Long before the term “orbital data center” entered the common vocabulary, we had ‘Star Wars,’ which depicted a world in which droids quietly repair spacecraft, maintain systems, and keep missions moving. While the setting prioritized fantasy over physics, the underlying idea was practical: as infrastructure expands, so does the need for an always-on maintenance workforce.
Today, the concept of maintenance robots has moved from the pages of movie scripts to engineering roadmaps, specifically for new computing resources planned for space. Given the growing demand for AI computing, companies are beginning to explore orbital data centers, with players like Axiom Space already advancing plans for computing and data-processing capabilities in space.
As someone who has spent a career advancing robots in space, I am excited this dream is coming to fruition. However, I also find it odd that engineers’ are mostly focused on putting computing infrastructure into orbit while paying comparatively little attention to a far less glamorous question: What happens when something breaks?
Because it will.
As orbital infrastructure matures, operators will need more than launch vehicles, power systems and communications networks; they will also need an operation and maintenance layer. On Earth, that layer consists of skilled technicians who diagnose problems, replace hardware, and keep facilities running. In space, keeping a human maintenance worker in orbit quickly racks up risk and expense.
For data centers to succeed in orbit, humanoid maintenance robots will become as essential as server infrastructure, shifting from futuristic curiosity to practical necessity.
Difference Between Servicing and Maintenance
Whether it’s a hyperscale facility in Virginia or a server room in Singapore, data centers depend on far more than processors and storage. There are power systems, cooling systems, cables, connectors, pumps, batteries, and control electronics. Maintaining them requires an ongoing cycle of inspection, diagnosis, repair, and replacement.
The same will be true in orbit. In fact, space accelerates those challenges. Electronics have to endure radiation exposure. Materials will expand and contract cyclically between sunlight and darkness. Mechanical components will age faster. Thermal management systems will need to operate continuously in an unforgiving environment.
The failures themselves are unlikely to be dramatic. More often, operators will contend with degraded connectors, failed compute modules, eroding power-conditioning hardware, broken cooling-loop components, and communications equipment and mechanical systems that no longer perform as intended. These are relatively routine maintenance events in terrestrial data centers. In space, however, a buildup of these minor issues compounds into a major strategic challenge.
So why not send astronauts?
After all, humans remain the most adaptable maintenance systems ever built. They diagnose problems, improvise solutions, and handle surprises better than any machine. But astronauts are also expensive and scarce. As orbital infrastructure grows, sending humans every time a component needs attention becomes difficult to justify economically.
That is where robotics enter the picture.
The space industry has already demonstrated the value of robotic maintenance. The International Space Station’s Canadarm2 and Dextre systems have performed servicing tasks that once required astronaut spacewalks, reducing both risk and operational burden.
Yet there is an important distinction between servicing a known target and maintaining a complex facility over time.
Robotic servicing systems excel when the task is predefined. Engineers know which component needs replacing; the interfaces are standardized and the procedure has been carefully rehearsed.
Real maintenance is rarely so accommodating. Anyone who has worked around critical infrastructure knows that failures have a habit of creating unexpected problems. A latch sticks. A cable is slightly out of position. A connector does not engage cleanly. A component exhibits symptoms that do not match the manual.
Maintenance is often less about following instructions than adapting when reality refuses to cooperate.
Many of today’s orbital servicing systems were designed around known targets, standardized interfaces, and carefully scripted procedures. They excel at replacing designated components but are less suited for the steady stream of unexpected maintenance events that large operational facilities inevitably generate.
An orbital data center may require a robot not only to service hardware, but also to diagnose problems, adapt to unexpected conditions, and physically work around them.
This is where the rationale for investing in and deploying humanoids begins to make sense.
Years ago, NASA pursued a deceptively simple idea through the Robonaut program. The goal was not to build a robot that looked like a person. The goal was to build a robot that could work in an environment already designed for people.
I remember how powerful that idea felt.
Spacecraft, workstations, tools, handrails, controls, and procedures have all evolved around human operators. Rather than redesigning infrastructure for a robot, NASA explored whether a robot could learn to use the infrastructure humans had already built.
Robonaut2 ultimately became the first humanoid robot in space, demonstrating how a dexterous, human-compatible system could interact with tools, switches, cables, and other station interfaces.
For space data centers, Robonaut provides a glimpse of a viable diagnostics and maintenance solution beyond simple servicing.
Why the Humanoid Form Factor Matters For Space Data Centers
When people discuss humanoid robots today, they often focus on dexterity: Can the robot grasp objects? Can it manipulate tools? Can it perform tasks that traditionally require human hands?
Those are important questions, but they miss what I believe is the bigger advantage: compatibility.
The goal is not to make robots look human. The goal is to enable them to operate in environments built around human tools, human reach, human maintenance procedures, and human-designed interfaces.
A future orbital data center will likely be designed around maintenance concepts engineers already understand, such as access panels, replaceable modules, handheld tools, service corridors, and standard operating procedures. A humanoid robot can fit naturally into that ecosystem.
The same repair procedure developed on Earth can be used to train engineers, guide a robot in orbit, and support a future astronaut if human intervention becomes necessary. The same interfaces work for both people and machines.
This interchangeability matters.
Every specialized robot requires infrastructure to adapt to its unique capabilities. A humanoid, by contrast, can adapt to infrastructure that was originally designed around human work, and again when that infrastructure changes. If orbital data centers are a success, it is likely their internal systems will be periodically gutted and replaced with the latest compute hardware. Humanoid maintenance bots will easily be able to move through the new configuration and learn new routines.
This does not mean autonomous systems are unimportant; quite the opposite. They provide the monitoring that underlies all maintenance fixes.
