Why here
Return Propellant Sizes Everything Else
The Moon is argued about as a supplier. Mars has no customer, and that changes what the industrial question even is.
A crew that lands on Mars cannot come home until roughly 100 t of methalox exists in a tank, and the only way that tank gets filled is locally. Under the direct-to-surface architecture — Starship lands, refuels, returns — the requirement is 683 t at about 876 kg a day, which needs 438–912 kW and is three to six times the power of the entire outpost around it. Under the orbit-staged ferry — the transit vehicle held at Phobos or in low Mars orbit while a 25 t ferry shuttles crew up and down — the requirement is 107 t at about 137 kg a day.
Same mission, same crew, a factor of six in propellant, and a different class of reactor: fission surface power in the 40 kWe class, which NASA already has under contract through Idaho National Laboratory, against a megawatt-class HALEU reactor that exists nowhere. Everything downstream hangs off that one number. The reactor is sized by it, the site is chosen by how much water the architecture needs — a sevenfold swing — the fleet splits in two because of it, and the settlement’s population ceiling is stored propellant divided by propellant per seat.
The programs in the field split the same way. SpaceX plans to land Starship and refuel it on the surface. NASA’s Moon-to-Mars path runs through Gateway and the March 2026 Ignition pivot, with the SR-1 Freedom nuclear-electric demonstrator at 20 kWe — propulsion, not surface power. China holds the only funded hardware that proves Mars ascent and Mars-orbit rendezvous: Tianwen-1 delivered the Zhurong rover in 2021, Tianwen-3 flies sample return in the 2028 and 2031 windows, and the announced crewed roadmap is orbit-first with nuclear propulsion, launched on Long March 10. Nobody anywhere is funding a Mars descent and ascent vehicle, which is the piece the orbit-staged architecture cannot do without. Building
Constraint topology
Operating Envelope: What Bounds Capability on Mars
Fixed properties of the place, designed around rather than defended against. None improve with time on station.
| Parameter | Value | Consequence | Rated |
|---|---|---|---|
| Arrival Δv | 4.9 km/s from LEO, aerobraked | The atmosphere pays part of the arrival bill — Mars needs roughly a fifth of the tanker flights the Moon does for a crewed round trip | Observed |
| Ascent Δv | 4.10 km/s to low Mars orbit · 5.30 km/s to Phobos | Cheap down, expensive up. Nothing goes up that does not have to, and bulk export never does | Observed |
| Synodic period | 25.6 months · 4.7 windows per decade | Crew travel is hard-quantized; spares, food and repair strategy size to a 25.6-month clock rather than to distance | Observed |
| Surface gravity | 3.71 m/s² · 0.38 g | Above the 1.67 m/s² anchoring threshold — the only destination where terrestrially proportioned machinery reacts its own working forces without ballast or anchors | Observed |
| Atmosphere | ~6 mbar, 95% CO₂ | Feedstock for oxygen and methane, enough for aerobraking, too thin for anything else. Every candidate site sits above the water triple point | Observed |
| Insolation | ~590 W/m² nominal · 59 W/m² in a global dust storm | The storm floor sits between Jupiter and Ceres. Solar cannot be primary; fission is a capability gate here in a way it is not on the Moon | Observed |
| Breakdown voltage | 292 V at a 1.39 mm gap — the Paschen minimum | High-voltage spans are fine in open Martian air at 3 kV; unsealed terminations are not. Generation and load cluster rather than scatter | Building |
| Light time to Earth | 6–44 min one way | No teleoperation, no rescue, and no decision needing a response inside about 48 minutes can be made on Earth. Autonomy is discovered here, not granted | Observed |
Hazard topology
Hazard Fingerprint: What Damages, and at What Rate
Hazards accumulate and are defended against. Constraints bound and are designed around. The two tables stay apart on purpose.
| Hazard | Magnitude / rate | Defense | Rated |
|---|---|---|---|
| Global dust storm | Planet-wide, weeks to months, ~10% of nominal insolation | Fission and stored energy. Storms are global, so a second settlement buys nothing here — the one failure mode dispersal does not decorrelate | Observed |
| Perchlorate dust | Chlorine at 0.5–0.8 wt% of soil; continuous, toxic, abrasive | Suitports and a hard dust boundary. The same chemistry is a feedstock asset and a health hazard, which is why it appears in both tables | Observed |
| Galactic cosmic rays | Chronic; thin atmosphere, no global magnetic field | Regolith overburden, locally free. Most of the mission dose is spent in transit, so faster transits buy surface working hours | Observed |
| Entry, descent and landing | Working assumption of 50% attrition for uncrewed cargo landings | Preposition divisible cargo early and accept losses; hold indivisible cargo until the rate is measured. Split the reactor across windows, not across landings | Projected |
| Thermal cycling | Diurnal swings of order 80–100 K at mid latitudes | Materials selection and buried volume. Cryogenic tanks cannot be switched off while a storm runs | Observed |
| Partial gravity | 0.38 g, chronic; no human data at any value between 0 and 1 g | None validated. Simultaneously the settlement case’s largest unknown and its only unbounded one | Modeled |
Two of these carry more weight than their row suggests. The dust storm is what makes fission a gate rather than a preference — at 59 W/m² a solar-primary site needs roughly 8,500 m² of array to hold a 150 kWe bus, and the storm arrives on its own schedule. And the 50% attrition figure is an assumption every downstream manifest decision depends on, which makes measuring it the cheapest high-leverage thing an uncrewed pathfinder could do.
Four readings
The Same Place, Four Questions
Each topology asks something different of the same body, which is why the maps are kept separate.
Fission, year one
The global dust storm sets the floor and the propellant plant sets the load. Reactors run flat out and buy energy; storage buys peak power; the propellant plant is the site’s load-follower and is curtailed first, because its product is storable.
The first site that closes
Water, carbon, atmosphere and halogens in one place, plus gravity high enough to excavate against. Mars is the only body where the semiconductor loop is chemically possible — the constraint on making devices here is power, not chemistry.
The abort clock
Light time forecloses help and the synodic period quantizes departure. A surface emergency resolves locally or not at all, and the return trip waits for the window whatever happens in between.
Dust, then the unknown
Dust is the operational enemy and it is mitigable. Partial gravity is the chronic one, and no amount of shielding, power or redundancy touches it.
Matter inventory
Martian Resources: Water, Carbon, Halogens and Salvage
No other crewed destination holds water, carbon and an atmosphere at once.
| Source | Yields | Notes | Rated |
|---|---|---|---|
| Shallow ice, mid latitudes | H₂O, H₂, O₂, shielding | Drives the northern site preference. Under orbit staging the water requirement falls from about 570 t to about 75 t per window, which changes which sites qualify | Projected |
| Hydrated regolith, equatorial | 3–8 wt% water-equivalent hydrogen | At 75 t per window ordinary regolith suffices — no clay deposit and no specialized site needed. Mapped by the Mars Odyssey gamma ray spectrometer | Observed |
| CO₂ atmosphere | O₂ by solid-oxide electrolysis; CH₃ with hydrogen | MOXIE flew this on Perseverance. It shares no feedstock, physics or hardware with the water route — genuine dissimilar redundancy on 78% of propellant mass | Observed |
| Perchlorate and chloride salts | Chlorine at roughly 300× lunar concentration | The halogen wall that holds the Moon to substrates does not exist here. The waste stream from water processing is the semiconductor feedstock | Observed |
| Bulk regolith | Shielding, berms, pads, cast and sintered structure | Free and unlimited, and gravity is high enough to excavate it with machinery that reacts its own loads | Observed |
| Landed cargo hulls | Steel, copper, actuators, ceramic tile, pressure volume | A parts bin rather than an ore body. A Mars lander arrives with thermal protection tiles usable as furnace refractory and four flaps carrying high-power electric actuators | Building |
| Crustal uranium and thorium | Fractions of a ppm — one to two orders above chondritic | The only body in the inner system where indigenous fissile material is geologically plausible, with the aqueous concentration mechanisms that ran here. Nobody has prospected for it | Modeled |
Makes and imports
What Mars Manufactures, and What Stays on a Ship from Earth
Sort every local product by the specific energy of its transformation and the build order falls out on its own.
Bulk mass at 0.001 kWh/kg, salvage at 0.04, structure at about 0.84 realized, water at 1.1–4.1, oxygen at about 7, metals at 20–40, devices at 10³–10⁵. Seven orders of magnitude, with each rung waiting for the power plant to grow about tenfold — so what Mars can make in a given year is an output of the reactor schedule, not an input to it.
The permanent imports are the trace materials, never the bulk ones: HALEU, plutonium-238, iridium unless the electrolyzer moves off platinum-group catalysts, helium, deliberately imported carbon fiber, and the component classes no local machine shop can make. Earth’s ore bodies were concentrated by liquid water and plate tectonics over billions of years; Mars ran those processes weakly and stopped. Full treatment on The Mars Industrial Regime.
| Item | Energy | Call | Rated |
|---|---|---|---|
| Shielding, berms, pads | 0.001 kWh/kg | Make | Building |
| Salvage from landed hulls | 0.04 kWh/kg | Make | Building |
| Structure, castings, repair parts | ~0.84 kWh/kg | Make | Projected |
| Water | 1.1–4.1 kWh/kg | Make | Projected |
| Oxygen and methane | ~7 kWh/kg | Make — the whole point | Building |
| Mineral fiber and matrix composite | Process-limited | Make — Mars can make both halves, which the Moon cannot | Modeled |
| Metals from regolith | 20–40 kWh/kg | Ship — an export rung, not a closure rung, while the scrap yard holds tonnes of alloy | Modeled |
| Food | 4–12 kW per resident, lighting-dominated | Ship, until the stay crosses two windows | Projected |
| Semiconductors and devices | 10³–10⁵ kWh/kg | Ship now, make eventually — the only body where eventually is real | Modeled |
| HALEU, Pu-238, iridium, helium, carbon fiber | No local route | Ship, permanently | Observed |
Corridors
Mars Corridors: Where the Node Connects
Departure energy is nearly identical from low Mars orbit, Phobos and Deimos, so node altitude is almost free on the way out and strictly costly on surface access.
| Route | Δv | Cadence | Carries |
|---|---|---|---|
| LEO → Mars surface | 4.9 km/s, aerobraked | every 25.6 mo | Cargo one way, hulls stay. Landing the cargo vehicle is correct; landing the crew vehicle is not |
| Mars surface → low Mars orbit | 4.10 km/s | ferry cycles | Crew up to the parked transit habitat. Ascent propellant is burned, never lifted |
| Mars surface → Phobos | 5.30 km/s | ferry cycles | The higher node, bought for shielding mass and dwell time rather than for energy |
| Mars system → Earth | 1.59–2.36 km/s TEI | every 25.6 mo | Crew home on node-supplied propellant, which is what takes surface ISRU off the first crew’s critical path |
| Mars ↔ Ceres | low, slow | every 38.2 mo | The named export market, and the worst cadence in the inner system. Mars and the belt are each other’s prerequisite, so neither can be the other’s first customer |
Contested
Deliberately Not Claimed
The load-bearing uncertainties, tagged with the confidence they actually earn.
The site anchor disagrees with itself, and that is the finding. Arcadia Planitia ranks first when the architecture is direct-to-surface and water demand is high; Isidis Planitia ranks first under orbit staging, with Amazonis Planitia the most rank-stable choice across the sweep. Both answers are correct for their own architecture, which makes the disagreement a demonstration of how site selection works rather than a contradiction to resolve quietly.
Phobos water decides the architecture and will not be known in time. Japan’s MMX mission arrives in late 2027 and lands on Phobos in 2029, returning samples in 2031. Deimos gets flybys only — no orbiter, no lander, no sample — so regolith strength, anchoring behavior and extractable water at Deimos stay unmeasured, and that gap is on nobody’s roadmap.
0.38 g physiology is unmeasured, and reproduction more so. Every human data point is microgravity or one g, with nothing in between. Two separate experiments are needed: an adult physiology ring in low Earth orbit, and an animal model carried through a full life cycle at 0.38 g. Only the second gates reproduction policy, which makes it the settlement’s critical path despite answering fewer questions.
Terraforming is an import problem, not a retention problem. Atmospheric stripping at the measured 2–3 kg/s would take 41–62 Gyr to remove a built 1 bar atmosphere, so whatever is built stays. The binding constraint is inventory: the accessible endogenous budget is of order tens of mbar against about 1 bar needed. The public debate has been arguing the retention term, and the retention term is negligible.
Record
Provenance and Data Sources
- Confidence
- Observed for orbital mechanics, atmosphere and surface chemistry — Building for architecture, power and program timelines
- Anchor site
- Arcadia Planitia, 39°N, working assumption; eight candidate sites carried
- Water
- Mars Odyssey gamma ray spectrometer; SWIM subsurface ice mapping; Phoenix and Curiosity soil chemistry
- Atmosphere and ISRU
- MOXIE on Perseverance; MAVEN escape-rate measurements
- Moons
- MMX — arrival late 2027, Phobos landing 2029, sample return 2031; Deimos flybys only
- Programs
- SpaceX Starship; NASA Moon-to-Mars and SR-1 Freedom; CNSA Tianwen-1, Tianwen-3, and the Long March 10 crewed roadmap
- Weakest link
- Partial-gravity physiology at 0.38 g, and the absence of any funded Mars descent and ascent vehicle