Who Repairs the Computers?

Part 13 of 30 · Series date: September 8, 2026

A robot with a wrench needs a reachable customer, the right spare part, and a repair worth making.

At a ground facility, a technician can pull a failed module, install a replacement, and carry the broken part to a workbench. The same task in orbit may require another spacecraft, a rendezvous, a robot, and a design that anticipated the repair years earlier.

Maintenance is where the romance of orbital industry meets the practical question of who brings the wrench.

Repair begins before launch

A repairable machine needs accessible connectors, predictable handling points, modular parts, and a safe way to isolate electrical and thermal systems. A robot cannot reliably replace a component buried behind equipment never intended to move.

The lesson applies on Earth too. Products designed for assembly are not always designed for disassembly. In orbit, the penalty for overlooking that difference is much larger.

NASA's in-space servicing, assembly, and manufacturing program treats refueling, repair, upgrades, and construction as related capabilities needed for more durable space infrastructure. These are development goals with different levels of maturity, not a single universally available service. NASA: ISAM.

A caution from an ambitious project

NASA discontinued OSAM-1 after technical, cost, and schedule challenges, along with changes in the market for servicing spacecraft not prepared for refueling. The cancellation is a reminder that a valuable goal can still be difficult to deliver economically. NASA: OSAM-1 status update, March 2024.

That does not prove orbital servicing is impossible. It argues for matching the repair method to hardware deliberately designed to accept it.

Three possible maintenance strategies

One approach is to replace the entire satellite. This may suit relatively small, standardized nodes if transportation and disposal are affordable. It avoids delicate component replacement but discards functioning equipment alongside failed or obsolete parts.

A second approach is to replace modules. Solar arrays, structures, and communications equipment might remain useful while processors are upgraded. The benefit depends on how cheaply a servicing mission can exchange and test the modules.

A third approach is to move equipment to a service platform or return it to Earth. That introduces transportation, handling, and potentially reentry requirements. It should be compared with simply launching a new unit.

These are architecture options, not a forecast that one will dominate.

Failure and obsolescence are different

A computer can still function while no longer earning enough to justify its place in the system. Newer processors may complete the same work using less energy or support applications the old hardware cannot run.

For a hypothetical platform, imagine arrays and structure that remain useful for ten years while the computing module becomes commercially unattractive after four. Replacing only that module could be valuable. But if the replacement requires a specialized spacecraft costing more than a complete new node, the apparent saving disappears.

The correct calculation includes compatibility, downtime, testing, and the risk that the repair fails.

Robots need proof, not choreography

A demonstration video may show a robot moving a part perfectly once. A commercial service must handle misalignment, unexpected resistance, degraded connectors, and incomplete information.

The robot also needs a safe fallback. If a connector will not release, can the system stop without leaving both vehicles endangered? If a coolant connection leaks, can the damaged section be isolated?

Robotic skill is only part of the answer. Standard interfaces and forgiving mechanical designs can reduce the skill required. The simplest successful repair may be better than the most impressive robot.

Diagnose before dispatching the robot

Imagine a computing module stops responding. Before launching a repair mission, the operator needs to know whether the fault lies in the module, its power supply, the network, or the software controlling it.

A robot carrying the wrong replacement can complete a flawless rendezvous and fix nothing. Remote diagnostics therefore deserve attention at the design stage. Useful measurements include temperatures, electrical behavior, error logs, and the response to controlled restart attempts.

The system should also preserve information around a failure. If every reset erases the only diagnostic record, the operator may repeatedly restore temporary function without learning why it disappears.

This resembles maintaining a vehicle far from a workshop. Knowing whether the problem is a flat tire or a failed fuel pump changes what help should be sent. In orbit, the cost of guessing wrong is much larger.

Reliable diagnosis can make some physical repairs unnecessary. It can also prevent repeated software restarts from being mistaken for a lasting fix when a component is steadily degrading.

A repair appointment has a route and a timetable

Two spacecraft described as being in low Earth orbit are not necessarily conveniently near each other. Their orbital planes, timing, and relative motion affect whether a servicing vehicle can reach the customer affordably.

A repair business therefore needs more than a capable robot. It needs a service region, enough customers in accessible locations, spare parts, and a transportation plan. A depot that serves one group of compatible spacecraft may be more useful than a nominally universal workshop that cannot reach most customers economically.

This is an architectural inference, not a claim about an existing depot market. It suggests that repairability can influence where future operators place hardware and which interfaces they adopt.

Maintenance begins to resemble a network industry. Compatible customers make a service route more valuable; an available service route makes compatible spacecraft more attractive. Building that relationship takes coordination as well as technology.

What makes a module worth saving?

Consider a deliberately simplified repair decision. Suppose restoring a platform would recover $3 million of expected future service value. A repair attempt costs $1 million and has a 70 percent chance of success. Ignoring other complications, its expected recovered value is $2.1 million, leaving $1.1 million after the attempt cost.

This is a teaching example, not investment advice or an estimate for a real mission. A serious decision would also include downtime, the risk of damaging other equipment, alternative replacement costs, and the uncertainty in future customer demand.

Now reduce the remaining service value to $1 million because the hardware is nearly obsolete. The same repair is much less attractive even though the robot's capabilities have not changed.

That is why technical repairability and economic repairability are different achievements. A part can be replaceable without being worth replacing.

The most useful maintenance strategy may combine approaches: repair expensive long-lived infrastructure, replace standardized computing modules, and retire units that no longer justify intervention. A mature industry would choose among those options with operating evidence instead of treating repair as either a universal solution or an unnecessary luxury.

Disposal is part of maintenance

An operator needs a plan for equipment that cannot be repaired. Abandonment can leave risk behind for other users. End-of-life propulsion, passive safety measures, and disposal costs belong in the original budget.

A service business should also explain what happens if the company itself fails. Hardware can outlast the organization that promised to manage it.

What would prove this?

Demonstrate repeated servicing on representative hardware, with measured cost, time, success rate, and recovery from abnormal conditions. Compare it with replacement rather than assuming repair is always greener or cheaper.

The future orbital workshop may be staffed by robots. Its economics will still depend on an old-fashioned principle: a repair is worthwhile when the useful life recovered is worth more than the resources spent recovering it.

Previous article · Next article