Low Earth orbit is undergoing the fastest expansion in its history. However, the nature of the payload is changing. We are moving past simple “bent-pipe” communication relays — where a satellite simply receives a signal and beams it back to Earth without processing it. With initiatives like StarCloud deploying and training a large language model in orbit, the era of the space data center has arrived. Broadband constellations and on-orbit processing missions are now multiplying faster than their underlying data-handling architectures can evolve.
The problem: You can’t put RAID in a cubesat
As we deploy high-performance computing (HPC) nodes in space, a critical vulnerability emerges. On Earth, data resilience is achieved through Redundant Array of Independent Disks (RAID), a data-protection method used in computing, where information is split and stored across multiple drives so that if one drive fails, the data can still be reconstructed from the remaining ones. If one drive fails, the data survives.
In space, this is impossible with a standalone node. A single satellite — restricted by mass, power and volume — cannot house the physical redundancy required for internal RAID. If a satellite carrying processed AI models or irreplaceable sensor fusion data suffers a radiation event or kinetic impact, that data is lost. Current designs still assume the satellite is an isolated device, a model that is a holdover from the early days of smallsat missions. Even advanced constellations such as Starlink, which use optical inter-satellite links to provide internet services, do not operate as distributed storage clusters.
The solution: The constellation becomes the computer
The industry’s current redundancy model — relying on ground replication or spare satellites — is out of step with constellation-scale operations. We must stop treating the satellite as the unit of storage and start treating the constellation as the storage medium.
A satellite constellation is, effectively, a spatially distributed computer. My own research suggests that a ‘constellation-level RAID’, or what I term an “Orbital Redundant Array of Independent Devices (O-RAID)”, is the only mathematical way to guarantee data survival without ground intervention. Disclosure: The author is the inventor of O-RAID, a patent-pending architecture (US Provisional Application No: 63/934,397). By striping data fragments and parity bits across multiple neighboring satellites via Inter-Satellite Links (ISL), we create a self-healing cluster. If one node burns up or fails, the surrounding nodes can mathematically reconstruct the lost data on the fly. This resilience is achieved through a coordinated mix of storage satellites that hold primary data fragments, parity satellites that preserve encoded redundancy and coordinator satellites that manage versioning and consistency across the cluster. Together, these roles form a distributed architecture that maintains data integrity even as individual spacecraft experience failures or orbital attrition.
Future horizon: The case for geostationary moon orbit
This architectural shift is a prerequisite for our next giant leap: the Lunar economy. There is a rush to establish data centers on the Moon to support Artemis and commercial lunar payloads. However, the greatest challenge to a lunar surface data center is the descent. Soft-landing racks of sensitive servers on the lunar surface is disproportionately expensive and fraught with risk.
A more elegant solution is the utilization of a Geostationary Moon Orbit (GMO). Instead of landing servers, we can place data nodes in a stable orbit around the Moon which remains a largely pristine orbital environment with minimal debris and far less congestion than Earth’s orbital shells. A GMO data constellation creates a high-bandwidth “orbital cloud” that services surface assets without the risk, complexity, or cost of landing. This idea builds on existing research into cislunar dynamics — particularly work examining Earth–moon Lagrange Point 1 (EML-1) as a location for station-keeping and communication relay satellites. EML-1 has long been studied as a stable operational outpost for standalone spacecraft. The same principles can be extended to create redundant satellite clusters, where fault-tolerant data nodes work together to preserve continuity of service even if an individual spacecraft fails. In this context, a GMO or Lagrange-point data layer becomes a natural extension of current cislunar mission architecture.
- The architecture: Surface rovers upload raw data to the GMO cloud for processing and storage.
- The backbone: GMO nodes link back to Earth via high-throughput Gateway nodes in Earth’s Geostationary Orbit (GEO).
- The resilience: Just like in LEO, these GMO nodes would utilize distributed parity. If a lunar orbit node fails, the data is preserved across the arc.
Who should pursue this?
This redundancy architecture is not aimed at any single entity. It speaks to a broad class of operators who rely on high-integrity data. Commercial service providers exploring orbital data centers, including operators planning massive AI-capable clusters in space stand to benefit from a fault-tolerant storage layer that does not depend on the ground. Moreover, government agencies could use constellation-level redundancy as a resilient backup for critical national records and citizen data. Financial institutions and other operators of essential digital infrastructure could adopt it as an off-planet disaster-recovery site. It could also serve as a secure long-term archive for critical medical datasets, such as genomic, epidemiological and biomedical research records, whose preservation is increasingly important for future healthcare resilience and for ensuring these assets endure even under extreme or existential risks to terrestrial infrastructure.
More broadly, relocating vital information to orbit is a strategic response to the growing risks facing centralized terrestrial data centers, including increasingly frequent climate-driven disruptions, escalating water and power consumption, deliberate sabotage of communication infrastructure and the broader vulnerability of physical infrastructure to extreme environmental events. In this sense, the proposal is a roadmap for any organization whose mission depends on the long-term survival of digital information.
Whether in LEO, MEO, GEO or GMO, the physics of space demands that we abandon the idea of the “standalone” satellite. As operators invest billions into multi-decade deployments, failing to modernize redundancy will hold back future missions. The conversation on distributed resilience needs to begin now.
Ravinda Meegama is a senior professor in computer science at the Department of Computer Science, Faculty of Applied Sciences, University of Sri Jayewardenepura, Sri Lanka. He has BSc degree (First Class) in Computer Science from University of Colombo, Sri Lanka, MSc from AIT, Thailand and PhD from NTU, Singapore. His research interests primarily focus on exploring the intersection of image processing and artificial intelligence.
SpaceNews is committed to publishing our community’s diverse perspectives. Whether you’re an academic, executive, engineer or even just a concerned citizen of the cosmos, send your arguments and viewpoints to opinion (at) spacenews.com to be considered for publication online or in our next magazine. If you have something to submit, read some of our recent opinion articles and our submission guidelines to get a sense of what we’re looking for. The perspectives shared in these opinion articles are solely those of the authors and do not necessarily represent their employers or professional affiliations.
