Every other system in the Lunar Ark — the cryogenic vaults, the data substrate, the autonomous swarm, the radiation monitoring — depends on uninterrupted power. A lapse of hours risks data integrity; a lapse of weeks would compromise the preservation function. The power architecture is therefore the only system in the Codex with triple-redundant generation, multi-chemistry storage, and zero acceptable single-point failures.
Vertical solar at the south pole
The site choice — Shackleton Crater rim — was driven in significant part by power. Rim peaks on the Connecting Ridge and de Gerlache massif receive 85-90% annual illumination. Vertical solar arrays mounted there (L3-PWR-SOL-ARRAY, L3-PWR-SOL-STRUCT) capture grazing sunlight throughout the lunar day and almost all of the lunar "night" — eliminating the 14-day darkness problem that plagues equatorial sites. The arrays are sized at multi-MW capacity with significant overprovisioning; the Codex assumes 30% degradation over 100 years from cumulative micrometeoroid and dust impact (L3-PWR-SOL-DUST), and still meets demand.
Nuclear: the floor
Solar is the primary generation source. Nuclear is the floor. The Codex specifies two layered nuclear systems: small modular fission reactors (L2-PWR-NUC, drawing on NASA's Kilopower / Fission Surface Power program for the L3-PWR-NUC-CORE design) for continuous baseline output, and radioisotope thermoelectric generators (L2-PWR-RTG, L3-PWR-RTG-GPHS) as the absolute last-resort backup. The RTG class is sized at a small fraction of total demand — enough to keep the cryo vaults and core data integrity systems alive even if every other generation source fails. Plutonium-238 powering an RTG decays at a half-life of 87.7 years; the Codex's 100-year horizon plans for 45% power loss across that window, sized accordingly.
"NASA's Kilopower demonstrator (KRUSTY, 2018) validated 10 kWe surface fission power at TRL 5. Scaling to multi-100 kWe for a permanent lunar facility is the explicit goal of the Fission Surface Power program for Artemis-era infrastructure."
— NASA / DOE Fission Surface Power Program, 2024 program status.
Storage: three chemistries, three failure modes
Even with 90% solar illumination, storage matters. The Codex's L1-ESS specifies three parallel storage technologies sized for different timescales: lithium-ion (L2-ESS-LITH) for short-cycle daily smoothing, regenerative fuel cells (L2-ESS-FC) for longer eclipse periods, and supercapacitors (L2-ESS-SCAP) for transient peak demand. The diversification is intentional — lithium degrades by cycle count, fuel cells degrade by membrane fouling, supercaps degrade by leakage. None of them fail in the same way. The L3-ESS-MGMT-SOH (state-of-health) node specifies continuous monitoring with the BMS (L3-ESS-LITH-BMS) decommissioning individual cells before they reach end-of-life, replaceable via the swarm.
Distribution that survives faults
L1-PDM is the layer that makes power generation usable. Triple-redundant bus architecture (L3-PDM-BUS-MAIN, L3-PDM-BUS-ESS, L3-PDM-BUS-EMER) lets any one bus fail without dropping load on critical subsystems. Solid-state circuit breakers (L3-PDM-SWGR-SOLID) isolate faults in microseconds — faster than any electromechanical relay — preventing arc damage to harnessing. Load prioritization (L3-PDM-LOAD-PRIO) sheds non-critical loads first under power stress: lighting before sensors, sensors before comms, comms before the cryo vaults. The vaults are the last load to lose power, by design.
The architecture is conservative on purpose. A thousand-year design horizon doesn't reward elegance; it rewards enumerated failure modes and explicit graceful degradation. Every line of the power-system decomposition asks the same question: when this fails, what stays alive?