A square meter of solar cell in Earth orbit does a different job from a square meter in Oklahoma. Above the atmosphere the solar constant is about 1366 W m⁻², unfiltered by air mass, cloud, or night. On the ground, even a clear-sky panel sees a spectrum already stripped of the ultraviolet and a clock that turns the array off for half of every day. Annual capacity factors for terrestrial utility solar sit near 20 percent in good mid-latitudes. A geostationary or sun-synchronous array can run above 95 percent. The ratio of energy harvested per kilogram of panel, year over year, is on the order of five. That is not a slogan. It is the solar constant, Planck’s curve, and a planet that rotates.
Cooling is the other half of the argument, usually omitted. In vacuum a radiator sees 3 K sky. There is no convective sink, but there is also no atmosphere to trap waste heat against the array. High-efficiency III–V cells, which overheat in desert air, can be run nearer their design point if the back side is a radiator. The same vacuum that makes life support hard makes thermodynamics easy. A factory that wants gigawatts of clean power, and does not want to wait on weather, is therefore not looking for a better desert. It is looking for a place with no air.
The Moon is that place with a surface. Escape velocity is 2.38 km s⁻¹ against Earth’s 11.2. There is no thick atmosphere to punch through. Silicon and oxygen are the two most abundant elements in the regolith. A launch window toward Earth or a high cislunar orbit opens roughly every ten days; a Hohmann window to Mars opens about every twenty-six months. Polar ice, if it can be mined, is propellant that never climbed out of Earth’s well. Those four facts — flux, vacuum, mass, and cadence — are why “Sun as power, Moon as node” is an engineering sentence and not a metaphor.
The Night That Lasts a Fortnight
Equatorial lunar night is 14 Earth days. That single number wrecks a naïve solar architecture. Storing a fortnight of habitat load in batteries launched from Earth is a mass disaster: solar-plus-electrochemical systems for the lunar surface sit near 10³ kg kW⁻¹ of Earth-launched hardware, an order of magnitude worse than a continuous fission plant, because the night dominates the bill of materials. Polar geography changes the arithmetic. Peaks of near-continuous illumination sit a few kilometers from permanently shadowed floors. A vertical array on a rim can see the Sun for most of a year; a drill in the crater never sees it at all. Darkness and light are adjacent real estate, not a global condition.
NASA’s Lunar Enabling Infrastructure Accelerator, released 8 September 2026, is written for that geometry. The solicitation asks industry for vertical solar arrays that generate, store, manage, and distribute power where terrain and shadow complicate the disk; for oxygen extracted from the oxides that make up the soil; for radioisotope Stirling generators that keep systems alive in the darkest, dustiest corners; and for in-space manufacturing that cuts the resupply line. Greg Stover, directing the Advanced Research and Technology Division, put the motive without romance: close the gaps that stop a long stay. That is a request for a grid, not a flag.
Ice as a Tank Farm
LCROSS, LRO’s neutron and ultraviolet instruments, Chandrayaan-1’s Moon Mineralogy Mapper, and later polar orbiters mapped hydrogen excess in permanently shadowed craters. The working interpretation is water ice mixed with regolith, stable because the floor never rises above a temperature that would let the ice sublime into sunlight. ShadowCam is tightening the bound on pure surface frost. None of that is an assay of tonnes. It is geometry plus a hydrogen signal. Geometry is already enough to aim a lander.
If the ice can be excavated and electrolyzed, the product is the oldest rocket fuel that still matters: liquid oxygen and liquid hydrogen. Regenerative fuel cells sit at 400–1000 Wh kg⁻¹, and their viability on the Moon is tied directly to polar water. A south-polar base that splits ice is not decorating a crater. It is standing up a tank farm at the bottom of a shallow gravity well. Every kilogram of propellant made there is a kilogram that did not pay 11.2 km s⁻¹ to leave Florida. NASA and CNSA are aimed at the pole for that reason. The umbra of a crater stores the volatiles. The umbra of a planet does not.
Forty-Two Percent Oxygen, Zero Atmosphere
Typical lunar highlands and mare soils are about 42 percent oxygen by mass, locked in silicates and oxides. There is no free O₂ to breathe. The oxygen is a mineral. Hydrogen reduction of ilmenite, molten-regolith electrolysis, and molten-salt reduction are the three process families that have been run in vacuum chambers and, in reduced form, on small lander payloads. Each trades power, temperature, and yield. All of them assume a kilowatt-class plant before they assume a settlement. The 2026 NASA call treats oxygen-from-regolith as infrastructure, in the same sentence as power and manufacturing, because a base that imports oxidizer is a base that is still a mission.
The slag is not waste. Reduced metal and leftover silicate are feedstock for shielding, landing pads, and, in the longer designs, the rails of a launcher that never burns. In-situ battery skins, regolith piled as a fission shield, and concentrated-solar thermal stores all appear in the 2026 lunar-energy architecture literature as mass that no longer flies. The scholarly point is small and sharp: the launched kilogram per delivered kilowatt falls by about a factor of two once the heaviest parts of the plant are dirt you already stand on.
The Catapult That Does Not Burn
A mass driver is an electromagnetic gun laid on the ground. Gerard O’Neill’s 1970s studies, and the NASA technical reports that followed them, treated the lunar surface as the one place in the inner system where such a gun is not a fantasy of power density. Vacuum, one-sixth gravity, and a 2.38 km s⁻¹ escape speed mean the rails do not have to fight air or a deep well. A payload leaves on a Lorentz force and does not spend a second stage. Polar siting is preferred in the conceptual designs because a high-latitude array can keep the coils in sunlight for roughly 80 percent of a day-equivalent, shrinking the battery that would otherwise have to float a multi-megawatt pulse through night.
One published polar layout sizes the photovoltaic field at tens of thousands of square meters to feed about 8–10 MW into the coils, with a notional throughput of 10⁵ tonnes of material per year once the plant is mature. Those numbers are design studies, not invoices. What they establish is the power class: a mass driver is a solar problem before it is a rail problem. The same polar photons that keep a habitat alive can, at larger aperture, throw the next satellite off the world that built it. Musk’s public sketch of a self-growing city that fabricates compute and launches it with a catapult sits on that literature. The agencies mapping ice and mineralogy are the ones that make the sketch dimensional.
Beaming the Harvest Home
Space-based solar power is older than the Shuttle. Peter Glaser’s 1968 microwave-beam architecture, JAXA’s decades of demonstration, Caltech’s 2023 Space Solar Power Demonstrator, and a 2026 wave of private optical-beaming labs are one research line: collect in continuous sunlight, convert, and send the energy as a beam to a rectenna or a photovoltaic receiver that can work at night. Overview Energy and others are now building toward kilometer-class optical beams aimed at terrestrial arrays. The physics is conversion efficiency and beam diffraction, not whether the Sun is up in Virginia.
Cislunar beaming is the nearer, quieter problem. A January 2026 JPL analysis by Slava Turyshev lays out phased-array laser power from near-rectilinear halo orbit to a Shackleton-rim site. A 2 m-class array at 1064 nm delivers on the order of 0.6–0.8 kWh per day to a 1 m² receiver in the reference geometry; a 10 m aperture at 100 kW transmitter power maps to tens of kWh per day at polar sites with typical single-orbiter visibility. Habitat survival, rover traverse, a comm node, and a pilot ISRU drill all fit inside those energy bands if the aperture grows. The paper’s claim is modest and therefore useful: power beaming is no longer only a terrestrial-grid dream. It is a way to light a crater floor that never sees the disk.
A Grid Without a Planet
Early polar loads are small. Survival thermal and life support sit at 1–3 kW. A rover on duty is hundreds of watts. Pilot ISRU is 0.5–5 kW when the drill is turning. The grid that matters first is therefore not a terawatt. It is a ring of vertical arrays, a fission or Stirling baseload for the true dark, and a storage chemistry that can be refueled from local water. NASA’s Fission Surface Power work and the radioisotope Stirling line exist because polar “near-continuous” illumination is still not a promise: terrain, dust, and seasonal node geometry punch holes in the envelope.
A 2026 review in npj Space Exploration compared the architectures on launched mass. Solar plus Earth-brought batteries is heavy because of night. Fission is lighter per kilowatt until you count the shield; pile the shield from excavated regolith and the launched specific mass halves. Use polar water in regenerative fuel cells and the storage term stops being a shipping problem. Phase the plant: kilowatts in the first winters, tens to hundreds of kilowatts as a polar ring closes around 2035 in the International Lunar Research Station sketches, megawatts only when something — a mass driver, a refinery, a compute hall — actually asks for them. That is how a scholarly grid grows. Not from a rendering of a city. From a load table.
Ten Days, Not Twenty-Six Months
The factory argument is often sold as Mars. The mechanics prefer the Moon. Earth–Moon synodic geometry gives a transfer opportunity about every ten days. Earth–Mars Hohmann windows open about every 26 months and then demand a nine-month cruise. A broken part on the Moon is a fortnight from a replacement if the launch pad on Earth is awake. A broken part on Mars is a planning cycle. For a factory that iterates hardware — solar blankets, rails, electrolyzers, compute boards — that difference is the whole business model.
Lunar silicon can, in principle, be reduced and doped for photovoltaics. Lunar oxygen is already in the rock. Lunar iron and aluminium sit in the same oxides. None of those process chains is a plant today. All of them are shorter than the chain that starts with a tonne in a Texas tank and ends with that tonne on Phobos. The Kardashev move — converting a non-trivial fraction of a star’s output — does not begin with a flag on another planet. It begins with an array that never sees night, a crater that holds ice, and a gun that throws the next array without burning the last of the fuel you made.
Three Measurements on One Table
Parker’s in-situ switchbacks and the 3.8-million-mile perihelia are the star as a machine that can be visited. Polar hydrogen maps versus a dry, radio-quiet far side are the Moon as a divided world: ice in the dark, a radio lab on the back, a factory on the rim. Space-based photovoltaics plus a mass driver plus local propellant are the industrial inference those two maps permit. Solar physicists and the people who fly tankers do not share a vocabulary. They now share an object.
What is not on the table: a ³He gold rush as the reason to go, a terawatt plant on a poster, or a city that appears before a 3 kW survival bus. Helium-3 is real in the soil at parts per billion; Interlune’s offtake for quantum cooling is a customer, not a grid. Terawatts are a cadence question for Starship and a materials question for lunar silicon. The city is a mass-driver question that waits on megawatts. The scholarly order is the load table, then the ice, then the rails. The photosphere is already doing its part. It has been, at 1366 W m⁻², for as long as there has been an Earth to stand on.
Sources
- NASA. Lunar Enabling Infrastructure Accelerator (solicitation, 8 September 2026): vertical solar arrays, oxygen from regolith, radioisotope Stirling generators, in-space manufacturing. Advanced Research and Technology Division.
- NASA. Fission Surface Power and Kilopower project documentation; radioisotope Stirling converter line for dark and dusty sites.
- Colaprete, A., et al. (2010). Detection of water in the LCROSS ejecta plume. Science, 330(6003), 463–468.
- Li, S., et al. (2018). Direct evidence of surface exposed water ice in the lunar polar regions. PNAS, 115(36), 8907–8912. (Moon Mineralogy Mapper / Chandrayaan-1.)
- Turyshev, S. G. (2026). Phased-Array Laser Power Beaming from Cislunar Space to the Lunar Surface. arXiv:2508.10855. Jet Propulsion Laboratory.
- Lunar surface energy infrastructure: from in-situ resource utilization to deep space exploration. npj Space Exploration (2026). Architecture comparison: ~10³ kg kW⁻¹ solar-plus-storage vs ~10² kg kW⁻¹ fission; ISRU shield and polar RFC mass reductions.
- O’Neill, G. K. (1974). The colonization of space. Physics Today, 27(9), 32–40. Mass-driver and lunar-materials arguments restated in later NASA SP studies.
- Glaser, P. E. (1968). Power from the Sun: its future. Science, 162(3856), 857–861. Founding microwave space-solar architecture.
- Caltech Space Solar Power Project. MAPLE / SSPP in-orbit power-beaming demonstration (2023) and subsequent conversion-efficiency reports.
- Kaczmarzyk, M., et al. (and related polar-illumination studies). Peaks of persistent light and permanently shadowed regions as adjacent thermal environments. Icarus / NASA LRO-based illumination models.
- Heiken, G., Vaniman, D., & French, B. (Eds.). Lunar Sourcebook. Cambridge University Press. Oxygen ~42% by mass in typical lunar soils; ilmenite and silicate hosts.
- Kopp, G., & Lean, J. L. (2011). A new, lower value of total solar irradiance: 1361–1362 W m⁻². Geophysical Research Letters, 38, L01706. (AM0; commonly rounded 1366 W m⁻² in engineering use.)
