The load, honestly stated
Concept studies tend to quote peak generation and leave the load vague. The load is the hard part, because almost none of it can be shed. Life support runs continuously. Thermal control runs continuously and gets harder at night, not easier. The ISRU plant can be throttled, and is the only significant flexible load on the base.
Why the array is vertical
At this site the sun stays low. A horizontal panel of the kind used in Earth orbit would see it at a grazing angle for the whole lunar day and produce a fraction of its rating. A vertical array, rotated slowly about one axis to follow the sun round the horizon, keeps the incidence angle near normal for as long as the sun is above the local horizon. The sun never climbs more than about 1.5 degrees above the horizon at the pole, the tilt of the Moon's spin axis to the ecliptic, and modelling based on the Lunar Orbiter Laser Altimeter (LOLA) gives the best Shackleton rim site 240 days of sun without a break.
This also solves a dust problem for free. A horizontal surface collects settling regolith. A vertical one sheds most of it, and what remains can be cleared electrostatically without anyone touching the panel face.
The night
This is the section that decides the architecture, and it starts by saying what the night is. At the equator it is 14.7 days, and at even the reduced 27.6 kW habitat load that is 9,740 kWh. Lithium ion at a generous 250 Wh/kg, before packaging, thermal control and depth of discharge margin, gives about 39 tonnes of battery, a thirty nine billion dollar battery at a million dollars a kilogram. That is the case most readers know, and it is not this site. LOLA based illumination modelling at 240 m resolution finds a site on the Shackleton rim lit continuously for 240 days with a longest dark spell of about 1.5 days; the 20 m resolution model gives the best rim clusters a longest shadow of 65 to 66 hours and the best Connecting Ridge cluster 112 hours, and puts the best sites at 2 m above the ground in sun 77 to 88 percent of the year. A 36 hour spell at 27.6 kW is about 1,000 kWh, roughly 4 tonnes of cell before packaging. A 66 hour spell is about 1,800 kWh and 7.3 tonnes, already heavier than the reactor. The reactor is in the architecture for three reasons that survive that arithmetic: the longest gap at the exact array position has not been computed, and a few metres of position or height changes it; the array's output over the year is 12 to 23 percent below what a lit array gives, and the winter brings the gaps in a series; and the ISRU plant and the sintering gantry are loads that should not follow the sun. That is a weaker argument than a fourteen day night, and this document says so.
| Option | Mass for the equatorial 14.7 day night | Landed cost | Mass for a 66 hour spell | Verdict |
|---|---|---|---|---|
| Lithium ion only250 Wh/kg, 80% depth of discharge (DoD) | 39 t | $39 bn | 7.3 t | Heavier than the reactor even for the rim spell, and the longest spell here is not yet computed |
| Regenerative fuel cellH2 and O2, 500 Wh/kg, near the top of NASA Glenn's 110 to 550 Wh/kg range | 19 t | $19 bn | 3.6 t | Lighter than cells, and NASA Glenn ran a full system on the ground in 2026, but it adds gas storage and thermal switching and it still follows the sun |
| Fission surface power40 kWe class | 6 t | $6 bn | 6 t | Viable, does not depend on the local horizon, and the only option that scales |
| Fission plus 210 kWh bufferthe architecture here | 6.9 t | $6.9 bn | 6.9 t | Buffer covers reactor transients and peak load |
Fission surface power
NASA's Fission Surface Power project set the class this architecture is sized around: a 40 kWe reactor of no more than six tonnes, a design life of at least ten years and no crew intervention. Three industry teams, Lockheed Martin, Westinghouse and IX, were awarded twelve month Phase 1 design contracts of about 5 million dollars each in June 2022 and completed them in 2023. In August 2025 NASA's directive on fission surface power moved the target to a minimum of 100 kWe with a closed Brayton cycle, carried on a heavy class lander and ready to launch by the first quarter of 2030. The 40 kWe unit here is therefore a class, not a product. If the 100 kWe unit flies first, one of them carries this base with margin to spare. It is still the single largest piece of infrastructure the base depends on that does not yet exist.
NASA's own 40 kWe point design, worked up by the Glenn Compass team for a south pole outpost, is the reference for what the unit is and where it stands. It is a heat pipe cooled reactor on high assay low enriched uranium, with eight 6.2 kWe Stirling convertors in opposed pairs, deployable radiators and a shadow shield rather than a full shield, carried to site in three sleds by a repurposed crew rover chassis. The study places the reactor about one kilometre from the crew, the distance at which the shadow shield holds the dose below 5 rem a year for permanently present crew, and it notes that moving the reactor further away or over the horizon does not reduce the shield mass. The base follows that. The unit stands about one kilometre out on the far side of a natural rise, with a regolith berm on the habitat side as a second line, and the crew's working area stays inside the shielded sector. Power leaves the reactor at plus or minus 2,800 V DC over an aluminium cable that the study puts at about 45 kg for the kilometre, and is bucked to 120 V DC at the habitat end.
Increment three depends on hardware that has not been built. The closest flown relative is KRUSTY (Kilopower Reactor Using Stirling Technology), the 1 kWe class Kilopower core that NASA and the Department of Energy ran at the Nevada National Security Site in March 2018, and a 40 kWe unit is a different machine. If fission surface power slips, the base does not stop, but the ISRU plant does. The plant is the largest flexible load and would be shut down through the night, which turns the outpost from a construction base back into a shelter with a resupply bill. That is the schedule risk worth arguing about, and it is not a habitat risk.
Storage, and what it is actually for
210 kWh of lithium ion is not there to carry the night. It is there for four things that all happen on timescales of minutes to hours: reactor start transients, the peak when the airlock cycles while the greenhouse is at full photoperiod, ride through when a string trips, and the shadow transits when a local ridge takes the low sun off the array faster than the reactor can ramp.
| Function | Energy | Duration | Why storage and not generation |
|---|---|---|---|
| Reactor ramp coveragecold start | 40 kWh | 1.5 h | A reactor cannot follow a step load |
| Peak clippingairlock plus greenhouse | 22 kWh | 40 min | Cheaper than sizing generation for a rare peak |
| String trip ride throughsingle fault | 28 kWh | 1 h | Buys time to isolate and reconfigure |
| Shadow transitwhen local relief occults the sun | 95 kWh | 3.5 h | Array falls off faster than the reactor ramps |
| Total sized | 185 kWh | 210 kWh installed, 12 percent margin |
Distribution
Everything runs at 120 V DC. There is no reason to carry an inverter and its losses when every source on the base is inherently DC and every load can be. The exception is the ISRU plant, whose electrolysis cells want low voltage and very high current, and which therefore has its own converter sited at the plant rather than at the habitat. The other exception is the reactor line: a kilometre at 120 V would need a cable of hundreds of kilograms, so the reactor's converter boosts to plus or minus 2,800 V DC for the run and a second converter at the field combiner brings it back to the bus voltage, as in NASA's study.
Dust, and the slow loss nobody budgets for
Regolith is electrostatically charged and moves near the terminator. It settles on everything. The Apollo dust detector experiments at the 12, 14 and 15 sites measured settling dust at about 100 micrograms per square centimetre a year, enough to cut the output of their solar cells measurably over the years they ran, and on Apollo 16 and 17 the rover batteries ran over their temperature limits because dust on their radiators could not be brushed off.
A vertical array sheds most of what lands on it. What remains is cleared by an electrodynamic dust shield, a set of transparent electrodes on the panel face driven with a travelling wave that walks charged particles off the edge. NASA demonstrated it on the Blue Ghost lander in March 2025, where it lifted regolith off glass and radiator surfaces, and it has no moving parts. The alternative, sending a crew member out with a brush, both wears the coating and consumes the scarcest resource on the base, which is crew time.
Mass and cost
| Element | Mass | Power | Increment | Note |
|---|---|---|---|---|
| Solar array and mastone axis tracked | 1,240 kg | 48 kW peak | 1 | Vertical, about 120 m² of cell at 30 percent efficiency and 1,361 W/m² |
| Battery, 210 kWhlithium ion | 890 kg | none | 1 | Includes thermal control and packaging |
| Fission surface power40 kWe class | 6,000 kg | 40 kW | 3 | Includes shielding and radiator |
| Power management and distributiondual bus | 410 kg | 1.1 kW losses | 1 | 120 V DC throughout |
| Cabling1 km reactor line and site runs | 320 kg | none | 1 to 3 | Aluminium, not copper, on mass grounds. The kilometre at 2,800 V is about 45 kg of this in NASA's study; the 120 V site runs are the rest |
| Electrodynamic dust shieldsarray face | 34 kg | not yet budgeted | 2 | Removes all crew time from cleaning |
| Total | 8,894 kg | 88 kW installed | $8.9 bn landed at $1M/kg |
Power is the single heaviest subsystem on the base, at nearly nine tonnes, and the reactor is two thirds of it. That is the price of not being dependent on the sun, and it is the reason the ISRU plant has to earn its keep: a base that lands nine tonnes of power hardware and then still ships oxygen from Earth has not made an argument for itself.
Failure modes
| Failure | Effect | Buffer | Response |
|---|---|---|---|
| Array string tripone of six strings | 8 kW lost | 210 kWh | Reconfigure. Daylight surplus absorbs it entirely. |
| Reactor scramin a dark spell | 40 kW lost | 210 kWh at 27.6 kW is 7.6 h | Shed ISRU and greenhouse, hold life support and thermal. Restart, or hold survival load until the sun clears the ridge. |
| Battery string failureone of four | 52 kWh lost | Remaining 158 kWh | Isolate. Shadow transits become tighter but remain covered. |
| Bus faultshort on one side | Half the base | Dual bus | Isolate the faulted side. Every load has a feed from both. |
| Dust accumulationgradual | Output falls | Weeks of margin | Electrodynamic clearing while the array is lit. |
What is not yet known
The longest dark spell at the exact array position has not been computed. The 1.5 day figure is for the best rim site in a 240 m per pixel LOLA model; the 20 m per pixel model gives the best rim clusters 65 to 66 hours and the best Connecting Ridge cluster 112 hours, so the design gap sits somewhere between a day and a half and five days until the site is modelled at the mast position. How much dust a vertical face collects against a horizontal one has not been measured; the Apollo detectors were horizontal cells. The shield's power draw at array scale is not yet budgeted. No fission unit exists. NASA's 2025 directive targets a minimum of 100 kWe ready to launch by the first quarter of 2030, and if it slips the ISRU plant runs only while the array is lit and the outpost falls back to a shelter with a resupply bill. That is the schedule risk worth arguing about. The communications document carries the mast at 0.9 kW continuous while the load table here folds 0.6 kW of comms into the cupola line; the two have to be reconciled.
Increment plan
Increment 1, solar and storage
48 kW of array and 210 kWh of battery, landed with the habitat. This supports the habitat while the array is lit and for 7.6 hours of dark at full load. Early increments accept a reduced night mode through the longer spells.
Increment 2, second string and the reactor's site
Electrodynamic shields on the array face, a second distribution string so no single bus fault can reach the habitat, and the reactor's ground work: the berm, the graded kilometre for the cable and the field combiner, laid while the unit is still being built on Earth.
Increment 3, fission surface power
The 40 kWe unit, delivered by rover to a point about 1 km out and cabled back at plus or minus 2,800 V DC. This is the step that makes continuous ISRU operation possible and turns the base from a shelter into a plant, and it is the one that waits on NASA's reactor schedule.
Increment 4, capacity for construction
Additional array to cover the sintering gantry at full duty. Construction is the load that grows fastest as the site expands.
Sources and further reading
Every figure in this document traces to one of these. Agency documents and peer reviewed work first, reporting only where it is the primary record.