Key takeaways:
- Orbital data centers are gaining momentum but remain unlikely to replace terrestrial AI infrastructure at hyperscale.
- Launch costs, supporting infrastructure, and system mass remain the largest economic barriers to orbital computing.
- Space-based solar power reduces reliance on terrestrial electricity but significantly increases deployment complexity and cost.
- Thermal management remains a critical engineering challenge because waste heat must be rejected through radiation rather than convection.
- High-bandwidth communications between orbit and Earth continue to limit orbital data centers serving terrestrial AI workloads.
- Near-term opportunities for orbital computing are strongest in Earth observation, defense, and autonomous space operations where data originate in space.
The Lux Take: Focus on terrestrial solutions rather than orbital data centers
Utilities and data center developers should remain focused on terrestrial infrastructure solutions rather than orbital data centers. While orbital compute may find niche applications in Earth observation, defense, and autonomous space operations, launch economics, thermal management, communications infrastructure, and reliability challenges limit its viability as a large-scale alternative to terrestrial AI infrastructure.
Where are the credible opportunities for orbital compute
As AI workloads continue to grow, data center developers are facing constraints related to power availability, land acquisition, permitting timelines, and water consumption, among others. In response, they are exploring alternate approaches to increase compute capacity, such as creating distributed data center networks or putting compute capacity underwater. Alongside these routes, a growing number of companies are exploring orbital data centers as a potential alternative to terrestrial infrastructure. Interest in the concept has been fueled both by increasing pressure on terrestrial data center infrastructure and by declining launch costs and rising launch activity, driven largely by commercial providers like SpaceX. The concept places compute hardware in orbit, powered primarily by solar energy and cooled through radiative heat rejection, with proponents arguing that it could bypass many of the constraints facing data center development on Earth.
Interest in orbital computing has accelerated over the past two years. Starcloud demonstrated AI inference and training workloads on an Nvidia H100 GPU operating in orbit. Cowboy Space Corporation is pursuing solar-powered orbital AI infrastructure and large-scale orbital compute networks. Axiom Space has launched orbital data center nodes intended to support cloud computing services in low Earth orbit, while satellite operators and infrastructure providers are exploring in-orbit processing capabilities for communications, Earth observation, and defense.
While recent demonstrations and announcements have created momentum around orbital computing, there are questions around the concept’s scalability and capability to absorb some of the growing terrestrial compute load. Here, we assess whether orbital computing can become a viable alternative to terrestrial infrastructure and where the primary technical and economic constraints are likely to emerge.
Launch costs remain the biggest barrier to orbital data centers
Launch economics remain the largest barrier to scaling orbital compute.While AI hardware receives most of the attention, the majority of mass in a large orbital computing platform would come from supporting infrastructure rather than the compute hardware itself. Solar arrays, radiators, structural components, shielding, communications equipment, and power management systems all contribute substantial mass. As a result, overall economics remain highly dependent on launch costs. Concepts like Starcloud assume launch costs on the order of USD 30/kg, whereas current commercial launch pricing remains closer to USD 3,000–USD 3,500/kg based on published Falcon 9 pricing. At scale, this two-order-of-magnitude gap becomes difficult to overcome because noncompute mass rather than the graphic processing units (GPUs) themselves increasingly dominates the system.
Solar power doesn’t eliminate the economics challenge
The energy argument further reinforces this challenge. While orbital data centers offer access to abundant solar energy, capturing and using that energy requires large solar arrays, power systems, and supporting infrastructure that add significant mass. This increases both launch costs and deployment complexity as systems scale.
Thermal management is still a major engineering constraint
Thermal management remains a fundamental constraint despite common perceptions that space naturally solves cooling challenges.Space is cold — temperatures in deep space plummeting to as low as 3 K (–270 °C) — but this doesn’t mean the cooling problem is solved. In a vacuum, heat cannot be removed by convection and must instead be rejected by radiation. While Starcloud demonstrated that hundreds of watts of GPU heat can be managed in orbit, scaling to megawatt- or gigawatt-class systems requires correspondingly larger radiator systems. While companies like Starcloud are exploring liquid cooling and deployable radiator systems, these approaches don’t eliminate the underlying scaling challenge: Larger compute deployments require proportionally larger heat-rejection infrastructure. Even under favorable assumptions, radiative heat rejection is measured in hundreds of watts per square meter, implying that megawatt-scale orbital compute would require thousands of square meters of radiator area, while gigawatt-scale systems would require square kilometers of deployed thermal infrastructure. Since nearly all electrical power consumed by AI compute hardware eventually becomes waste heat, cooling becomes a materials, area, and launch–cost challenge rather than disappearing altogether.
Data transfer limits large-scale orbital computing
Serving terrestrial workloads requires solving a difficult data-transfer problem between orbit and Earth.Orbital compute can create value when data originate in space, particularly for Earth observation, defense, and autonomous space operations, where processing data in orbit reduces downlink requirements. However, serving terrestrial cloud customers requires moving large volumes of data between orbit and Earth. NASA’s TeraByte InfraRed Delivery demonstration achieved optical downlink speeds of up to 200 Gbps, showing that high-bandwidth communications are technically feasible. Yet delivering hyperscale-grade connectivity remains challenging because optical links require line-of-sight connections to ground stations, are affected by cloud cover and atmospheric conditions, and depend on a distributed network of receiving infrastructure to provide continuous coverage. As a result, communications remain a significant constraint for large-scale orbital compute serving terrestrial users.
Reliability and maintenance increase operational complexity
Reliability becomes more difficult once compute infrastructure is deployed in orbit. Terrestrial data centers benefit from routine maintenance, hardware replacement, and controlled operating environments, while orbital systems must operate in radiation-rich environments with limited opportunities for repair or upgrade. Starcloud-1, for example, is expected to operate for roughly 11 months before deorbiting. Scaling orbital compute would therefore require managing hardware failures, technology refresh cycles, orbital debris risks, and end-of-life disposal across potentially thousands of satellites, increasing operational complexity and cost. The same solar energy and cooling equipment would have a far longer lifetime on Earth, meaning capex will go further in terrestrial applications.
Outlook: Orbital data centers will complement, not replace, terrestrial infrastructure
Orbital data centers have progressed from a theoretical concept to early commercial experimentation, but significant questions around economics, scalability, and operational reliability remain unanswered. The most credible near-term opportunities are in Earth observation, defense, and autonomous space operations, where processing data in orbit can reduce transmission requirements and improve responsiveness. Utilities and data center developers should continue prioritizing terrestrial power and grid infrastructure investments, as orbital compute won’t materially offset growing AI-driven demand for data center capacity.
Explore the technologies shaping innovation in 2026
Orbital data centers are just one example of the emerging technologies challenging conventional assumptions about infrastructure, energy, and AI. But not every promising concept is positioned for commercial success.
Download “Rising Technologies Shaping Innovation in 2026” to discover which technologies are gaining real momentum, which are losing steam, and where innovation leaders should focus their investments over the next 12 to 24 months. Backed by Lux Research’s proprietary Tech Signal and expert analysis, the e-book helps separate lasting opportunities from hype.