The Moon's first power grid runs at 3000 VAC and 1000 Hz, because DC runs out of switches

What the work claims

This is an architecture and standards paper from NASA Glenn's Power Management and Distribution branch, funded under the Space Technology Mission Directorate's Game Changing Development program; everything in it is analysis and ground development, not flight hardware. The authors argue that sustained Artemis operations and any lunar commercial economy require growing from self-sufficient power elements to a connected grid, and that the grid needs a common interface at the user side and a higher-voltage backbone at the transmission side. Their answer is the UMIC: a modular, bi-directional converter whose load/source side connects ISPSIS-compliant sources and loads at 120 VDC, and whose grid side connects to other UMICs over long-distance cables at 3000 VAC, 3-phase, 1000 Hz.¹

The load-bearing numbers: the UMIC is designed for 10 kW nominal and 12 kW peak, and is parallelable to higher ratings; the grid voltage decision of 3000 VAC at 1000 Hz follows directly from a trade study covering DC, single-phase AC, and three-phase AC at 100 to 1000 Hz and 600 to 6000 V; and the study behind it found that about 1.5 kV is the maximum realistic DC voltage, because it would take roughly five series stages of switches derated to 300 V each, the best available space-rated rating.¹

How it works

The growth path is staged. Early Artemis elements are self-sufficient: every habitat, rover, or plant carries all the generation and storage it needs to survive the lunar year (Figure 1). Elements within about 100 m can already exchange power at the standard ISPSIS voltage of 120 VDC (Figure 2). As power demand grows into the hundreds of kilowatts and distances toward 10 km, an external source feeds a localized grid still at 120 VDC (Figure 3). Beyond that, voltage must rise, and the paper projects a regional grid, then a utility-scale system at megawatt demand and distances approaching 100 km.¹

The voltage-raising problem is where the physics bites. I-squared-R losses fall with voltage, so cable mass falls; but at some point insulation mass grows faster than conductor mass shrinks, and the curve turns up. The cited trade study, run for a system generating and consuming about 150 kW over a total transmission-line distance of 10 km, found mass trends roughly architecture-independent but consistently ordered: DC is lightest at any voltage above 100 VDC, three-phase AC is slightly heavier, single-phase AC is heaviest, and higher AC frequency saves mass only marginally. Raising voltage is the effective lever, until insulation mass takes over.¹

DC should have won that comparison on mass, and would have, except for one component: the switch. The study notes that no space-rated switch, used at proper derating, exists that can enable a high-voltage DC-DC converter without additional series stacking stages. With about five stages of switches derated to 300 V, the maximum realistic DC voltage is about 1.5 kV, and more stages mean compromised reliability. Against a 1.5 kV DC system, higher-voltage AC provides significant mass savings, scales to still-higher voltages by a transformer redesign, and resembles the terrestrial system. Hence 3000 VAC, 3-phase, 1000 Hz, with the UMIC carrying grid-forming and grid-syncing inverters so that compliant 120 VDC elements can plug into the backbone without each element redesigning its power electronics.¹

The frequency choice carries its own physics. Kilohertz-class AC shrinks transformer mass and volume, but raises line reactance, which on longer cables invites power-quality problems: the Ferranti effect (voltage rising toward the far end of a lightly loaded line), higher losses, and the need to size the system for reactive power. The authors are explicit that no AC power transmission has ever operated on the lunar surface, and that the interaction between power cables and lunar regolith is unknown; they suggest analytical models and terrestrial thermal-vacuum chambers with chemically equivalent lunar simulant as partial paths, with surface testing likely required before power-quality standards can be finalized.¹

Steelman: the best case for this architecture

The strongest argument is that the decision is forced by a component that exists today, not by a projection. The 300 V switch-derating limit is a here-and-now procurement fact; an architecture that routes around a real parts constraint instead of waiting for better parts is engineering, not ideology. Choosing AC also buys extensibility for free: moving from 3000 V to a higher backbone voltage is a transformer redesign, while the DC alternative was already at its reliability limit at 1.5 kV.¹

The UMIC pattern is equally well aimed. A 10 kW-class, parallelable, bi-directional converter that speaks 120 VDC ISPSIS on one side and grid AC on the other decouples element designers from grid designers, which is precisely the standardization move that lets a habitat, an ISRU plant, and a fission unit share one network without a common owner. The paper's observation that the M2M Architecture objective LI-1 calls for an incremental, evolvable system capable of scaling to global power utilization means the interface converter, not any single power source, is the enabling artifact for the industrial lunar case.¹ And the authors' restraint about standards is a virtue: they argue against drafting binding grid standards before lunar surface data exists, which is the opposite of the premature-freeze failure mode that standards bodies often fall into.

Where a skeptic should push

The most load-bearing assumption is that the trade study's mass ordering survives contact with a real 1000 Hz cable lying on regolith. The 1.5 kV DC ceiling is real, but it is a present-day component limitation, and component limits move: a space-rated switch family at 600 to 900 V, or a solid-state DC breaker that avoids series stacking, would reopen the DC case, and DC was the lightest option in the study's own mass ranking. The paper's defense of AC leans on extensibility and transformer scaling, which are genuine, but the mass comparison that eliminated DC was made against a DC system deliberately capped by today's switches.¹

Second, the unknowns are not peripheral, they are the design. No AC transmission has run on the Moon; cable-regolith coupling at kilohertz frequencies is unmeasured; the Ferranti effect on a 10 km lightly loaded lunar line is a calculation, not an observation. The authors concede that surface demonstrations may be required before standards can be finalized, which is an admission that the 3000 VAC/1000 Hz selection is a pre-data commitment, albeit a reversible one at the converter level.¹ Third, everything here is analysis: the UMIC is a development converter under a ground maturation program, the trade study is paper, and the staged Figures 1 through 5 are concepts. Nothing in this paper has been demonstrated on the lunar surface, and the gap between a 10 kW converter on a bench and a megawatt-class utility is several orders of deployment.

What a 3 kV lunar AC grid means for space power

Non-obvious implication: the binding technology for planetary power grids is the humble switch, not the generator. A 1.5 kV ceiling set by switch derating is what pushed a planetary grid to AC at kilohertz frequency, and the same constraint will reappear on Mars, on large spacecraft buses, and in any cislunar transport architecture that wants high-voltage DC for mass reasons. Anyone planning high-power DC distribution in space should read this paper as a map of exactly which component qualification unlocks which architecture.¹

Genuine threat: a kilohertz AC backbone on an uncharacterized regolith dielectric, with power-quality failure modes (Ferranti rise, reactive loading, unknown cable coupling) that can only be retired by surface testing, is a standards hazard. If the community freezes connector, voltage, and quality standards from terrestrial analogy before the first long-line lunar demonstration, early hardware could bake in the wrong grid, and retrofit on the Moon is expensive in a way retrofit on Earth never is.¹

Opportunity: the UMIC interface point is a procurement spec waiting to be written. A vendor-neutral, 10 kW-class, bi-directional ISPSIS-to-AC converter module, qualified for lunar thermal-vac and regolith-adjacent operation, is a concrete product that a terrestrial medium-voltage drive industry could build now, and every element that complies with it becomes grid-ready without knowing what the grid will be. The paper's staged Figures 2 and 3 (power exchange under 100 m, localized 120 VDC grid) define near-term demonstrations that need no new physics at all.¹

The bottom line

Established by this paper: a coherent, staged architecture for lunar surface power from self-sufficient elements to a utility grid; a specific converter design point (10 kW nominal, 12 kW peak, 120 VDC ISPSIS to 3000 VAC 3-phase 1000 Hz); and a defensible, component-grounded reason AC beat DC. Not established: any lunar-surface performance of a kilohertz AC line, any measured cable-regolith behavior, or any flight UMIC; the converter is a ground-development item and the grid is a concept. Confidence is high that the switch-derating constraint is real today, moderate that AC remains the right choice once space-rated switch voltages improve, and low in any specific timeline for the megawatt utility phase. What would confirm the architecture is a long-line, kilohertz AC demonstration on lunar simulant in thermal-vac; what would break it is that test showing the Ferranti and coupling problems are worse than modeled, which would push the community back toward DC and a switch-development campaign.