As Electorate Rejects Terrestrial Data Centers, Satellite Industry Considers Space Compute
Small onboard computers could analyze satellite imagery to flag crop problems for farmers and identify potential threats for defense agencies
Akul Saxena
NEW ORLEANS, Nov. 18, 2026 – Voters in 10 states backed restrictions on data center development on Nov. 3. Satellite executives say those restrictions strengthen the case for putting some computing equipment in orbit.
Executives at the Broadband Nation Expo and Data Center Excellence event here took up what that rebuke means for computing in space Wednesday.
Early compute applications
Satellites collect more imagery than their radio links can send to Earth, said Rick Ward, founder and chief technology officer of OrbitsEdge, a Central Florida startup developing small computers that other companies can carry aboard their satellites. Processing the imagery in orbit and sending down only the results gets around that bottleneck.
For a farmer, it’s the information that’s useful. “They need the information, not necessarily the pictures,” Ward said.
Defense agencies are likely early customers, Ward said. Builders cannot start with gigawatt data centers, he said, so the first computing tasks have to be small ones where processing on board pays off.
Early warning is another use case. A satellite that analyzes its own sensor data can send down an alert instead of raw imagery, skipping the wait to move the pictures to the ground first. "If you're doing something like Starshield, you need early warning protection first," Ward said.
Starshield is SpaceX’s satellite line for government agencies.
OrbitsEdge’s Edge1 design weighed less than 4.4 pounds and required 10 watts from its host satellite, Payload reported in June 2025.
Meanwhile, Elon Musk’s rocket company applied to the Federal Communications Commission in January to operate up to one million data center satellites. That is about 59 times the 16,910 satellites the European Space Agency counted in orbit at the time, Broadband Breakfast reported.
Moving from small onboard computers to facilities serving customers on Earth requires overcoming three major constraints: Cooling, launch costs and transmission delays.
The case for orbit starts with scarce power on the ground
A data center brings modest value to the town that hosts it, even when its value to the country is large, Ward said. That mismatch has made the facilities a hard sell locally, and it is the case for orbit, he said.
Power is the other terrestrial bottleneck. A data center goes up in about two years, Ward said. Two nuclear reactors, by contrast, took close to 20 years, he said. For example, the two reactors at Georgia Power's Plant Vogtle took 15 years and $36.8 billion before entering service in 2024, took 15 years and $36.8 billion before entering service.
Until new plants arrive, data centers on the ground rely on gas turbines and large solar arrays while their demand for electricity continues to surge, Ward said.
Heat caps how big a machine can get
Data centers on the ground often draw on local water to cool their servers while a spacecraft uses none. On Earth, servers shed heat three ways, through conduction, convection and radiation, Ward said. In space it’s only through radiation.
Cooling 1 megawatt of computing takes about 13,000 square feet of panel, according to SatNews. That is nearly a quarter of a football field for 1 megawatt. Large data center campuses on the ground run to a gigawatt, 1,000 times more.
Launch sets the price, and supply just tightened
However, computing in orbit costs far more today than on the ground, Ward said. But the launch is a one-time bill, and hardware with few moving parts can run for decades. He called orbital compute a closed-loop system since “it's just mass."
A Vulcan rocket failed Sunday, Nov. 15, destroying 40 Amazon Leo satellites, and could stay grounded for months. Falcon 9 launches from Florida paused for at least six weeks after Oct. 20 to make room for Starship.
Altitude determines which computing tasks can go up
The farther a satellite sits from Earth, the longer a signal takes to make the round trip. That delay, called latency, determines which computing tasks can run in orbit at all. A round trip to a geostationary satellite, about 22,236 miles up, runs 600 milliseconds or more. Low Earth orbit cuts that to 25 to 60 milliseconds.
That gap matters for TCP, the protocol that carries most internet traffic. "There's no TCP-based network application that works better with more latency," said Grant Kirkwood, chief executive of Contrivian, which routes traffic across satellite and ground networks.
Space traffic management
Space traffic management matters more as the technology outruns policymakers, Ward said. The FCC has a role through licensing while civil traffic coordination moves from the Defense Department to the Commerce Department, said Tom Stroup, president of the Satellite Industry Association.
A 2022 FCC rule requires low Earth orbit satellites to deorbit, or reenter Earth's atmosphere, within five years of mission end, replacing a 25-year guideline.
The turnover in Congress
With United Launch Alliance's Vulcan grounded and SpaceX's Falcon 9 paused in Florida, there are few rockets to go around. SpaceX can sell room on a launch to another company or fill it with its own satellites. Computing in orbit could easily take up all available capacity, Stroup said, which raises competition issues.
New members of Congress take office in January. With that significant turnover, he said, policymakers will have a lot of questions about the industry.
Some have called for breaking up SpaceX, which builds both the rockets and the satellites they carry. The model is working, Ward said, and splitting it would serve no real benefit because the pieces would very likely still be each other's customers.
