Every doubling of cumulative production cuts unit cost by a roughly constant fraction — Wright's law, the most durable empirical rule in industrial economics. MetroVolt's fleet economics apply it at a published 14.5% learning rate, bucket by bucket.
The series decomposes plant capital into buckets with cited, published learning behavior — superconducting magnets, power electronics, structures, balance of plant — rather than applying one optimistic rate to everything. The composite 14.5% rate yields per-unit capex factors of ≈0.50 by unit 20, ≈0.46 by unit 30, and ≈0.41 by unit 50 against first-of-a-kind capital.
Compact geometry is what makes the law bite: an R₀ = 5.75 m machine is a factory product with a supply chain, not a decade-long civil project. The learning that took photovoltaics and batteries down their cost curves needs production cadence — and cadence needs a machine you can actually iterate.
The difference between $84 and $48 per MWh in the ladder is not a physics breakthrough — it is manufacturing repetition at a documented rate. Investors can diligence a learning assumption bucket-by-bucket against published industry data; that is a materially easier bet than diligencing a miracle.
| Learning rate | 14.5% (bucket-decomposed, cited) |
| Capex factor @ unit 20 | ≈0.50 × FOAK |
| @ unit 30 / 50 | ≈0.46 / ≈0.41 × FOAK |
| What it drives | the $84 → $48 ladder descent |
| Precedents | PV, batteries, aerospace composites |