The Deltas.

Heat & Power Charges

On Earth: ~1 GWe out, with waste heat dumped into water and air the plant never runs out of.

Computed — and we were wrong twice. We expected the Martian surface to pay back; it charges. In July we called the charge mild, crediting the thin Martian air with carrying half of the waste heat; recomputed in September, that air carries about a tenth. At practical radiator temperatures the plant still keeps 83–97% of its Earth output — the steam cycle sets that — but the field of steel radiators roughly doubles, to three to five square kilometres. A mass tax, not a thermodynamic wall, and the field itself is the natural first product of Martian steelmaking. Deep space charges hard: shedding gigawatts by radiation alone forces the radiators hot, which surrenders 400–800 MW of output — and the reference design’s water-cooled blanket, built for Earth, does not travel there at all. One environment taxes you; the other rewrites the machine.

Fuel Supply Charges

On Earth: the reactor breeds its own tritium from lithium that arrives by truck.

Mars inverts the supply problem. Deuterium is local and abundant; lithium, needed to breed tritium in the blanket, runs around ten parts per million — too thin to feed itself. Computed, the whole axis turns on one number: if the blanket breeds as designed, the plant feeds itself on about half a tonne of imports a year, and a thirty-year lithium stock rides in on a single resupply window. If the breeding margin — seven percent on paper, about the size of the uncertainty in calculating it — falls short, tritium must be imported forever, and tritium decays in the hold on the way. The fuel cycle is not a cost; it is a coin toss the blanket designers hold.

Landed Mass Charges

On Earth: a twelve-metre machine is craned into place at a grid site.

We ran the first honest mass budget: about eighty thousand tonnes landed on Mars, plant and radiators together (seventy thousand before the September correction to the radiator field) — over a thousand landings if every part is shipped. And it cannot be shipped: a single wound superconducting coil is an indivisible three-to-five-hundred-tonne object, several times any lander’s capacity, so the coils must be wound on site or not at all. The wall is bought down not by cheaper rockets and not by shrinking the machine — shrinking makes it worse, because the metre and a half of neutron shielding does not shrink with it — but by building: local steel and local ground displace roughly five kilograms for every one a seed factory costs to land. The quiet surprise in the arithmetic: per landed kilogram, the fusion plant still beats the small fission reactors planned for the same ground.

Magnets & Cryogenics Pays back

On Earth: superconducting coils are held cold against a warm 300-kelvin ambient.

The one plainly easier thing — and, computed, the one plainly small thing. The cold surroundings do reduce the cost of holding superconducting coils at twenty kelvin, exactly as expected. The saving is a few megawatts against a gigawatt plant: real, favourable, and less than half a percent of output. We predicted the right sign and the wrong size, and we report both. Honesty about a small dividend is worth more than enthusiasm about it.

Neutron Damage & Replacement Charges

On Earth: a neutron-damaged wall is swapped from a supply chain.

Fusion neutrons wear out the wall that faces the plasma, and this — not fuel — turns out to be the real recurring bill. Computed over a thirty-year life, replacement steel outweighs the fuel stream a thousand to one: hundreds of tonnes of blanket and wall riding every resupply window, staged one window ahead, forever. On Mars, local steelmaking can cut the stream several-fold — if reactor-grade steel can be made there at all, which no one has shown. In deep space there is no such relief: the plant must fly with its own lifetime of spares, at a multiple of its own mass. Lifetime stops being a maintenance schedule and becomes the design constraint.

Running It Alone Pays back

On Earth: operators stand in the loop, seconds from the machine.

This is why a stellarator, of all fusion machines. With no large plasma current it cannot suffer the sudden disruptions that plague the tokamak, and a fusion plant cannot run away — at worst, it turns itself off. A machine that fails safe and runs steady is the only kind you can trust to operate alone, minutes of light-lag from the nearest human. Off-world does not merely tolerate autonomy; it demands it — and this is the machine built for it.

Where Each Delta Falls