Eighty-five days after launch, LINK reentered Earth's atmosphere without touching the telescope it was built to save.
The spacecraft was Katalyst Space's attempt to rescue NASA's Neil Gehrels Swift Observatory, a 22-year-old satellite sinking faster than expected after solar storms thickened the upper atmosphere. NASA awarded the contract in September 2025. LINK launched on July 3, 2026.
The mission ended on September 25. NASA and Katalyst had already descoped the capture and boost in August, after an attitude-control fault left LINK without enough propellant to grab a satellite that was never designed to be serviced.
The rescue failed. The commercial case for robotic servicing did not.
Nine months from contract to orbit
Katalyst designed, built, tested, and launched LINK in nine months. The mission cost $30 million. That timeline is the part of the story the space industry keeps repeating, because most servicing concepts spend longer in studies than LINK spent in total.
The spacecraft was not a bespoke science platform. It was a prototype adapted for a customer with a deadline. Swift was expected to reenter by the end of 2026, which turned a technology demonstration into a race with the atmosphere.
LINK mission duration
Launched July 3, 2026 and reentered September 25, 2026. Katalyst Space, 2026
The fault that ate the mission
During commissioning, an electrical power system fault took two of LINK's three reaction wheels offline. The spacecraft entered a multi-axis spin. Communications became sporadic.
Engineers wrote new guidance, navigation, and control modes around the surviving hardware. They combined the reaction control thrusters with the one working wheel, then used the hall effect thrusters to reduce the body rate from 9 degrees per second to 1.47 degrees per second.
Electrical fault → two reaction wheels offline → multi-axis spin → partial recovery on thrusters → propellant burn too high for capture and boost.
The fix worked. The cost did not. Attitude control on the reaction control system burned propellant far faster than the planned configuration, and Katalyst and NASA concluded that LINK could not safely execute the capture and orbit-raising phase.
That left a spacecraft with fuel for something, just not the mission it launched with. The team pointed it at Swift anyway.
What 12 kilometers bought
LINK closed to within 12 to 15 kilometers of Swift. It collected unresolved imagery and exercised all three robotic arms, each about a meter long with a gripper at the end. The team also ran simulated capture operations and lifetime tests on the hardware.
That is not a rescue. It is a flight-qualified data set for the next attempt, and it arrived from a spacecraft that had already lost two reaction wheels.
“We knew from the beginning that this was an ambitious mission on an aggressive timeline. We also knew that there was a meaningful possibility that we would encounter challenges that could not be solved in time to achieve every ambition.”— Ghonhee Lee, CEO, Katalyst Space
Swift is expected to follow LINK into the atmosphere in early November. NASA had kept the observatory above roughly 300 kilometers with operational changes before the boost attempt, buying time for a mission that ultimately ran out of propellant instead of time.
A demo is not a business
As we wrote in early October, the servicing market already has a price anchor: a $310,000 seed round chasing a $30 million life-extension job. LINK just tested the expensive half of that equation in public.
The competitive field is no longer a set of isolated demos. Northrop Grumman's SpaceLogistics launched its Mission Robotic Vehicle with DARPA's RSGS payload in July, beginning a yearlong climb to geostationary orbit. Starfish Space reported autonomous proximity operations about 200 meters from another company's satellite in October, alongside a $52.5 million Space Development Agency contract for end-of-life disposal.
| Program | Backer | Status | What it proved |
|---|---|---|---|
| LINK | NASA / Katalyst | Reentered Sep 2026 | RPO to 12–15 km; arm and capture tests |
| MRV / RSGS | DARPA / Northrop | Launched Jul 2026 | First private operational GEO servicer en route |
| Otter Pup 2 | Starfish / SDA | Flying | Autonomous proximity at about 200 m |
Three programs, three different answers to the same question: who pays for a robot that extends the life of someone else's satellite? NASA paid for a rescue. DARPA paid for a capability. The Space Development Agency is paying for disposal.
Why propellant decides servicing missions
A servicing spacecraft spends most of its fuel before it touches anything. Rendezvous requires phasing, approach, and station-keeping. Every correction costs propellant, and the vehicle has to reserve enough to retreat if the approach goes wrong.
LINK's recovery changed that budget. The reaction control thrusters held attitude after two reaction wheels failed, but they did so at a much higher consumption rate than the design assumed. By the time the team stabilized the spacecraft, the remaining fuel could not cover both a safe approach and an orbit-raising burn.
NASA and Katalyst made the conservative call in August. Capture and boost came off the manifest. The spacecraft would spend its remaining life exercising systems and collecting proximity data instead.
The constraint is structural. Satellites in low Earth orbit were not built with standard docking interfaces, refueling ports, or cooperative targets. Swift was larger than LINK and never designed to be grabbed. A servicer has to arrive with enough autonomy to handle a target that is not helping.
That is why the September data matters. LINK showed that the guidance and control stack could be rewritten in flight, that the spacecraft could stabilize on degraded hardware, and that the robotic arms could still operate after the anomaly. None of those are capture. All of them reduce the unknowns in the next attempt.
Operating life matters too. LINK spent 85 days on orbit, most of it after the anomaly. That is short by satellite standards, but long enough to expose software and thermal behavior that ground tests miss. The flight data now feeds a platform with four missions to prove itself on.
Four missions, one platform
Katalyst raised a $12 million Series A on June 16, led by Geodesic Capital with Fortitude Ventures and other investors. The company says the round supports NEXUS, a standard robotic spacecraft platform meant to carry servicing, inspection, maneuvering, and deorbit work across multiple customers.
The follow-on pipeline is already visible. Katalyst has four missions and programs in development, including a Deorbit-as-a-Service award selected by the Defense Innovation Unit for the Space Development Agency, and the ASGARD power-beaming mission with the U.S. Naval Research Laboratory under OECIF. The company's first GEO mission is targeted for 2027.
Rendezvous and proximity operations, robotic manipulation, propulsion, and spacecraft control — the same building blocks, applied to different customers and orbits.
The economics only work if the platform is reusable across missions. A one-off rescue cannot amortize the engineering. A standard bus with interchangeable payloads can, at least in principle, spread the fixed cost across inspection, life extension, and disposal contracts.
LINK did not prove that model. It proved that a small team can field the hardware fast enough to matter, and that the hardware can survive its own failure long enough to produce useful flight data.
The test after LINK
The next milestone is not another dramatic rescue. It is a paid, repeatable service that a satellite operator buys without a government agency underwriting the risk.
A nine-month contract-to-orbit cycle is now a reference point, not an outlier.
Flight data from a partially failed spacecraft is still an asset for the next design.
The servicing market now has multiple funded vehicles, which makes the missing piece the customer contract, not the robot.
Katalyst lost the satellite. It kept the playbook, the data, and the next four missions. In a market that has spent years promising orbital mechanics as a service, that is a more useful outcome than a single saved telescope.