Despite the excitement surrounding space-based data centres, the reality is that there is currently a limited number of GPUs orbiting the Earth. However, the landscape is gradually evolving, with businesses starting to explore the potential of orbital computing.
In January, Kepler Communications from Canada launched the largest compute cluster in space, consisting of 40 Nvidia Orin edge processors across 10 active satellites. These satellites are interconnected using advanced laser communication, allowing for efficient data processing in orbit. Kepler has secured 18 customers, including its latest partner, Sophia Space, a startup developing innovative software for orbital computing.
Industry experts predict that large-scale space data centres, akin to those proposed by companies like SpaceX and Blue Origin, may not materialise until the 2030s. For now, the focus will be on enhancing data processing capabilities for sensors used in space, beneficial for both private and government entities.
Kepler envisions itself not merely as a data centre but as essential infrastructure facilitating applications in space. Their CEO, Mina Mitry, emphasises the company’s role in providing networking services for other satellites and aircraft.
Sophia Space is tackling one of the critical challenges for orbital data centres: cooling high-performance processors. The startup is developing passively-cooled space computers to prevent overheating without relying on heavy and costly active-cooling mechanisms. As part of their collaboration, Sophia will test its proprietary operating system on Kepler’s satellites, marking a significant milestone in testing software for space-based applications—an operation previously only achievable in terrestrial data centres.
This partnership also underscores Kepler’s utility in orbit. Currently, it processes data collected from the ground or from payloads on its satellites. As the sector advances, Kepler anticipates collaborating with third-party satellites to expand its networking and processing services. Mitry highlights that satellite operators are beginning to design future capabilities around this model, particularly for processing demanding sensors like synthetic aperture radar, which is vital for military applications, such as missile defence systems.
The concept of edge processing—handling data where it is generated for more immediate responses—is where these orbital data centres can make an impact. This approach distinguishes Kepler and Sophia from established players in the industry pooling resources toward larger data centre models.
Mitry believes in a future where distributed GPUs will perform inference tasks continuously rather than relying on a single powerful GPU for training, which often goes underutilised. He argues that their method of operating GPUs at full capacity will be more effective.
As these technologies gain traction in space, the potential for innovation expands. Sophia’s CEO, Rob DeMillo, notes the growing restrictions on data centre constructions on Earth, suggesting that such limitations make space-based alternatives increasingly appealing, further hinting at a pivotal shift in the data processing paradigm.
Fanpage: TechArena.au
Watch more about AI – Artificial Intelligence
