Recent advancements in space technology are reshaping the landscape of data centre capabilities, with a growing focus on orbital computing. While there has been significant excitement around establishing data centres in space, the current number of graphics processing units (GPUs) in orbit remains limited. However, change is on the horizon as initiatives like Canada’s Kepler Communications take shape.
In January, Kepler launched the largest compute cluster in orbit, featuring around 40 Nvidia Orin edge processors across ten satellites, interconnected via laser communication technology. Now supported by 18 customers, including the newly announced partnership with startup Sophia Space, Kepler aims to provide a robust infrastructure for space applications rather than serving as a conventional data centre.
Sophia Space is working on innovative passively-cooled computers designed to mitigate overheating issues faced by powerful processors in space. As part of their collaboration, Sophia will upload its proprietary operating system to Kepler’s satellites, deploying it across six GPUs on two spacecraft. This will mark the first attempt to operationalise software in orbit, crucial for de-risking prior to Sophia’s first satellite launch in 2027.
Kepler’s CEO, Mina Mitry, emphasises their goal of establishing a network service layer supporting other satellites and aerial vehicles. Current operations involve data processing from terrestrial sources or payloads aboard their satellites, but as the industry matures, collaborations with third-party satellites for processing services are anticipated. The implication is that satellite companies are beginning to design their future projects around this model, benefiting from offloading tasks from power-intensive sensors, particularly for military applications.
Processing data in orbit enhances responsiveness, a crucial feature for applications such as synthetic aperture radar used by the U.S. military for missile detection. Kepler has already demonstrated capabilities for space-to-air laser connectivity to the U.S. government, showcasing the potential value of edge processing in space.
In contrast to larger space companies like SpaceX and Blue Origin, Kepler and Sophia are focused on immediate, practical applications rather than large-scale data centres. Mitry advocates for a distributed approach, highlighting the need for numerous GPUs that conduct inference rather than relying on one powerful GPU designed for extensive training loads. This efficiency allows their GPUs to operate at maximum capacity continuously, in stark contrast to traditional systems.
Recent legislative measures, such as Wisconsin’s ban on new data centre constructions, could position space-based alternatives as increasingly appealing solutions. Sophia’s CEO, Rob DeMillo, notes the significant shift towards space computing, suggesting a future where terrestrial facilities struggle to compete with their orbital counterparts.
As these technologies continue to evolve, the potential for orbital computing remains vast, setting the stage for a new era in data processing.
Fanpage: TechArena.au
Watch more about AI – Artificial Intelligence
