Despite significant buzz around space-based data centers, the number of GPUs currently in orbit remains low. However, this situation is evolving, and the immediate future of orbital computing is beginning to solidify.
Canada’s Kepler Communications deployed the largest orbital compute cluster in January, featuring approximately 40 Nvidia Orin edge processors distributed across 10 active satellites, interconnected via laser communication links.
With 18 clients, Kepler recently revealed Sophia Space as its latest customer. Sophia Space is a startup planning to test its distinct orbital computer’s software using Kepler’s satellite constellation.
Industry experts predict that extensive data centers, similar to those conceptualized by SpaceX or Blue Origin, are unlikely to materialize before the 2030s. The initial focus will involve processing data gathered in orbit to enhance space-based sensors utilized by both private enterprises and government entities.
Kepler views itself not as a data center provider, but as foundational infrastructure for in-space applications, according to CEO Mina Mitry. The company aims to offer network services for other satellites in orbit, as well as for drones and aircraft operating within Earth’s atmosphere.
In contrast, Sophia is innovating passively cooled space computers. These could address a major hurdle for large orbital data centers: preventing powerful processors from overheating without requiring the development and launch of bulky, costly active-cooling systems.
Under this new collaboration, Sophia intends to upload its proprietary operating system to a Kepler satellite and then deploy and configure it across six GPUs spanning two spacecraft. While a standard procedure in ground-based data centers, this marks its inaugural attempt in orbit. Ensuring the software’s functionality in space is a critical de-risking step for Sophia before its anticipated satellite launch in late 2027.
For Kepler, this partnership validates its network’s capabilities. Currently, the company handles and processes data either transmitted from Earth or gathered by hosted payloads on its own satellites. However, as the industry evolves, Kepler anticipates connecting with external satellites to deliver networking and processing services.
Mitry states that satellite firms are now designing future assets with this model in mind, highlighting the advantages of delegating processing for high-power sensors such as synthetic aperture radar. The U.S. military is a significant client for such services, particularly as it develops a missile defense system reliant on satellite-based threat detection and tracking. Kepler has previously showcased a space-to-air laser link during a demonstration for the U.S. government.
Edge processing, which involves handling data at its collection point for quicker response, is where orbital data centers will first demonstrate their utility. This approach distinguishes Sophia and Kepler from established space entities like SpaceX and Blue Origin, and from startups such as Starcloud and Aetherflux, which are securing substantial funding to build extensive data centers featuring traditional data center processors.
Mitry explained to TechCrunch, “Our philosophy is that inference is more crucial than training, so we prioritize a network of distributed GPUs for inference over a single, high-capacity GPU designed for training workloads.” He added, “If a system draws kilowatts but operates only 10% of the time, it’s not very efficient. Our GPUs, however, are operational 100% of the time.”
Once these space technologies are validated, the possibilities are vast. Sophia CEO Rob DeMillo noted that a Wisconsin city recently prohibited data center construction, a sentiment also gaining traction among some Congressional lawmakers. From their perspective, any restriction on terrestrial data centers enhances the appeal of space-based alternatives.
“No more data centers here [in this city],” DeMillo reflected. “Things are going to get interesting from this point forward.”