The race to build out artificial intelligence infrastructure is rapidly running into terrestrial limits. Communities are pushing back against new data center construction, the electrical grid is straining under unprecedented compute demand, and water resources used for cooling are growing scarce. In response, a new frontier is opening above our heads: orbital data centers, or ODCs, which place racks of high-performance compute on satellites in low Earth orbit and beyond.
A Marketplace Model for Space Compute
The emerging orbital data center industry is being structured to feel familiar to enterprise customers. Instead of a bespoke aerospace procurement process, the model resembles walking into a cell phone store: a customer selects a compute configuration, picks a connectivity speed, and receives a monthly bill. Behind the scenes, a marketplace platform aggregates supply chain partners — satellite bus manufacturers, propulsion vendors, and other space-qualified component providers — and assembles an optimized vehicle for the customer's needs.
The contracts are designed to mirror terrestrial agreements as closely as possible. A typical master service agreement runs for five years and is built around standard data center units. A "full rack," for example, corresponds to a 42U configuration drawing roughly 105 kilowatts, but instead of sitting in a warehouse, it is integrated into a satellite chassis and launched into orbit.
Hardware for a Hostile Environment
The physical reality of operating in orbit is uncompromising. At an altitude where data center satellites operate, objects move at roughly 17,000 miles per hour — Mach 21. At those velocities, even a 3-millimeter aluminum sphere becomes a projectile capable of punching a clean hole through conventional materials. Protecting expensive compute chips from micrometeorite and debris impacts therefore requires specialized shielding, and one of the most important hardware enablers of this industry has been the development of extremely light, thin space armor designed to absorb such impacts while keeping the satellite's mass budget manageable.
Servicing presents another challenge, since astronauts cannot easily perform repairs on every satellite. The answer lies primarily in software: redundancy is engineered directly into the system so that if a memory flip or compute fault occurs, workloads can be rerouted around the affected hardware. Beyond that, the economics of AI argue against extensive servicing in any case. New AI accelerators arrive roughly every 18 months, so each generation of orbital data center is designed for a five- to seven-year operational life and then refreshed. At end of life, the satellites de-orbit and burn up in the atmosphere, the same disposal pattern already used by large constellations such as Starlink.
Why Move Compute Off the Planet
Two distinct forces are pushing compute into orbit. The first is purely earthly: some U.S. states are now outlawing or banning new data center construction, and even where projects are permitted, they face increasing scrutiny over their effects on local power grids, consumer electricity costs, and water consumption. Orbital infrastructure sidesteps these constraints by drawing on the free, continuous power of the sun and the free cooling of the vacuum of space, eliminating the need for terrestrial water and grid power from the moment a satellite reaches orbit.
The second force is intrinsic to space itself. As humanity expands its activities beyond Earth, there are applications that simply cannot rely on terrestrial data centers. Spacecraft, sensors, and surface operations on other bodies generate data that must be processed locally — particularly when the round-trip latency back to Earth is prohibitive. Lunar operations are the clearest example: the light-time delay from the Moon back to Earth makes round-trip cloud compute impractical for many real-time tasks, which means the build-out of a lunar economy will require compute infrastructure on or near the lunar surface itself.
This is not hypothetical. The first orbital data center in lunar orbit is already under contract, with a constellation of six satellites slated to provide compute services in support of activities such as the Artemis program and broader lunar resource development. Solar arrays are being supplied to power those six data center satellites, an early sign that the supply chain for off-world compute infrastructure is taking concrete shape.
Today's Cost Versus Tomorrow's Economics
Orbital compute is currently more expensive than its terrestrial equivalent, primarily because launch costs remain higher than the industry would like. But as launch prices continue to fall and as terrestrial constraints tighten, a growing list of use cases tips toward orbit. Workloads that demand sustainable power, applications subject to local political restrictions on new data centers, and any compute that natively serves a space-domain customer all begin to make economic sense even at today's launch prices.
The sustainability story is particularly compelling. By relying on solar generation and radiative cooling into space, orbital data centers avoid both the carbon footprint of grid electricity and the freshwater consumption of evaporative cooling systems. The trade-off is the launch energy required to put hardware in orbit in the first place — a one-time cost that is distributed across years of zero-water, zero-grid operation.
Security in Orbit
Security considerations also tilt favorably toward space. Recent geopolitical conflict has demonstrated that terrestrial data centers can be physical targets — they are stationary, identifiable, and relatively easy to strike. A satellite traveling at 17,000 miles per hour is a far harder target, and orbital infrastructure is inherently airgapped from ground networks, eliminating an entire category of attack surface. Cyber concerns remain real, of course, but they are addressed through standard enterprise-grade controls such as AES-256 encryption — the same algorithm that protects banking transactions.
A Rising Tide for the Space Economy
Public awareness of the broader space economy is also accelerating. The expected mega-IPO of the largest commercial launch provider this year is bringing mainstream investor and consumer attention to a sector that has historically been dominated by government programs. As one phrasing in the industry puts it, a rising tide raises all orbits — broader awareness of space as an investable, commercial domain helps every company building out infrastructure above the atmosphere.
Conclusion
Orbital data centers represent a convergence of several trends that have until recently been considered separately: the AI compute boom, the limits of terrestrial infrastructure, the maturation of the satellite supply chain, the falling cost of launch, and the early stages of a lunar economy. By packaging these into a marketplace experience that feels familiar to enterprise buyers — pick your rack, pick your bandwidth, sign a five-year agreement — the industry is positioning space-based compute not as exotic science fiction but as a practical extension of the data center business. The first satellites in this new architecture are already flying or under contract, and the question is no longer whether AI workloads will move into orbit, but how quickly the economics will pull more of them upward.