Ship or Make

The opening argument

Every Load-Bearing Edge Runs Out of Earthmoving

Two independent derivations end at the same machine: a low-gravity excavator nobody has a program to build.

Ship or make is the decision every site faces for every item: launch it from Earth or produce it locally. Closure - making enough locally that the site no longer needs resupply - looks like a long chain of dependencies. Tested edge by edge, most of the chain is not a chain. The orderings that hold all run back to one capability: moving regolith. Shielding a habitat, building a landing pad, anchoring a machine-tool base and static-firing a return vehicle all wait on earthmoving, and the analysis of hardware primitives found the same machine on its own critical path. Modeled

Closure is replacement, not construction. A site arrives complete: at 100 t per landing, a 232.5 t minimum viable outpost is 2.3 cargo Starships. It then decays, shedding about 4.6 t of hardware a year - 2% of itself, across roughly 240 failures averaging 19 kg - while about 12 t of consumables run out completely. Upper rungs serve the decaying stream and lower rungs the depleting one, and only the depleting stream is urgent. Modeled

Local production comes in two kinds. A displacement capability makes something that would otherwise be shipped: oxygen, water, spare parts, structure. An enabling capability makes something nobody would ship at all: a berm, a road, a foundation, a buried habitat, a landing pad. Payback arithmetic applies only to the first. A grader does not repay the mass of the road it grades, because nobody was going to launch the road; its value is everything the road makes possible.

The machine

Low Gravity Takes Away Force, Not Energy

The earthmoving set is the second-largest line at the lunar outpost and the least specified object in the analysis.

A digging machine develops breakout force by reacting it through its own weight. Regolith resists a cutting edge through cohesion, friction and the interlocking of unweathered grains, and none of those change with gravity; the force the machine can apply does. A 5 t machine weighs 49 kN on Earth and 8.1 kN on the Moon. The machine does not lose power off Earth - it loses grip. The 1.67 m/s² threshold that forces machine tools onto poured foundations is the same physics one step earlier: a lunar mill has 0.65 kN of friction against 0.4-3.0 kN of cutting force, and an excavator has 5 kN of drawbar against a cut that wants more. The Moon sits 3% under the line; Mars, at 3.71 m/s², is the only body comfortably above it. Modeled

There are two escapes. A machine that fills a ballast bin from its own first bucket doubles its working weight using mass already on site, and dumps the ballast to travel - a factor of two back where gravity took six. The better answer is a different machine: a hydraulic backhoe concentrates the whole cut in one bucket, the architecture low gravity punishes, while a bucket-wheel excavator spreads the same removal across many small teeth taking thin, continuous cuts. Bucket wheels already move lignite overburden by the hundred million tons on Earth. Modeled

About 5,800 machine-hours is three machines at 35% duty for a season - roughly 97 working days each at 20 hours a day. Earthmoving is a capability gate, not a schedule gate: the work cannot be shipped, but it does not take long. Hauling is four times the excavation, which makes the distance from pit to hull - where the ice is against where the pad and the habitat go - a larger design variable than bucket size. Teeth are the one consumable, a few hundred kilograms of simple steel a year against 66 t of hull steel at the pad; the hardfacing that makes steel a wear alloy is still an import.

Open questions: the cutting resistance of regolith below the top few centimeters, which a penetrometer on a lander would measure; whether a bucket wheel tolerates the fines it throws in vacuum; and tool life in unweathered regolith, an abrasion test on simulant that sets whether wear is a few hundred kilograms a year or a few thousand.

Real orderings

Most of the Ladder Is Not a Ladder

An edge is logical if the later step is impossible without the earlier one and the missing item cannot be landed; economic if it is merely dearer; otherwise there is no edge.

A flat graph can be reordered by paying for it, so closure sequencing is a budgeting problem rather than a critical path - except at earthmoving. Buildout order is set by the vehicle - its capacity and how much of the manifest can move itself - not by the finished network.

The ladder

The Closure Ladder Is an Energy Ladder

Sort the rungs by the specific energy of each transformation and the order falls out by itself.

Seven orders of magnitude separate the bottom rung from the top, and each step up waits for the power plant to grow about tenfold. The closure schedule is an output of the power buildout, not an input to it. Salvage sits off the curve: 0.04 kWh/kg liberates material carrying 10-15 kWh/kg of embodied energy, a ratio near 300 to 1.

Rung zero

A Landed Cargo Hull Is a Parts Bin, Not an Ore Body

The steel was never the binding import; components are.

Extend the site's interface standards to the vehicle - common fasteners, bus voltage, connectors, bearing and seal sizes, systems designed for disassembly - and every hull becomes a spares depot. The landed hull also serves as the site's survey monument: an inertial unit, star tracker and known touchdown state give a position datum with no prior survey. The Starship page carries the vehicle side of the same argument.

Dispatch

The Rungs Time-Share One Power Bus

Unlock costs total 76-102 t of energy chain against a 55 t energy layer; the rungs cannot all run at once.

At 150 kWe, 100 t a year of oxygen draws about 80 kW and 100 t a year of water draws 57 kW at 5% ice grade or 114 kW at 1%. At low grade the pair does not fit on the bus; at high grade 13 kW is left over, less than a machine shop's 31.6 kW installed load. Closure is a dispatch problem: everything on the ladder is interruptible except the rung in production. That gives the minimum viable outpost a working definition - the largest capability set that fits on one bus at once. Modeled

The seam

Below the Spares Rung, Making Pays in Mass; Above It, Only in Schedule

Closure ratio is annual product divided by plant mass plus 0.37 t per kWe of average power.

The mass curve and the schedule curve run opposite with distance and cross between the Moon and Mars. Close to Earth, resupply is cheap and making spares is hard to justify; past Mars, a 26-month queue makes availability worth more than mass. A machine shop's product is time: a failed bracket not on the spares list becomes a one-to-three-day problem instead of a next-launch-window problem.

A few capabilities reproduce industry. A locally cast machine base lets later machine tools arrive as precision assemblies without tons of structure; a mineral-fiber line turns imported feedstock into far more finished material. Each installation makes the next one cheaper - the industrial counterpart of a reusable launcher.

Regimes

One Ladder, Two Orderings, Cut at Different Heights

Rung order is physics and does not change by destination. Gravity changes the order twice; the Element Register sets how high each site can climb.

The smelter

Metals Reduction Is an Export Rung, Not a Closure Rung

The popular account of off-Earth industry has this exactly inverted.

Reducing regolith to metal at 100 t a year takes 228-457 kW, which is 84-167 t of energy chain before any smelter is counted - and no off-Earth smelter mass has been derived anywhere. Against it sits a site demand of about 400 kg of metal a year and 66 t of stainless steel in every landed hull. For closure, the scrap yard retires the smelter. For export it does not: the Moon's best product is area - radiator and array substrate - and that is a metals-reduction product. Modeled

Device making is gated by halogens: semiconductor processing is chlorine and fluorine chemistry end to end, and lunar chlorine sits at 10-30 ppm. Mining supplies makeup for losses, not the working inventory, so the real specification is how much of the loop is recovered.

Bootstrap

Only the First Machine Tool Must Ship Complete

A capability is self-gated when its cheapest first instance needs an instance of itself.

There are four ways out: import the seed, degrade the first instance, borrow from another site, or find a construction that converges on accuracy by itself. Five of six candidate cases fail a first-principles test, because self-mobile payload drives off the ship and an inertial unit fixes position with no prior survey. The first machine tool stands: finishing a cast base to tolerance needs a tool as accurate as the one being built. Joseph Whitworth's three-plate method - lapping three plates against each other in rotation - produced accurate flats in the 1830s with no accurate machine upstream, paid for in labor, which a shop idle 85% of the time has. Modeled

China

China's Ladder Starts at the Structure Rung

The Chinese lunar program has tested regolith building material before landing any salvage or water plant.

Qualification

Making Propellant Is Not the Same as Having Propellant a Crew Will Burn

Acceptance testing is a rung the usual closure diagrams leave out.

Locally made oxygen or methane can pass a chemical assay and still carry contaminants that affect ignition, injector deposits, turbomachinery or combustion stability, and a remote site has no supplier certificates, second sources or analytical laboratories behind it. The landed vehicle helps again: its small reaction-control thrusters burn the same propellant at a few kilonewtons, which a parked hull reacts through its own weight - a screening test available on landing day. A full engine test does not fit: one Raptor at 40% throttle produces about 920 kN against a 120 t hull weighing 194 kN on the Moon, so it needs about 1 MN of hold-down and a plume trench. Full qualification sits downstream of earthmoving too. Modeled

Decision rule

Four Questions Decide Ship or Make

Applied item by item, with the reason stated whenever the mass test fails.

• Does local production displace an import? If not, it is enabling infrastructure and is judged by what it unlocks, not by a mass payback.
• Do the plant and its share of the energy system repay their landed mass within the planning horizon? Water, oxygen and some structure pay back almost immediately; higher rungs often do not.
• Does the process fit the dispatch schedule? A furnace or electrolyzer that fits the annual energy budget but collides with another critical peak is not free capacity.
• Does the destination have the chemistry, or has the design substituted around what is missing? Aluminum for copper, sodium for lithium, mineral fiber for carbon fiber, nickel-based alkaline electrolysis for platinum-group catalysts, resharpenable tools for disposable carbide.

When the mass test fails and local production is still attractive, there are three valid reasons: availability, latency and self-expansion. Modeled

Independence

The Goal Is Not Autarky

Earth is progressively removed from the mass ledger, not replicated outward.

A working settlement may import microelectronics, catalysts, pharmaceuticals, specialty polymers, optics, bearings and trace chemicals long after it stops importing oxygen, water, shielding, structure, pressure vessels and ordinary spares. Dependence becomes dangerous when the imported stream is massive, continuous, unpredictable, or needed faster than the transport network can deliver. A few tons a year of compact, high-value, long-shelf-life material can support a large local economy; hundreds of tons a year of water or oxygen cannot. The order runs: stop shipping dirt, then water and oxygen, then simple structure, then common spares - and push into metallurgy and chemistry only when energy, demand and feedstock justify it. What never closes becomes trade. Modeled

Contested

Deliberately Not Claimed

The load-bearing uncertainties, tagged with the confidence they actually earn.

Record

Provenance and Data Sources