Background Long-duration energy storage research is overwhelmingly device-centric: it focuses on improving specific energy of materials, extending device cycle life, developing conversion chains, and optimising market or transmission design. Under device-centric paradigms, delivered-energy costs for multi-day to seasonal storage durations remain high and typically scale approximately linearly with the number of days of storage required, reflecting the assumption that energy capacity cost is proportional to volume of active material. Gap The literature lacks a formal account of architecture-and-operation regimes that structurally alter the scaling of delivered cost with storage duration. Specifically, no widely adopted mechanism class exists in which cheap bulk storage media, passive state retention, and low-activity dispatch together produce sub-linear cost scaling with duration — an architecture-first regime in which marginal cost per additional storage-day decreases rather than remaining constant. The Named Binary distinguishing Device-First Storage Economics (DFSE) from Architecture-First Storage Economics (AFSE) does not appear in the literature. Approach We formalise a levelized delivered-energy cost (LDEC) model that explicitly incorporates architectural scaling exponent α, passive daily loss fraction L, utilisation decay exponent β, amortisation horizon Y, and charging cost Pᵢn. We derive the sub-linear admissibility condition: the formal inequality on α that must hold for LDEC (d) to decrease with duration d. We present the Strongest Formulation in the four-part conditional-mechanistic-consequential-honest template, a pre-registerable Collapse Counter-Scenario (CCS), boundary conditions for regime validity, structural invariance confirmation across three independent engineering domains, and a Weil Protocol practitioner review pack. Results The admissibility condition for sub-linear LDEC scaling is: α < 1 − β (1 − CAPEXP/CAPEXE (d) ) − L·d/ (1 − L·dᵣet). Under boundary conditions (medium cost ≤ 5 £/kWhₑ, passive loss L ≤ 0. 02/day, minimum utilisation N (d) ≥ 2 cycles/year, structural amortisation Y ≥ 25 years, and charging cost Pᵢn approaching zero during curtailment), architecture-first designs can achieve α < 1, yielding sub-linear LDEC scaling over the 7–180-day duration range. Three distinguishing predictions specify empirical tests that DFSE models cannot make. Implications Architecture-first long-duration storage shifts the R status is INCOMPLETE.
José Caetano de Mattos (Tue,) studied this question.