Habitats
Launch the Parts, Not the Ship
You cannot launch a habitat big enough to spin. So you launch a shipyard, and you let the robots build it.
The last piece made the argument that a long-duration ship has to spin, because bones and eyes and hearts do not hold up without weight. The obvious next question is how you build something that big when the biggest rocket we have is nine meters across.
The answer is that you do not build it on Earth. The first real step off this planet is not going somewhere. It is standing up a place in orbit where a ship gets assembled, and then going somewhere in it. We cannot just get up there and leave. We have to get up there and build.
What I saw on a tank
I have spent a lot of my career on jobs where somebody was putting up a big field-erected storage tank — fuel, water, produced water. The way it goes: the plates come in flat on a truck, get stood up in a ring, and then a little machine gets set on the top edge of the ring. It straddles the plate, one side inside the tank and one side outside, and rolls along the seam laying a weld the whole way around.1 When that course is done, the crane lifts it up and sets it on the next ring of plate, and it goes around again. One operator, one machine, x-ray quality seam, and nobody hanging off scaffolding.2
That machine is the ancestor of the thing I want built in orbit.
The placer
Here is the difference. The panels we have been publishing on this site — the interlocking triangles, STL files at the bottom of the page — do not get welded. They hook. Slide one into its neighbor, lay the next course on top, and the course above pins the course below so it can never back out. No heat, no filler wire, no weld head, no shielding gas. So the orbital machine is not a welder. It is a placer.
Picture the tank welder with a magazine on its back. The triangles are stacked in a cartridge like rounds in a rifle magazine. The robot grabs one, runs out to the edge of what has already been built, slides the panel down into its lock, comes back, grabs the next one. Ten thousand times. It rides on the structure it just finished — the hull is its own scaffolding — and it holds on by a groove cut into the panel geometry, not by magnets. I will come back to why not magnets.
Never send one
You never send one robot. You send one, it breaks down, and the whole job is dead in the water with nobody to drive out and look at it. Everything mechanical fails eventually. There is always something. So it is a crew — four, five, however many it takes to keep panels going down on separate seams at the same time — plus a mechanic. A robot whose job is to crawl around and fix the others where it can: swap a gripper, clear a jammed panel, tow one that quit back to the dock. That is not a nice-to-have. On a job site you cannot drive to, redundancy is the design.
Who is driving
Ideally, a person. Someone on the ground, a screen, a set of controls, and the robot as their hands. The boring part — grab, slide, seat, index over, repeat — will end up automated, because that is what automation is good at. What a person is for is the exception: the jam, the panel that seats wrong, the gripper that will not let go. A machine handles the situations it has already seen. A person handles the one it has not.
I will say plainly that this might turn out backwards. If the autonomy gets good enough to clear its own jams and diagnose its own failures, the yard will run better without anyone at the controls, and I would take that. So I do not want to bet the design on either answer. Build the machine so a human can drive it, and let the software take over as much of the job as it proves it can hold.
Either way, orbit is the one place this actually works, and it only works if the yard is in geosynchronous orbit. Signal to a geosynchronous satellite and back takes about 240 milliseconds — just under a quarter of a second.3 Workable. Try that at the Moon and it is closer to three seconds; at Mars, minutes. And a geosynchronous yard sits still in the sky, so the operator is never waiting for it to come back over the horizon mid-seam.
A note on lasers, because I asked the question myself. Laser links do not make the signal faster — light is light. What they buy you is bandwidth: more cameras, higher frame rate, less compression lag. For driving a machine by video, that matters a lot. It is just not a latency fix.
Why so high
Geosynchronous is a long way out — 35,786 kilometers — and there are two prices to pay for that. Every panel costs more to get there, and the yard sits outside the protection that low orbit gets from the Van Allen belts, so the electronics have to be hardened and any crew that ever goes up to look at it is taking more dose.
There are two reasons it is worth paying. The first is continuous control, above. The second is that low Earth orbit is filling up. Starlink alone has put more than seven thousand satellites up,4 and that is one operator. A shipyard is not a compact satellite. It is a structure growing toward hundreds of meters with loose panels and robots moving around it, and it cannot dodge a conjunction the way a Starlink can. One hit seeds a debris field through the whole yard. Better to go where the traffic is not.
Getting the panels there
You do not launch ten thousand loose triangles. You launch cartridges. And the way to launch a cartridge already exists: Starlink satellites are flat, they stack in the fairing without a dispenser, and they go up dozens at a time.4 A panel cartridge is a flat stack by nature. It rides the same way.
What it should not do is fly itself up. Starlink raises its own orbit on argon Hall thrusters,5 which is efficient but slow — months of spiraling to reach geosynchronous, and every day of it inside the radiation belts. For a shipyard, launch direct. Fewer trips, bigger cartridges.
What it is made of
Steel. Specifically stainless. Printed plastic is the prototype — it proves the geometry, it proves the lock — but plastic will not hold pressure and it will not stop radiation. SpaceX made the same call for Starship: they started on 301 stainless, moved to 304L for its toughness at cryogenic temperatures, and have since developed their own alloys in the 300 series.6 Musk’s argument for steel over carbon fiber was that it holds strength when hot, it is cheap, and you can form and weld it in a field instead of an autoclave.7 That is my argument too.
Which brings me back to magnets. The obvious way for a robot to hold onto a steel hull is magnetically. But 304 stainless is austenitic, and a magnet will not stick to it. You would have to go to a 400-series ferritic grade and give up some of what makes the 300 series good. So the robot holds on mechanically — a groove in the panel, a rail on the machine. Better anyway. No propellant to run out, no thrust plume, and it does not care what you print in.
Ten thousand garages
The panels stay small — ten or twelve inches on a side — and that is on purpose. A real orbital hull would rather be built from six-foot plates; fewer seams, fewer robot cycles. But a ten-inch triangle fits on a hobbyist 3D printer, and that means the manufacturing base is not one factory, it is ten thousand garages. The size follows the manufacturing model, not the structure.
Small panels mean a lot of seams, so the seams have to be very strong, and the hull is built in layers stacked several deep with the seams staggered so no joint in one course sits over a joint in the next. Same reason you cross a brick course, same reason plywood is plywood. A failed seam stays local. It cannot run.
If you have seen Andor, you have seen the picture: prisoners in separate compartments pressing together parts they do not understand, and only at the end do you find out they were building sections of the Death Star. Strip out the prison and that is distributed manufacturing. Thousands of people making one piece each, none of them seeing the whole.
This is our thought. Who can build it?
So here is where I am. The panels exist — you can print them tonight from the files below. The machine that places them does not. It has to do four things I can name and probably a dozen I cannot yet:
- Hold a triangle square while it slides it home, because a triangle only goes together from one side and only one way.
- Feel when the lock seats, rather than force it.
- Brace itself against the reaction, because in orbit, shoving a panel forward shoves the robot backward exactly as hard.
- Be drivable from the ground by a person — enough cameras, and enough bandwidth behind them, that an operator can work through a quarter-second of lag and still feel what the machine is doing.
I am going to buy a printer, run off a stack of panels, and build a crude first crawler and watch it fumble them. That will tell me where the design fights back. But somebody reading this has built pick-and-place machines, or welding carriages, or climbing robots, and knows things about this problem I have not thought of.
This is an open proposal. Knock it down, or build it.
From one explorer to another, we appreciate you reading this far. We are excited about what this endeavor brings, and we would like you along for the rest of it.
Come join us. Because we are Star Bound.
John LangEditor
Ideas in Star Bound are published open — take them, build them better. Our open proposal.
A note on where this comes from. I am not a metallurgist, a spacecraft engineer, or a degreed engineer of any kind. But I am a practical one, with more than forty years in heavy industrial construction: welding, millwright work, hydraulics, mining, and thirty-three years of electrical construction and commissioning on the North Slope of Alaska, in a place that punishes anything built wrong. Much of that time I have worked as a field engineer, which means I am the one standing in front of the equipment when the drawing turns out to be wrong, and the fix has to be made on the spot, with what is on hand, and has to work the first time. Anyone who has done that long enough learns that a lot of engineering gets corrected in the field by people with more experience than the people who drew it. That is the eye I bring to these subjects: not what should work on paper, but what actually gets built, and what breaks. The facts I take from other people are footnoted so you can check them and read further. The ideas built on top of those facts are my own, and I would rather see them knocked down by somebody who knows better than left standing because nobody looked.
Sources
- Koike, "Automatic Girth Welders (AGW) for Storage Tank Shells" — the machine straddles the shell plates on flanged wheels and welds the horizontal seam. https://www.koike.com/positioning/tank-fabrication/agw/ ↩
- Bug-O Systems, "Girth Welder Machines for Tank Fabrication" — self-propelled units that run directly on the tank and eliminate scaffolding. https://bugo.com/girth-welders/ ↩
- Satsig, "Geostationary satellite latency and time delay" — 240 ms for the up-and-down hop at 35,786 km, up to about 280 ms at the edge of coverage. https://www.satsig.net/latency.htm ↩
- eoPortal, "Starlink Satellite Constellation" — flat-panel design, stacked for launch without a dispenser, more than 7,000 satellites deployed. https://www.eoportal.org/satellite-missions/starlink ↩
- SpaceNews, "SpaceX launches first upgraded Starlink satellites" (Feb. 28, 2023) — V2 Mini satellites use argon Hall-effect thrusters in place of krypton. https://spacenews.com/spacex-launches-first-upgraded-starlink-satellites/ ↩
- Wikipedia, "SAE 304 stainless steel" — early Starships used 301, moving to 304L in 2020; 304 strengthens by cold work. https://en.wikipedia.org/wiki/SAE_304_stainless_steel — and Basenor, "SpaceX Has Moved Beyond 301 Steel" on the proprietary 30X-series alloy. https://www.basenor.com/blogs/news/spacex-has-moved-beyond-301-steel-starship-now-flies-on-proprietary-alloys ↩
- IMDEA Materials, "Why Elon Musk uses stainless steel for his new spacecraft" — stainless versus carbon fiber and aluminum-lithium; cold-forming and field weldability. https://materials.imdea.org/why-elon-musk-uses-stainless-steel-for-his-new-spacecraft/ ↩
More from Star Bound
- Top of the Loop, Bottom of the LoopThe two animals nobody budgets for, and what happened the last time we forgot them
- Light Is the ThrottleWhat a mine in Idaho and a garden in Alaska taught us about growing things fast — and what continuous daylight does to a plant that never gets to sleep
- Commission It FirstWe have already run phase one. It cost one man 150 million dollars of his own money, and it very nearly killed the people inside.
- Life in the BarrelO’Neill gave us the barrel in 1976 and got very little of it wrong. What he did not finish was the inside — and the standard the inside has to meet isn’t twenty years. It’s forever.
- Worms: The Backbone of CivilizationWhy the first Martians will need a bucket of worms more than another rocket
- Send the Robots FirstWhy the road to Mars runs through the moon, and why the first lunar miners will never leave Earth
- Wet the Ground and WaitA passive water generator for the desert, and why we should build one before we need one on Mars
- Send the Refinery, Not the OreWhy the belt gets mined by factory ships, and why that is the one job we cannot run from a chair
- Grind the Asteroid, Print the PartWhy there will never be a steel mill in space, and what replaces it
- A Thousand GaragesBefore there is a printer in space, there has to be a shipyard on the ground, and the shipyard does not have to be in one place
- Powder or WireThe two ways to print metal, why neither one is ready for the Belt, and the questions somebody has to answer first
- Bones Don't Travel WellThe one problem no rocket solves, and why Gerard O'Neill worked out the answer in 1976
- Blow the Rock Like GlassMy father's answer to how you build something big enough to spin, and the man who worked it out in 1963
- Built for Earth, or Built for Something Bigger?The tools a colony needs are being perfected right now, by people who say they are building them for us
- Mission Watch: September 2026What private enterprise and the agencies actually did this summer, in the order that matters
- The Question I Asked Buzz AldrinIn 1996 I asked the second man on the Moon what part private enterprise would play in space. Thirty years later, the answer is checkable
- Fish, Salt, or DirtThree ways to feed a habitat, what each one costs you, what each one can't survive, and a question for the reader.
- Thirty Thousand Volts and a Piece of Tin FoilI drove an hour and a half to call a man a liar. He turned on a computer monitor instead, and I have been thinking about it for twenty years.
- A Message in a Bottle, Thrown HardEvery article in this magazine is about how to build something. This one is about how to pay for it.
- The Box Nobody TestedIn 1991 eight people were sealed inside three acres of glass to prove a closed world could keep them alive. It nearly didn't, and the reason was sitting in the walls the whole time.
- Sixteen Inches of SteelThe number that breaks the mile-wide barrel, four ways around it, the one we would bet on, and a file you can print tonight
- That’s Not a Rocket, It’s a Freight TrainHow I wired the one on my bench in 1994, the one I drew and never built, and why a mass driver runs on railroad logic, not rocket logic.
- Out the Side of the MountainTwenty miles of vacuum tube under a Hawaiian volcano, and a freight train that never comes back down. What a ground-based mass driver would actually take, and why it exits the flank instead of the summit.
- Cold, Vacuum, and a Hundred Thousand JointsWhat geosynchronous orbit does to the ten-inch panels in "A Thousand Garages," starting with the question every steel man asks first — and one thing that article got wrong.
- One Press or Ten ThousandElon Musk casts a car underbody in one shot with a machine the size of a house. We proposed ten thousand garages. It turns out we may both be right, on different walls — but nobody knows until somebody breaks fifty panels on purpose.
- Cast It or Print ItThree ways to make a hundred thousand habitat panels. What each one buys you, what each one costs you, and where each one belongs. The first entry in a department that exists to write both columns down.
- Powder or PlateThe garages stay. What's in them changes.
- Refuel It or Strip ItThe graveyard belt is the closest junkyard we've got.
Comments
Star Bound is not a forum, so comments run the old way: if you have comments, please write to us at hello@starboundmagazine.com and they will be published here in the next update, with your name on them. Say if you would rather stay anonymous.
No comments posted on this one yet. Yours could be the first.