercises, so many different components…with each team and each group doing it their own way each time, and they're all over the world, right? So what are the logistics of getting people and classified edge devices and applications and whatnot transported all over the world? We decided to program manage it centrally and run it out of my team,” Martin said.

That’s giving AWS an insight into another difficult barrier to faster information sharing, she said: how to share with foreign partners in a way that still protects U.S. secrets.  

To speed up data transfer, AWS is putting their own satellites in space, via Project Kuiper, 

“It’s not only just the government entities that are space-centric, but you know, there's a number of commercial entities that [rely on AWS.] We work across all of those partners, working towards some innovative new ideas on how we think about computing in space…We're launching our LEO satellites in (the fourth quarter of 2024) in support of that again, the resiliency of marrying the terrestrial infrastructure with the space-based infrastructure,” said Martin.

Unlike Google and Microsoft, AWS is in no rush to put a public generative AI product in front of consumers, particularly one that make embarrassing mistakes. Martin described AWS’s role as making sure that companies and government customers that want to build their own AI tools have the tools as well as the computing power they need.

One of those partners building AI tools with AWS is Maxar. “We build a lot of our app [with AWS] from an application-build perspective, as well as we're starting to use some of their ground-station capabilities and compute capabilities,” said Smoot.