In many regions, data center developments are now facing seven-year power queues in some markets, water scarcity, and regulatory/community resistance, all of which threaten deployment timelines.

Meanwhile, orbital compute has moved from concept to early operational demonstrations, with recent GPU-in-space demonstrations showing the technical feasibility of orbital AI compute and SpaceX recently revealing its first generation of orbital data center satellites. The question is not whether compute can run in space. The more important question is whether orbit removes greater bottlenecks than it introduces. Can these space-based data centers help bypass terrestrial constraints, and what hurdles still need to be overcome before they become operational?

The growing interest in orbital compute reflects a broader shift in AI infrastructure. For decades, compute itself was the scarce resource. Today, access to power, cooling, permitting, and grid connections increasingly determines how quickly new AI capacity can be deployed. As infrastructure constraints become more important than compute constraints, the conversation has expanded beyond faster chips to alternative architectures, including orbital compute.

The challenges to data center growth

Despite massive investments in AI data centers and a growing need for the compute they provide, deployments are being held back by physical factors. Access to sufficient always-on power, grid connections, and water for cooling are all becoming constraints. Lengthening regulatory timelines and growing community pushback impact construction schedules and increase stakeholder management costs. According to Bessemer Venture Partners, of the 110 data center projects that were expected to go live in 2025, more than a quarter were delayed due to power, permitting, and construction constraints.

These obstacles threaten the economic viability of projects as infrastructure that arrives after the model-refresh cycle delivers diminished returns. At the same time, orbital compute is transitioning from concept to early operational demonstrations. A recent demonstration successfully tested an H100-class GPU payload in space, marking a tangible step toward space-based AI infrastructure. Ahead of its flotation, SpaceX announced its AI1 orbital data center satellite, along with plans to open a factory in Texas to produce them starting in 2027.

Understanding orbital data centers

Low Earth orbit (LEO) satellites form the foundation of orbital data centers. Operating between 400 and 1,400 km (249 and 870 miles) above the Earth’s surface, they travel around the Earth every 90- 120 minutes. Certain orbital configurations can provide near-continuous solar exposure while also offering lower communication latency than deep-space deployments.

Orbital data centers will not replicate hyperscale facilities in space. For example, SpaceX’s AI1 is reported to provide approximately 150kW peak compute power and 120kW on average. In comparison, hyperscale terrestrial data centers offer power capacity measured in gigawatts.

They will instead be built around a modular, networked constellation of compute satellites. These will be designed for workloads where orbit provides structural advantages, such as near-continuous solar exposure to power, a radiative thermal environment that avoids water-based cooling but still requires sophisticated thermal management, proximity to space-generated data, and geopolitical resilience.

The engineering foundations required for orbital success

While hardware is advancing, effective orbital data centers have to be built on six key engineering foundations.

1. Power. While certain orbital configurations can provide near-continuous solar exposure to power satellites, technology has to cope with the harsh conditions of space. High-specific-power, radiation-tolerant solar arrays and resilient energy storage are needed to handle transients and contingency eclipse events within strict mass and reliability constraints.
2. Thermal management. Every orbit faces a few minutes of shadow, meaning temperatures will range from +120°C to -250°C (+ 248°F to −418°F). Efficient thermal management, including heat spreading, conservative power density, and intelligent workload scheduling, becomes key to performance.
3. Compute. Radiation hardening, redundancy, and autonomous operation are baseline requirements for orbital data centers, along with the ability to refresh hardware, such as through swappable units.
4. Network. Non-geostationary module large orbital data centers require robotic assembly of modular units and periodic hardware refresh. Standardization and automation are therefore crucial to operating and refreshing hardware at a competitive rate. Switching on orbital compute
5. Launch economics. Launch costs currently account for about 40 percent of total investment. These are falling, with SpaceX’s Starship targeting sub-$100/kg versus historical rates of $2,000-$10,000/kg, improving the economics of orbital deployment.
6. In-orbit assembly and servicing. Large orbital data centers require robotic assembly of modular units and periodic hardware refresh. Standardization and automation are therefore crucial to operating and refreshing hardware at a competitive rate.Switching on orbital compute

For operators, orbital capacity will sit alongside terrestrial options and will only be used where it removes a greater bottleneck than it introduces. Essentially, it will provide a new data channel, similar to the one emerging in mobile communications with OneWeb, Starlink, and Kuiper.

Currently, the applications for orbital data centers fall into one of three categories. First, in-orbit edge compute processes data such as Earth observation imagery, RF signals, and spacecraft telemetry at the source, allowing operators to transmit insights rather than raw datasets back to Earth. For many orbital applications, reducing the amount of data that needs to be transmitted back to Earth is more valuable than adding raw compute capacity, reducing communications bottlenecks while accelerating decision-making.

Secondly, they enable digital sovereignty and resilience by providing off-planet archives of critical datasets, model checkpoints, and immutable logs for continuity scenarios where terrestrial redundancy is insufficient. Finally, they are well-positioned to handle latency-tolerant batch compute where energy availability is more important than millisecond responsiveness.

Orbital data centers – challenges and opportunities

Orbital data centers are no longer science fiction, but to move to deployments at scale will need to offer a compelling economic alternative to terrestrial data centers. Their additional launch and deployment costs must not exceed the costs of terrestrial delays such as power, water, permitting delays, and timeline risk. They must be able to operate autonomously and efficiently to offer a competitive cost per compute hour, as well as enabling regular refresh rates. Finally, orbital availability, spectrum allocation, and cybersecurity frameworks will shape deployment speed, permissible actors, and operational boundaries.

Increasingly, the data center industry’s constraints are external to technology. Orbital compute will not eliminate every bottleneck, but for specific workload classes, it offers a way to avoid power queues, heat limits, and permit timelines by converting physical constraints into architectural opportunities.

Orbital compute is unlikely to replace terrestrial infrastructure. Instead, it will become another architectural option within the broader AI infrastructure ecosystem, one that makes sense only where it removes a greater bottleneck than it introduces. As technical demonstrations mature and commercial models evolve, organizations should begin evaluating which workloads could benefit from orbital compute as part of a broader hybrid infrastructure strategy.