The analyses underpinning the lunar surface program, published in full. Each stands on its own and is meant to be evaluated independently.
The energy figures are reproducible from the open-source model at
github.com/dubthree/lunar-propellant-energy-model.
These are working drafts, not peer-reviewed publications; versions and numbers will move as the model is pressed.
Canonical model
A Common Electrical-Energy Basis for Comparing Lunar Oxygen and Propellant Production Routes
Walter Kueffer · Draft v0.13 · 2026-07-01 · 12 pp
Published energy figures for lunar oxygen extraction are not comparable: hydrogen reduction is reported as electrical kWh/kg, carbothermal as thermal, and the electrolysis routes not at all. This paper places five routes (hydrogen reduction, carbothermal, molten regolith electrolysis, molten-salt electrolysis, and polar water-ice mining) on one electrical-equivalent kWh/kg O2 basis under a single explicit system boundary, with Monte-Carlo propagation of literature parameter ranges and continuous standing losses charged to every route, the water route included. It validates by independently reproducing the one clean published anchor (Leger et al., PNAS 2025) and finds the PSR water route cheapest in 89% of paired trials on an all-electric basis, because it is the only low-temperature route; a solar-thermal sensitivity shows that concentrated sunlight at a sunlit site inverts the ranking.
Companion analysis
Compute Waste Heat as a Low-Grade Thermal Offset for Lunar ISRU
Walter Kueffer · Companion v0.2 · 2026-07-01 · 3 pp
A narrow, Second-Law-safe claim: data-center / GPU waste heat (~315-350 K) can supply the low-grade thermal demand of the PSR water route (sublimation at ~273 K), about 11% of that route’s production energy after heat-exchanger pinch and effectiveness, but cannot drive any high-grade reduction or electrolysis route (1073-1900 K). On energy grounds a 100 kW compute facility’s rejected heat could serve ~500 t O2/yr of water mining, an upper bound set by energy balance, not by conduction-limited delivery. The value is that the heat is otherwise dumped, and that it lands specifically on the only route that yields complete propellant.
Position paper
Permanently Shadowed Regions as a Shared Compute / ISRU Hub
Walter Kueffer · Position paper v0.2 · 2026-07-01 · 3 pp
A siting argument: a PSR is simultaneously the best lunar location to reject compute waste heat and where the water resource is. Co-location amortizes the one hard thing both systems need, power delivered into permanent shadow. Under an explicit radiator energy balance with realistic vertical-panel geometry, a PSR saves a median ~10 t of radiator per MW of compute (IQR ~7-16). The conclusion is deliberately narrow: co-location is justified, if at all, by compute siting economics; the ISRU heat cascade is a cheap bonus to capture once there, not a reason to go.
Related: the lunar surface overview and the
cosmo-regulus radiation-tolerance model.