Why a shop
The Shop Buys Time, Not Mass
Carrying raw stock instead of finished spares does not lighten the manifest past the Moon. The shop is justified by availability, iteration and self-expansion.
An off-Earth machine shop is a 3.09 t facility that can make or repair 69% of what breaks at an outpost. Its first justification was mass: one rack of bar stock serves hundreds of part types, cutting safety stock by the square root of the part count - 16.6 times at 276 types. But machining from bar wastes about 44% of each billet, and that loss outweighs the pooling gain everywhere past the Moon. The mass argument for the shop was withdrawn. Modeled
Three arguments hold. Availability: a spares list is a bet on a specific list, and a shop is a hedge against the list being wrong; at Mars it cuts function-months lost to unmanifested failures from 1,248 a year to 389. Iteration: a design change at Mars takes 790 days through Earth and one to three days through a local shop. Self-expansion: once the site can cast machine bases, each new machine ships as 17% of its terrestrial mass. The first years of an outpost are a research program in an operations costume, and a site that cannot modify its own hardware can only report problems and wait. Modeled
| Function | What it demands | Sizing driver | Rated |
|---|---|---|---|
| Repair | Diagnose, refurbish, regrind, re-thread, splice a seal - hours to days | Peak response; sets metrology and fixturing | Modeled |
| Manufacture | Make a replacement from stock when no spare exists - days | Breadth of material stock | Modeled |
| Prototype | Make, iterate and qualify something new - days to weeks | Design-to-part cycle time; sets the CAM chain and compute | Modeled |
Demand
The Shop Covers 69% of Failures; Bearings and Electronics Always Ship
Failure shares are engineering estimates for a mechanically dense site, not measured data.
| Failure class | Share | Shop? | Rated |
|---|---|---|---|
| Seals, gaskets, O-rings | 22% | Yes - from cord stock and a vulcanizing kit; the elastomer is imported | Modeled |
| Wear parts | 15% | Yes | Modeled |
| Electronics, sensors, windings | 14% | No - ships finished | Modeled |
| Rolling-element bearings | 13% | No - hardened and ground to micrometers | Modeled |
| Structural: brackets, mounts, guards | 10% | Yes | Modeled |
| Fasteners | 9% | Yes | Modeled |
| Hose, tube, fittings | 9% | Yes | Modeled |
| Springs and clips | 4% | Yes | Modeled |
| Optics and unclassified | 4% | No | Modeled |
With 120 part types under a commonality program, the site sees about 240 failures a year and 166 the shop can address - three jobs a week. A facility idle 80-90% of the time that must still respond within hours cannot justify a dedicated person, which makes robotic operation the default. Eight kilograms of O-ring cord and a vulcanizer replace several hundred seal part numbers.
By destination
Farther Out, the Shop Saves Time and Costs Mass
Spares mass favors the shop only at the Moon; availability and the design loop favor it everywhere.
| Destination | Resupply lead time | Spares mass saved per cycle | Function-months lost per year, without / with shop | Design loop, Earth / local | Rated |
|---|---|---|---|---|---|
| The Moon | 3 months | +0.11 t, 7.5-yr payback | 144 / 47 | 90 d / 1-3 d | Modeled |
| Mars | 26 months | -0.07 t | 1,248 / 389 | 790 d / 1-3 d | Modeled |
| Ceres | 30.8 months | -0.13 t | 1,478 / 461 | 938 d / 1-3 d | Modeled |
| Callisto | 3.82 years | -0.34 t | 2,198 / 684 | 1,395 d / 1-3 d | Modeled |
Near-net-shape processes - casting or wire deposition at about 15% waste - flip the mass result positive, 0.42 t a cycle at Mars and 0.51 t at Callisto, but still take 16-24 years to pay back. Recycling swarf into the wire and casting stream has the same effect without changing the cutting process.
The MVP
Three Tiers, 3.09 Tons to the First Working Shop
Each tier is useful on its own and is justified by a different function.
| Tier | Mass | What it adds | Rated |
|---|---|---|---|
| Tier 0 - the bench | 348 kg | Hand tools, TIG welder, polymer printer, grinder, bandsaw, metrology, O-ring cord kit; about 35% of failures | Modeled |
| Tier 1 - the MVP shop | 2,745 kg | CNC mill-turn (900 kg), 400 kg stock, fixturing, sealed swarf handling, wire DED, press, oven, sheet metal, surface plate, laser interferometer, part library; 69% of failures | Modeled |
| Tier 0 + Tier 1 | 3.09 t | About 55% above the 2.0 t first carried for the whole fabrication line | Modeled |
| Tier 2 - production | ~1.5 t shipped for ~4 t of capability | Wire EDM, surface and cylindrical grinding, casting, 5-axis machining on locally cast bases | Modeled |
| Composite bay | ~1.05 t | Extruder, platen press, thermoplastic welder, winder, oven - see Manufacturing | Modeled |
Physics
Below Mars, the Machine Walks Across the Floor
Five problems Earth shops never meet: hold-down, swarf, coolant, fire and calibration.
| Machine | Hold-down, Earth | Mars | Moon | Ceres | Rated |
|---|---|---|---|---|---|
| Benchtop mill, 600 kg | 2.35 kN | 0.89 kN | 0.39 kN | 0.07 kN | Modeled |
| Small machining center, 1 t | 3.92 kN | 1.48 kN | 0.65 kN | 0.11 kN | Modeled |
| Production mill, 3 t | 11.77 kN | 4.45 kN | 1.94 kN | 0.34 kN | Modeled |
A 12 mm end mill in aluminum develops 400-900 N; a roughing cutter 1,500-3,000 N. A 1 t mill on the Moon has 0.65 kN of friction, so every machine below Mars is bolted to a poured, cured, surveyed foundation - the same anchoring problem the excavator faces on Ship or Make. Swarf drifts and clings at 0.16 g, conducts, and in titanium and magnesium can ignite, so cutting is fully enclosed with suction and electrostatic capture and a briquetter compacts the chips into feedstock.
| Problem | Earth default | Off-Earth choice | Rated |
|---|---|---|---|
| Metal additive | Laser powder-bed fusion | Wire-fed directed energy deposition - no loose powder, one wire for welding and printing | Modeled |
| Coolant | Flood coolant | Cryogenic CO₂ on Mars from the atmosphere; dry machining and minimum-quantity lubrication on the Moon | Modeled |
| Cutting tools | Carbide, discarded | High-speed steel, resharpened 10 times - tungsten import falls from 1.2 to 0.33 kg a year | Modeled |
| Bearings | Rolling element | Bronze or PTFE plain bearings where duty allows; 20 minutes on a lathe | Modeled |
| Calibration | Traceable gauge chain | Laser interferometer for length; three-plate lapping for flatness | Modeled |
| Fire | Air atmosphere and sprinklers | Hypoxic or inert atmosphere while robots run it - a third pressure zone | Modeled |
| Machine base | Shipped cast iron | Cast basalt made on site; the machine ships as 17% of its mass | Modeled |
Power and space
6.7 kW Average, 31.6 kW at Peak, One Habitat Module of Volume
The peak matters more than the average because the shop's value is response time.
| Item | Figure | Note | Rated |
|---|---|---|---|
| Installed power | 31.6 kW | 21% of a 150 kWe site if everything runs at once | Modeled |
| Average power | 6.7 kW | 4.5% of site power; load-shed above the robot charger, below life support | Modeled |
| Radiator | ~13.4 m² at 320 K | Its own loop; the shop is a concentrated heat load | Modeled |
| Floor and volume | 40-55 m² · 100-140 m³ | About one habitat module | Modeled |
| Workload | 3.2 jobs a week | 12-33% of one operator; sized for peak response | Modeled |
| Swarf | 324-972 kg a year | 800-2,400 liters loose; larger than the 400 kg of stock consumed | Modeled |
Self-expansion
A Machine Tool Ships as 17% of Its Mass Once the Base Is Cast Locally
Machine-tool mass is mostly cast structure for stiffness and damping, and weight does no hold-down work off Earth anyway.
About 73% of a machine tool is cast structure and 7% sheet metal. The part that must ship - spindle and bearings, controller, ways, ball screws, encoders and servos - is about 17%. A 900 kg mill becomes a 153 kg shipment assembled onto a locally cast base, and a four-machine expansion ships as 0.53 t instead of 3.10 t. Cast basalt is a mature terrestrial material with roughly cast-iron compressive strength and better damping; a base works in compression, where basalt is strong. The first machine still ships complete, because nothing on site can finish a casting until it exists. Modeled
The shop pays back its own expansion, not the spares inventory. Closure is about 80% by mass, low by part count, very low by value and zero by capability: the shop is a mass amplifier, not a self-replicator.
Staging
From a Bench in the Garage to a Regional Shop
The shop grows with the site's foundations, power and casting capability.
| Phase | Tier | Prerequisite | Rated |
|---|---|---|---|
| Phase 0 - first cargo, uncrewed | Tier 0 bench, in the garage | None | Modeled |
| Phase 1 - site established | Tier 1 MVP shop | Poured and surveyed foundation, third pressure zone, 32 kW peak, 13 m² radiator | Modeled |
| Phase 2 - crewed, casting online | Tier 2 production | Induction melt and mold preparation | Modeled |
| Phase 3 - industrial | Regional shop, locally built | A metals supply chain; machine-tool assembly becomes a local product | Modeled |
The shop sits beside the garage - big door - not the habitat, and runs its own dirtier-than-habitat, cleaner-than-garage dust regime. The validated part library of CAM programs is the most valuable item in it and weighs nothing. Hard automation problems are fixturing and diagnosis, not cutting; designed datum and clamping features on every serviceable part are the fix.
Flown so far
Polymer and Metal Printing Have Flown; No Machine Tool Has
Every fabrication process flown in space so far is additive; no subtractive machine has cut metal off Earth.
| Hardware · organization | Process | Status | Rated |
|---|---|---|---|
| Zero-G printer · Made In Space, NASA | Polymer printing on the ISS | First part printed in space, November 2014 | Observed |
| Additive Manufacturing Facility · Redwire | Commercial polymer printing on the ISS | Operating since 2016 | Observed |
| Metal 3D printer · ESA, Airbus | Laser-melted stainless wire | First metal part on the ISS, August 2024 | Observed |
| Continuous-fiber printer · CAST, China | Carbon-fiber composite | Two parts printed on the Long March 5B test flight, May 2020 | Observed |
| Lihong-1 · CAS Space and CAS Institute of Mechanics, China | Metal printing in microgravity | Suborbital flight, 12 January 2026 | Observed |
| Containerless furnace · Tiangong, China | Refractory alloy melting | Above 3,100 °C, reported August 2025 | Observed |
Contested
Deliberately Not Claimed
The load-bearing uncertainties, tagged with the confidence they actually earn.
The failure population is an estimate, not data. Nearly every number on this page scales off it. Maintenance records from Antarctic stations, offshore platforms, submarines and mine sites would turn it into evidence with no hardware at all.
Cast basalt is unproven as machine structure. A university-scale test - cast a base, mount commercial ways and a spindle, measure stiffness, damping and thermal stability against cast iron - would validate the largest finding.
Swarf behavior in low gravity is unmeasured. No parabolic-flight or orbital test has handled machining chips at partial gravity.
Robotic fixturing of broken parts is unsolved. It is a terrestrial robotics problem, and nobody has addressed it for an outpost.
Record
Provenance and Data Sources
- Confidence
- Observed for flown, fielded or measured - Building for hardware and programs in development - Projected for announced plans - Modeled for original analysis
- Status date
- Q4 2026
- Analysis
- The Machine Shop (The Primitives, Annex B, Rev A, July 2026); The Composite-Structure Question (Annex C); Closure Sequencing
- Flight record
- NASA and Made In Space (2014); Redwire; ESA and Airbus (2024); CAST (2020); CAS Space (2026); Tiangong reports
- Weakest link
- The failure-population estimate and the untested cast-basalt base