27. Make vs Ship Tradeoffs
On Mars, in those first decades of settlement we argued about pride when we should have argued about physics.
“Self-sufficiency” sounds noble on a podium; in a control room, the question is simpler: does it cost you less risk to make this here, or to ship it on the next hand of celestial cards?
In that first century we learned to treat the decision like pressure management — measured, logged, and adjusted as evidence accumulated. When we did, tempers cooled. Numbers are better at mercy than slogans.
♦
The cadence that mattered through the first hundred years was the Hohmann Cycle.
Every twenty-six months or so, the sky’s door cracked open just long enough to tempt you into depending on it. Between openings, distance turned convenience into silence. That silence calibrated the trade: for any family of parts or materials, we compared the penalties of shipping — mass, volume, launch cadence, fragility handling, buffer inventory, the chance a disruption would land — against the burdens of making — capital to stand up the line, energy and labor to run it, the yield we would actually achieve, and the long tail of spares and metrology.
We did not pretend to algebra in briefings; we spoke in plain outcomes. If downtime at the wrong week would threaten breathing, pumping, sealing, or signaling, we made it, even if it stung. If an item arrived twice a decade, weighed nothing, and demanded a temple of QA to produce, we shipped it until the frontier moved.
“Good-enough autonomy” beat perfection in every decade that mattered. A robust ninety percent local solution with drills that passed under load always outlived a fragile ninety-nine point nine that shattered at the first bad run. The frontier moved as yields rose, buffers thickened, and launch reliability wobbled. We kept the map current and we forgave ourselves for yesterday’s answers when today’s evidence improved.
Of course, all this “make vs ship” arithmetic rides on ships: arcs and burns, ascent and EDL, cranes that sing in thin air, ports that think like epidemiologists, and signed corridors where code crosses borders slower than gossip. Between windows the link thins to whispers; in-window it roars, and courage does the bookkeeping. How we built ports as redundancy engines (not dependency machines), why every route ends at a refuge, how mass, people, and meaning moved without importing fragility — that saga is too large for this page. I keep it in another codex on rocket science, space travel and inter-planetary supply chains. Here, accept the premise: logistics sets the grammar; manufacturing chooses the words.
♦
Our chokelist started as a chalkboard and became a civic ritual. Some families were “must make” from the moment dust touched our boots. Oxygen, obviously. It is continuous-use, massive to ship, and intolerant of two-year holes. We learned to split production across methods — electrolysis where water flowed, oxide reduction where it didn’t — and to pair output with buffers that were audited like bank reserves. Water processing joined the same club: filtration, distillation, and scavenging were not projects but organs. Seals, filters, gaskets, and basic glass followed quickly; they are bulky in practice, consumed daily, and unreasonably fond of failing on holidays. Fertilizers, resins, base polymers, primary structural metals — each migrated to “make” as soon as we could promise tolerances and cleanliness, because uptime does not care to wait for shipping calendars.
Other families lived on the “should make” edge and crossed over when the sums finally bent. Pumps, valves, impellers — dull, essential shapes that stopped being exotic as our machine-shop and metrology caught up. Medical disposables, IV fluids, and the reagent lines that feed diagnostics — fragile to import at scale, but tractable once we proved recipes under our own air. Printed circuit boards were a long quarrel. Early on we imported them and repaired mercilessly. As line downtime started to bite and our chemistry rooms earned their keep, we brought up single- and double-layer boards, then multilayer with controlled impedance. The day a local rework cell cleared a week of backlog in a morning, the debate ended itself.
There were also the “never make for now” pieces, and discipline saved us there. Frontier-node chips wanted their cathedrals: optics, lithography, resists, cleanrooms that ate budgets for breakfast. In those first hundred years (and beyond in many cases) we could not justify the burden while the cadence stayed low and the risk of a missed shipment could be covered by stock and substitutes. What we did instead was embrace the useful-node principle. Keep electronics repair-first. Build enough semiconductor capability to produce power devices, microcontrollers, converters, and sensors at mature, forgiving nodes. Keep boards flowing. Keep drives switching. Keep control planes alive. Import the clever brains until volume, risk, or politics forced a different answer. It was less glamorous than a factory tour on Earth. It was also what kept the miners breathing when a board failed at 03:00.
♦
Heuristics became habit. Localize what is bulky, time-critical, continuous-use, or miserable to stockpile. Localize anything whose failure would shutter a system while the sky is closed. Import what is fragile, rare, low-cadence, or demands metrology you do not yet possess — until your yields, instruments, and drill results say otherwise. We learned to check our pride against three quiet indicators. First, the import-mass slope per person: when it bent downward and stayed there, localization was working. Second, the spares backlog half-life: when it shrank quarter after quarter, our “make” lines were honest. Third, drill pass rates: when black-start, traffic-pause, and dark-Earth exercises passed using locally made replacements, no one needed a speech.
Site selection made saints and sinners of us all. A hub rooted in silica-rich plains found glass and glazing drifting from “should” to “must” earlier than others, because inputs were underfoot and kilns were modest compared to the pain of shipping panes. Virtual watershed — the practical reach of volatiles, ores, clays, and aggregates under our logistics — mattered as much as pretty geology. We extended those watersheds with roads, rovers, and haven chains, then redrew the chokelist accordingly. Some of the fastest wins of the mid-century came not from invention but from admitting that a quarry a hundred kilometers away was, for our purposes, “local.”
We rode a wave of materials and manufacturing advances that would make a catalog groan — binders that laughed at dust, additive processes that printed what used to be purchased, coatings that turned abrasion into a footnote. They deserve their own appendix. What mattered to operators in the first hundred years was that the frontier kept shifting left. What mattered to me, on more than one night shift, was that the valve that failed was a valve we could make, and that the person who installed it had practiced in daylight.
♦
The frontier moved when we stopped lying to ourselves about what quality costs. A line you can start is not a line you can trust. In the fourth settlement decade, a fashionable dome tried to print its way past metrology. The parts looked like parts. Under pressure they behaved like wishes. After three polite failures and one impolite one, the lesson crystallized: buy measurement before machinery. Micrometers before mills. Gas analytics before glossy flasks. Yield follows instruments the way ice follows winter. Once we funded inspection like life support, local lines began to keep their promises, and the make–ship boundary slid toward us without drama.
♦
Good-enough autonomy felt like heresy until it felt like relief. I remember the council that wanted frontier-node chips because their brochure gleamed. We built something else: a repair-first electronics bench that could rework boards in hours, not weeks, and a “useful-node” capability that produced power devices, microcontrollers, converters, and sensors at mature, forgiving nodes. It lacked glamour and saved whole weeks of downtime. When a control stack hiccupped during a dust storm and the miners kept breathing because a local rework cell cleared the fault by breakfast, the argument ended without applause. Not perfection. Sufficiency that survives drills. Mars’ mantra.
The same pattern played out in glass. Early on we shipped panes and paid in cracked corners. Then a hub parked itself on silica-rich plain, built a modest kiln with metered anneal, and discovered that glazing was not a luxury but a lever. Once greenhouses, viewports, and process lines stopped waiting for ships, schedules changed shape. In medicine, we ceased importing the end of every tube and began making IV fluids, disposables, and a short list of reagents whose recipes we could defend under our own air. The vaccine line came later, and with it a quiet drop in outbreak anxiety you could feel in how people walked.
Oxygen remained the cleanest proof. No one writes poetry about compressors (even AIs don’t). The day we split production methods — electrolysis near water, oxide reduction where ice was a rumor — and paired each with buffers audited like finance, oxygen left the agenda. That is autonomy: not a slogan, a subtraction from the meeting.
♦
Resource reach redrew the map more than any speech could. We coined “virtual watersheds” because geology on a page was less useful than geology under wheels. A quarry at a hundred kilometers becomes local when your haven settlement chain is honest. A brine pocket becomes part of your chemistry if your rovers and their crews have a bed and a broth at each depot.
The chokelist shifted as our reach did. Bulky inputs drifted into “must make” the moment logistics made them domestic. Fragile goods remained imports until metrology and muscle memory said otherwise. In time, three hubs learned to trade not just materials but evidence: yields, pass rates, defect maps. The frontier stopped being a guess and became a shared chart.
And droids kept the factory lights civil. Material handling, welding, pick-and-place, anneal loading — tasks that growl at lungs — belonged to machines. Humans stayed where judgment and feel mattered: diagnosis, QA, bring-up, the first shaky weeks of a new line. We learned to commission in batches rather than drips, so training, SOPs, and spares libraries matured together instead of fraying separately. And we kept heterogeneity on purpose. Two vendors for critical inputs. Two process recipes for a seal compound. Two sets of control firmware that refused to update on the same week. Elegance is a fine thing in music. In survival it is a correlation coefficient, and we learned to keep it low.
♦
The places we faltered are as instructive as the places we flew. Vanity manufacturing built cathedrals for frontier chips and neglected the gaskets that actually leaked. QA theater declared “local” with no destructive testing and a spreadsheet of imagined yields. Import complacency let buffers shrink to the comfort of a press release; the first traffic pause turned comfort to apology. Over-optimization chased the last nine-tenths of a percent while a pump line begged for a second impeller mold. When we broke these habits, clocks compressed without cheating physics. When we kept them, committees grew eloquent and nights grew long.
The governance texture mattered. Hubs that published their evidence — import mass per head trending down, backlog half-life falling, drill pass rates with local substitutes — were the hubs whose councils could say no to shiny mistakes. Budgets that funded instruments before festivals and buffers before vanity shrines pulled the make–ship line toward sanity. Standards discipline — staggered patching, signed corridors, no cross-hub auto-update — kept one hub’s bad idea from becoming a planetary outage.
♦
There was a quieter frontier in the ledger: things Mars could make that Earth could not, or not without wincing at the bill. Partial gravity and clean, cold vacuum did odd favors. Some semiconductor steps liked 0.38g—the way dopants settled, the way long anneals behaved when convection lost its temper, the way litho alignment jitters calmed in rigid, dust-damped stacks. Fiber drawing and exotic laminates found new regimes; crystals grew as if unhurried by weather. Certain chemistries — slow-precipitation catalysts, long-chain polymers, aerogels you could almost bruise with a glance — came off lines with yields that made Earth’s accountants stare. Even biologics learned a Martian accent: a handful of niche drugs stabilized better under our gravity and thermal cycles than under Earth’s humid moods.
By the end of the first century, a trickle of these oddities went home. Not a new economy — more like ballast with invoices. The transfer plans already budgeted return tonnage; the incremental cost for cargo downwell was close to nothing and sometimes improved trajectories. We packed what survived the trip and mattered to someone — power devices at “useful nodes,” specialty fibers, stubborn reagents, a crate of impossible foams — and the slow commerce taught both planets a lesson: independence and interdependence are not enemies when physics is the broker.
♦
By the last decade of the first century, the frontier had an elegance born of scars. Make what is heavy, constant, and impatient: oxygen, water processing, seals, filters, glass, fertilizers, base polymers, pressure parts, pumps, valves, PPE, boards and power devices at useful nodes, disposables and reagents you can prove. Ship what is rare, fragile, and slow: leading-edge chips, specialty optics, odd catalysts, until cadence or risk says otherwise. Revisit the boundary every quarter; move families across only on data; drill until the boredom returns.
We sized our minimum populations with these choices in mind. We loaded starships not with pride but with proofs. And somewhere in the middle of a long, dusty century, the make–ship argument became a design practice. The right answer was rarely absolute. It was written in yields and in the quiet of a control room where the valve that failed at 03:00 was a valve we knew how to make, and the person who fit it had done so last week, in sollight, for practice.