Data centers in space

getty

In some communities, there has been growing backlash against the construction of data centers, seen as gobbling up too much power, too much water, and even creating some forms of pollution. In addition, there are questions whether all the new data centers dotting the land, along with adjoining power sources, can even meet the demand for AI in the near future. These concerns are causing second looks, skyward, at the efficacy of positioning data centers in orbit.

The possibility has been pushed aggressively by Elon Musk, chairman of SpaceX, who envisions his company employing a constellation of space-based data centers in the near future. The economics make it compelling he said, predicting back in April that “the cost of deploying AI in space will drop below the cost of terrestrial AI much sooner than most people expect. I think it may be only two or three years.”

Notably, SpaceX filed for FCC approval at the beginning of this year to launch and support a constellation of one million data center satellites.

“It is easier than ever to get data into orbit,” a new report out of JLL states. “The challenge now becomes managing the logistics of an orbital data ecosystem, mirroring the subsea cable deployment revolution of the 1990s that enabled global internet connectivity. Just as those undersea networks moved data infrastructure closer to where information was generated and consumed, the proliferation of satellites is creating a parallel dynamic above Earth's atmosphere.”

However, there are also a number of obstacles that get in the way of such a vision. Launch economics, operational reliability, and debris management are not yet favorable to the idea of space-based data centers, the report’s authors, led by Andrew Batson of JLL, caution.

Do the economics make sense? Not yet, but there’s potential, the JLL authors state. For the economics, to become meaningful, a weighted launch payload of $500 per kilogram is the threshold where space-based compute could become cost-competitive. SpaceX’s Starship program is targeting $200 per kilogram, versus $2,700 per kilogram for the Falcon 9 rocket today.

The economics may be more favorable as these efforts scale up. One set of estimates from Starcloud, cited in the report, suggests a 40 MW space-based data center could deliver $138 million in potential energy savings over 10 years in space. "Orbital data centers propose a fundamentally different energy equation, the JLL team suggests. "Continuous exposure to solar radiation in low Earth orbit offers the potential for abundant, predictable power generation without grid interconnection, backup generation, or terrestrial permitting constraints."

There are latency issues with moving data from outer space and back as well. “Modern Earth Observation satellites capture terabytes of data a day,” the report’s co-authors state. “Sending raw data to earth for processing is slow and expensive, often limited by alignment to ground stations.”

This may call for enhanced edge-processing and analysis capabilities in space as well, "processing data in orbit and sending only results back to earth rather
than raw data," they state. “Data centers in space will likely act as edge computing solutions for on-orbit processing reducing downlink bandwidth pressure.”

Cooling in space is another issue. Space is naturally extremely cold, but there is a "fundamental engineering challenge: heat cannot be convected away from equipment without air,” the report states. “Instead, waste heat must be emitted through large passive radiators, requiring far greater surface area than terrestrial systems.” This means more up-front costs for cooling systems.

.