How much battery do you actually buy? The BOL overbuild, computed.

The multiplier builds up in four steps

Start at the number the offtake contract cares about — 1.0×, the usable AC energy delivered at the point of interconnection — and walk backward to the DC cells you have to buy. Each loss layer divides what came before, so the multiplier compounds. No single step is dramatic; stacked, they push procurement past 1.6×.

The last step does most of the work. Degradation alone — sizing so the fleet still meets the contract at 70% state of health in year 20 — lifts the running multiplier from about 1.14× to 1.63×. That is the tell for what really drives procurement.

None of this means you write one cheque on day one. The alternative to front-loading is augmentation — install less at BOL and add cells mid-life to claw back the fade — which trades a smaller up-front buy for future procurement and integration risk. Either way, the degradation term is the one you are paying for; the only question is whether you pay it now or in installments.

One assumption dwarfs the rest

Vary each lever across the range a real project actually sees, holding the others at the base case, and the overbuild swings by wildly different amounts. The end-of-life definition — the SOH percentage the warranty is written to — moves it further than the SoC window, one-way efficiency, and auxiliary load put together.

Shifting the end-of-life definition from 80% to 65% SOH swings the overbuild by 0.33× — larger than the SoC window, efficiency, and auxiliaries combined (0.28×). Negotiate the warranty’s EOL number and you move more nameplate than any engineering optimisation on this list.

Method

The multiplier is 1 ÷ (one-way efficiency × auxiliary-keep × DoD × SOH). Because contracted energy and duration cancel out of that ratio, the overbuild is the same for a 2-hour or an 8-hour battery — it is set entirely by the loss and warranty assumptions, not by how big the project is. Only the one-way discharge efficiency enters, not the ≈ 90% round-trip number: the contract pays for energy delivered to the meter, so the charge-side loss doesn’t change how much nameplate you install. The base case uses 95% one-way efficiency (≈ 90% round-trip), 3% auxiliaries (the site’s own thermal management, controls, and transformer no-load losses that never reach the meter), a 95% SoC window, and 70% SOH at year 20; the range sweeps 93–97% efficiency, 1–5% auxiliaries, 90–98% DoD, and 65–80% SOH. Every figure on this page is recomputed at build time from the same open method behind the free sizing calculator — run your own contract through it, or read the full walk-through in the sizing guide.