Before the reactor. Because the fuel does not mind waiting, and the plant earns its keep regardless. What follows is a recipe. Every number on the card is verified against the published literature; where a figure is assumed rather than measured, it says so.
assets/stage0_hero.jpgrecommended 2400×1000 px (wide)The only fusion fuel that exists in harvestable, stable form off-world: tritium decays in 12.32 years1, and helium-3 is effectively absent on Mars2. One kilogram, fully fused, releases about 24 gigawatt-hours of heat as a floor — and up to roughly 96 if every secondary product is burned. The floor alone is around 21 kilotons of TNT per kilogram — about one Hiroshima — delivered slowly, and on purpose.3
Mars water is naturally deuterium-rich. The ground ice we would actually mine runs about five times Earth’s seawater (assumed — not yet measured in place)4; the polar caps, eight times or more (inferred)5; the deep primordial rock, only one to two times6. We choose Arcadia Planitia, near 40° north: shallow ice, up to three-quarters of the ground by volume7, below the landing-altitude ceiling8. Not the richest ice on the planet. The reachable ice.
assets/stage0_where.jpgrecommended 1200×1200 or widerTap the water plant you were landing anyway.
assets/stage0_process.jpgrecommended 2000×600 px (wide band)The energy all of this costs is well under a thousandth of the energy the fuel will return11. The thing to engineer is landed mass and reliability, not the power bill.
First: heavy water moderates a fission reactor so it can run on natural or lightly-enriched uranium — less cargo from Earth, because deuterium barely absorbs neutrons12. Second: a stockpile of stable deuterium waits, without spoiling, for the reactor that will burn it. Third, and longest: the deuterium itself is the fuel for deuterium-deuterium fusion — surface power for the dark and the dust, the day a machine exists to ignite it. (Tritium, if wanted, is bred from lithium, not from deuterium.14)
Because the fuel is the easy half and the reactor is the hard half — and there is no reason to make the first wait on the second. No machine yet burns deuterium-deuterium at net gain, on Earth or on Mars; even deuterium-tritium net gain exists only as a laboratory flash13. But the mining, the purifying, the separating — every step has a working terrestrial twin.
So we build the supply, document the architecture, and let the reactor catch up to a fuel line that is already running. FUNNY ships the recipe. Someone else lights the stove.
Technical figures verified against the published literature (deuterium reservoir measurements, subsurface-ice mapping, heavy-water separation chemistry) and an internal physics review, June 2026. Reservoir D/H, fusion energy density, and separation chemistry are established science; the landing-site ground-ice enrichment is an assumption pending in-situ measurement.
Every figure on the card is tagged by how it stands up. Click a number in the text to land here. Where a claim is an assumption rather than a measurement, it carries the orange tag — and so does the one number in this whole recipe that nobody has measured yet.
Reservoir and ice values are peer-reviewed; fusion energetics and separation chemistry are established science; the landing-site ground-ice enrichment is an explicit assumption pending in-situ measurement. Internal physics review, June 2026.