On-orbit servicing arms: $310K seed against a $30M price tag for a life extension

The arms were folded against the flanks of the servicing vehicle when it left Cape Canaveral on 21 July 2026. Two of them, each about three metres long, each built around seven rotational degrees of freedom. Over the following year they will coast outward to geosynchronous orbit, GEO, before they grip anything at all. When they do, they will perform the first privately owned, operational robotic servicing in that orbit, a description that reads like paperwork until you price the manoeuvre. Sold to a commercial operator as a life extension, that same job has been discussed at roughly $30 million.

Eleven weeks after that launch, a company of about six people in Puyallup, Washington, is trying to sell the arm that performs it as a standard component. Orbital Robotics has raised $310,000 in total. It has flown six demonstrations of AI-directed robotic arms through NASA and U.S. universities, and it lists Sophia Space among its partners. It has announced no customer, no contract and no revenue. The distance between a $310K raise and a $30M quote is where the supply layer of orbital servicing actually sits in October 2026.

Two arms, one Falcon 9, and a $30M question

The Mission Robotic Vehicle rode to orbit on a Falcon 9 from pad SLC-40 at Cape Canaveral as part of DARPA's Robotic Servicing of Geosynchronous Satellites programme, run as a public-private partnership with SpaceLogistics, the Northrop Grumman subsidiary that builds and flies the Geosynchronous Servicing Spacecraft family. The robotic payload was developed with the U.S. Naval Research Laboratory, USNRL: two 3-metre, 7-DOF arms and more than 20 cameras for situational awareness.

Two ownership details matter more than the hardware. SpaceLogistics owns and operates the integrated spacecraft. It holds the rights to use the robotic payload commercially, including installing company-funded Mission Extension Pods, MEPs, that dock with a client spacecraft and extend its life by up to six years. A 2024 agreement with NASA brought civil spaceflight and robotics expertise into the same program.

So the vehicle that launched on 21 July 2026 already carried a revenue path, and that path does not run through an arm manufacturer. It runs through whoever owns the vehicle and the pod.

The arm layer sits underneath that. It holds the least differentiated position in a stack that now has a mission in orbit, two government programs selecting suppliers, and primes buying robotics subsidiaries. It has also never been easier to be one of four suppliers bidding for the same buyer.

Four years, six flights, $310K raised

Orbital Robotics was founded in late 2024 by people who built Blue Origin's BE-7 lunar lander engine, New Glenn, Orbital Reef and Blue Ring. The wider team includes crews with more than 100 shuttle missions behind them. Aaron Borger is chief executive, Doug Kohl is chief operating officer, and Riley Mark and Sohil Pokharna came out of the same lineage. Caplight lists a headcount band of five. The company describes itself as Seattle-based, and its registered footprint sits in Puyallup.

The money is small and precisely dated. Caplight records a seed round closed on 1 November 2025, with total funding raised of $310,000. GeekWire reported the same figure in January 2026, noting Plug & Play as a backer alongside the AI House programme.

The catalogue is short: the ORA-Astrosfera manipulator, AstroBot, ASTRA-P, NavIQ for guidance, navigation and control, and ORBtos as the flight software. Orbital Robotics claims six flight demonstrations of AI-controlled arms through NASA and U.S. universities, and describes its current state as a technology readiness level 4, TRL-4, patent-pending system that fuses vision-based navigation with neural network planning and control.

Positioning runs hot for a company this size. It appeared on the TechCrunch Disrupt space stage in 2025, was showcased by The Aerospace Corporation, and carries a Black Flag 100 listing for 2026. The company reports more than ten industry partners. One Orbital Robotics illustration shows arms reaching for the Hubble Space Telescope. That is a concept image. No contract attaches to it yet, and the difference matters for everything that follows.

Why the arm is the bottleneck

Capture is hard in one specific way. A servicing arm does not approach a stationary object. It approaches a vehicle that is itself coasting, tumbling and reacting the instant the arm touches it. Orbital Robotics calls that condition dynamic coupling: arm and target both in motion. Every contact force the arm applies becomes an input to the target's trajectory, and the arm has to close that loop without a human in it.

Our approach is more you take position, orientation, velocity, things like that, and then do very small actions. Each neural network is trained in maybe attitude control or translation control and we can verify each of those pieces.— Aaron Borger, chief executive, Orbital Robotics, on the TechCrunch Disrupt space stage

The quote describes an architecture built around verification. Each neural network is trained on a narrow slice of behaviour, attitude control or translation control, and each slice can be checked before the pieces are composed. That decomposition is what turns a demonstration into something a customer can put an acceptance test against.

Compare it with the incumbent path. The MRV-1 payload uses two USNRL-developed arms and more than 20 cameras: heritage components, decades of qualification, and a government customer able to absorb integration risk. Orbital Composites has pushed robotic additive manufacturing for in-space assembly and manufacturing through Air Force Research Laboratory Catalyst and Small Business Innovative Research, SBIR, awards, another arm-adjacent supplier building the same verification trail.

That is the bar six flight demonstrations cleared and the bar the contract has not. Judges at The Aerospace Corporation flagged precisely this combination — hardware and software that must work together — as the hard part of the pitch. Borger said the company was pushing as much as possible into simulation and terrestrial test beds while lining up launches for the following summer and autumn.

None of this makes the arm category unattractive. It makes the revenue event narrow. A servicing contract is never won on a datasheet. It is won when a vehicle operator accepts that the arm's behaviour is flight-critical, and that acceptance arrives through a program, not a demo.

Two programs picked different orbits

Two government programs now sit on opposite sides of that acceptance problem, and they chose different orbits. MRV-1 works in GEO, where the customer is a communications operator deciding whether to spend $30 million on more life. NASA's Fly Foundational Robots mission, FFR, works in low Earth orbit, LEO. It launches in late 2027 and operates a commercial arm supplied and flown by Motiv Space Systems, now Rocket Lab Robotics after Rocket Lab closed that acquisition in May 2026.

DARPA (2026) and NASA Goddard / NASA Space Technology Mission Directorate in-space assembly and manufacturing portfolio (2026).

FFR's arm carries a harder brief than a refuelling visit: dexterous manipulation, autonomous tool use, and walking across spacecraft structures in zero or partial gravity. The payload installs an orbital replacement unit, an ORU, to extend or enhance the spacecraft's function, and a second payload called A-STEP is manifested alongside it. Named partners include Astro Digital, Arkysis, SpaceWorks, iBoss and Picknik.

The orbit choice pays more per job and arrives slower. A year of transit happens before the arms do useful work. The LEO option buys frequency instead: repeat visits, cheaper transport, a customer who is already building hardware to be serviced. An arm supplier must pick a vehicle architecture, a thermal environment and a buyer, and these two programs reward opposite answers.

The layer is not empty elsewhere. MDA Space, the former SSL, holds a $20.7M DARPA contract for two RSGS robotic arm systems, per SpaceNews. A flexible robotic arm was demonstrated on the Xiyuan-0 platform in April 2026, and the Shijian line has run GEO refuelling and mission-extension tests since late 2025. Multiple programmes, one shared bottleneck: the party that can prove the capture works owns the conversation.

Where the $30M comes from

We've already talked to satellite operators and we know they need it. Our GEO customers have indicated they would pay thirty million dollars for a life extension, and LEO customers want to be able to maneuver without regret.— Aaron Borger, chief executive, Orbital Robotics, recapped by The Aerospace Corporation

Read the wording closely. Indicated willingness to pay is not a purchase order, and a number spoken on a conference stage is not a booked contract. The defensible version of this story is that a customer-shaped price exists for GEO life extension, and that no public transaction at that level has been announced.

The annualisation is what makes the number plausible. Spread across the six years a Mission Extension Pod can add, $30 million is $5 million a year, charged against the alternative of retiring the asset and buying a replacement. The appeal is an operating expense on a mission the operator already owns.

Whether the money comes out of the operator's budget or an insurer's is the open question, and it is not hypothetical. As we wrote on 2 October, Charter Space raised $5M on a finding that 97% of operational spacecraft carried no insurance at all. A fleet that cannot be serviced is a fleet an insurer prices as a probable total loss. A fleet that can be serviced is a different asset class, and the two prices are one negotiation viewed from opposite ends.

The gap between TRL-4 and a contract

ORBITAL ROBOTICS ORA-ASTROSFERA — TRL 4/9
─────────────────────────────────────────────────────────────
  TRL 1–3        TRL 4–6        TRL 7–8        TRL 9
  🔬        ────  🧪        ────  ◉ NOW    ────  ✅
  Research       Pilot          Scale-up       Production
─────────────────────────────────────────────────────────────
[████░░░░░░] 40%  ·  Flight demos flown; no operational service yet.
Source: Orbital Robotics company materials, 2026; TRL scale per NASA/ESA convention

TRL-4 in practice means flight demonstrations without an operational service behind them. Orbital Robotics has cleared the flight half of that bar six times through NASA and university programs, and describes the underlying system as patent-pending at this level.

The commercial half is thinner. GeekWire reported in January 2026 that the company was working with a stealth venture on an orbital rendezvous project for the U.S. Space Force, with a series of missions scheduled across the following year and a half. Borger's framing of that schedule was blunt: the longer you wait, the more difficult it is. It is also a line about a partner's program rather than a disclosed contract of Orbital Robotics' own.

The same shape showed up in a neighbouring layer two days ago. On 3 October we covered Foundational, which raised £8.2M to turn orbit laser ranging into a service. Ranging is a measurement, an arm is a manipulation, and both are infrastructure sold as a recurring service against hardware already in orbit. Both hit the same wall: the invoice only clears after the vehicle's owner signs.

Assembly in orbit as the second market

There is a second market for these arms, and it sits closer to manufacturing than to servicing. Large structures assembled in orbit need capture, docking and material handling for the same reason a GEO operator needs refuelling. Orbital Robotics lists in-space assembly as a capability and names Sophia Space as a partner.

Sophia Space is building TILE, an orbital compute platform, with a memorandum signed with Redwire at the end of September 2026 and a demonstration targeted for 2027. That timeline is the whole argument for the arm layer's second act. Assembly programs need tooling years ahead of their first flight, and tooling is the one purchase a small supplier can win on engineering rather than on balance sheet.

Then the counter-case. Rocket Lab bought Motiv Space Systems in May 2026 and renamed it Rocket Lab Robotics, putting a Perseverance and CADRE heritage shop inside a launch prime. SpaceLogistics holds the commercial rights to the servicing payload on MRV-1. Northrop Grumman flies the vehicle. Each of those is a place where arm value gets absorbed internally rather than passed down to a supplier.

The arm maker is the thinnest layer in a stack where every other layer has consolidated around fewer names. That is not an argument against the hardware. It is an argument about which orders would have to land first, and who would be holding the vehicle when they do.

What to watch by mid-2027

Two dates bracket the next evidence. NASA's FFR launch in late 2027 is the first scheduled flight of a commercial-supplier arm, operated in orbit by the company that built it. The 18-month window Borger described runs into the second half of 2027 as well, so both land within weeks of each other.

A third signal costs nothing to watch. If SpaceLogistics books a customer-funded Mission Extension Pod instead of funding one itself, that single line item would tell you whether $30 million is a market price or a conference answer.

Where the value lands

The servicing layer stopped being a demonstration in July 2026. What has not changed is the distance between the number a customer says out loud and the number a supplier books.

Orbital Robotics has built a credible arm, flown it six times, and raised $310,000. Its nearest neighbours fly on government money, and one of them now sits inside a launch prime. The $30 million is the market's own estimate of what a GEO life extension is worth. Between the two numbers there is exactly one thing worth looking for: a disclosed contract.

It will show up in the boring places. A round above the seed. A flight under the Space Force rendezvous program. A customer-funded pod on a vehicle somebody else flies. Until one of those appears, the arm layer stays the smallest part of the largest opening in orbital operations.

Sources