Orbital data centers and space solar power: the next frontier for AI energy
Generative AI is driving a surge in compute demand, and with it a huge appetite for electricity and cooling. Data centers on Earth are already straining local power grids and raising sustainability questions. Some companies are now pitching a radical fix: move compute and its power source into orbit.
Aetherflux, a space-based solar power startup, has outlined a plan to build modular satellites to harvest solar energy in low Earth orbit and to use that power for computing off-world. The firm calls its initiative the Galactic Brain, an effort to place data center nodes in orbit where cooling and continuous power can be handled differently. Aetherflux aims to launch sunlight-harvesting satellites to low Earth orbit in 2026 and to put the first Galactic Brain node up by around the first quarter of 2027.
The company favors many smaller satellites in low Earth orbit rather than massive arrays at geostationary altitude. Those LEO satellites will pass through Earth’s shadow, but a large constellation is intended to smooth out those gaps and provide steady energy. Instead of radio transmission, Aetherflux plans to beam power back using infrared lasers.
That same approach, the company suggests, could free compute from the terrestrial constraints of heat and power distribution. Aetherflux says its technology will “remove the limits faced by Earth-based data centers.” The firm describes space-based solar power as a “foundational” part of its Galactic Brain project.
The newly proposed orbiting data center idea is not unique to Aetherflux. Established figures in tech and aerospace have signaled similar ambitions and made moves in that direction. Earlier this year, former Google CEO Eric Schmidt acquired a space startup and positioned himself to pursue orbital hardware for compute capacity.
SpaceX has also publicly flagged the concept, noting its Starlink constellation and laser-linked satellites as potential infrastructure for on-orbit data handling. “Simply scaling up Starlink V3 satellites, which have high-speed laser links, would work,” Musk on social media platform X. SpaceX’s existing launch cadence and satellite experience give it a significant head start if it follows through.
Aetherflux’s founder and CEO Baiju Bhatt, who co-founded Robinhood, has framed the push as part of a broader race for compute and energy. “The race for artificial general intelligence is fundamentally a race for compute capacity, and by extension, energy,” Aetherflux founder and CEO Baiju Bhatt said in the company statement. The startup pitches orbit as a place where power and compute can scale without the same terrestrial bottlenecks.
Bhatt has also taken the argument into geopolitical terms on social posts, warning about global competition for space-based capabilities. “The race for space solar power is accelerating, and we intend to lead at Aetherflux,” he . “China is taking it seriously, and the U.S. Government should prioritize investments in space solar power to ensure America doesn’t cede energy and space leadership to China.”
Practical hurdles remain large: launch costs, orbital congestion, regulatory regimes, and the engineering of long-duration, high-power systems in orbit. Earth’s orbital environment is already crowded, and adding power-beaming satellites and data center nodes raises collision and debris risks. Still, proponents argue those technical and policy challenges are solvable given the potential payoff for AI-scale compute.
Analysts warn the demand for data center power is poised to surge as AI workloads scale, with some estimates projecting dramatic increases within the decade. One major financial firm estimated AI-driven energy demand could rise 165 percent by 2030. Whether orbital systems become a mainstream part of that supply mix will depend on engineering progress, economics, and how regulators balance innovation with orbital safety.
https://x.com/elonmusk/status/1984249048107508061