The augmentation planner. A contract crossing a curve.

Overbuild, stage, or blend — who pays for the fade, and when

Front-loading buys cells that sit underused for years and spends capital early; back-loading defers spend but exposes the project to future cell prices, integration labour, and the near-certainty that the original cell SKU is discontinued — formats and chemistries turn over every few years, so a year-10 top-up almost never uses the original part. The historical case for deferral is the price of the thing being deferred: lithium pack prices fell roughly 90% from 2010 to 2025, from around $1,200/kWh to about $108/kWh (BNEF), and the time value of money argues on the same side.

But the bet is symmetrical — deferral converts a known year-0 price into a forecast of a price nobody controls, and that forecast is yours to make: the planner takes your block cost and your discount rate and publishes no price forecast of its own. Lenders' independent engineers scrutinise the augmentation reserve directly, because the capacity warranty and the revenue contract both depend on the plan holding.

Timing is not only financial. At constant MW output the real C-rate rises as capacity fades — a fleet built at 0.25C runs near 0.3C once retention reaches about 80% — and both the BMS current limits and the warranty degradation model must budget for that creep unless new capacity relieves it. Adding capacity relieves per-cell duty — lower depth of discharge and C-rate spread over more installed energy — which slows the cycling contribution to fade; it does nothing for calendar ageing, which arrives on schedule regardless. The alternative is concrete: periodic capacity tests come in below the guaranteed table, and the shortfall lands as liquidated damages.

Three working defaults

These bands are practitioner judgement — the author's operating rule of thumb from real projects, not an industry statistic. Nobody publishes a number here; treat them as a starting position to beat with your own curve and prices.

• Most cost-effective, lowest availability: ~0–5% overbuild at commercial operation (COD) with roughly annual augmentation — the least idle capital, and the most outages, orders and retests.
• Most expensive, highest availability: ~30–40% oversize on day one — the floor is cleared by construction for years, at the price of cells that idle.
• Balanced: ~10% overbuild, first augmentation around year 3–5, then roughly every five years — the shape most real projects converge on.

Oversize day one vs frequent augmentation

Table is wider than the screen — scroll it sideways →

What you must reserve before the first block fades

A staged plan is only cheap if the addition is an installation, not a construction project inside an energised plant. That is decided at the original design: reserved land with foundations, trenching and conduit run; spare MV feeder positions; PCS and transformer headroom for the future blocks; a DC voltage window the year-8 market can still match; and priced module options in the supply agreement. Augmentation space left as bare land tends to be value-engineered away or quietly repurposed — a deferral plan with no reserved plot, no conversion headroom and no answer on cell availability is not a strategy; it is an assumption that year-8 engineering will be free.

Mixing vintages is an engineering problem, not a spreadsheet line

A series string is limited by its weakest module, and racks paralleled on a shared DC bus are forced to the weakest string's operating window — so adding fresh, high-state-of-health (SOH) strings alongside aged ones creates a mismatched fleet the BMS and EMS must manage. DC-coupled designs handle it with per-string DC/DC conversion (the copper box in the diagram above), and vendors typically require augmentation racks on a separate DC bus or a dedicated PCS input. Every addition must also respect the as-built design — above all the DC voltage window (typically up to 1500 VDC) and the safety case: container footprint and separation under NFPA 855, and the site's explosion-protection basis, where UL 9540A test data informs NFPA 68/69 measures. New containers generally must match — or formally re-open — the commissioned safety case with the authority having jurisdiction (AHJ), under whatever code edition is in force at the time of the addition.

Where the obligation actually lives

You meet augmentation across five documents. The battery supply agreement carries the capacity warranty and its warranted degradation curve, with operating-envelope conditions — cycles per year, depth of discharge, temperature, C-rate, resting SOC — that void coverage if breached. A capacity-maintenance or long-term service agreement commits the integrator to keep deliverable energy on that curve, augmenting at its own cost or at pre-agreed unit prices. The tolling or capacity contract defines the tested energy at the POI and the liquidated damages for shortfall, while availability guarantees cap how much outage time an augmentation installation may consume. The interconnection study, finally, fixes whether new blocks even fit the approved plant.

Working a project, ask five questions:

1. Where is guaranteed energy measured — DC, PCS terminals, or net at the POI — and does the vendor state the degradation curve at that plane?
2. Who bears future cell-price and technology risk? Fixed-price augmentation options are valuable and rare — ask how capacity is priced per rack, how long the option holds, and what happens if the vendor substitutes a different cell mid-life.
3. What physical provisions are actually built on day one?
4. How does the warranty treat a blended fleet — does the supplier restart the degradation clock only for newly added equipment, leaving the aged fleet on its original curve?
5. Does the interconnection agreement already permit the addition, or does it trigger a fresh study cycle?

Executing an event, and re-baselining after it

An augmentation event is an outage with paperwork on both sides. Before it: the order placed by the decide-by date, the outage window agreed inside what the availability guarantee allows, and the safety case confirmed with the AHJ. During it: isolation, installation, and the commissioning capacity test that resets the contractual baseline. After it: the degradation model is re-baselined, because the schedule is steered by rack-level state-of-health telemetry against periodic capacity tests — not by the financial-close spreadsheet. Fade curves are warranted only under a specified duty; if the market pulls the plant into heavier cycling, real fade outruns the curve, and extrapolating early state-of-health data linearly to year 20 flatters the budget from the other side, because degradation has a late-life knee. Re-forecast at each capacity test and each major regulatory change.

The checklist, ordered by when each item stops being reversible

1. At contract signature: the measurement plane of the guaranteed MWh; the warranted curve at that plane; fixed-price augmentation options; warranty treatment of a blended fleet.
2. At design freeze: pad, foundations, conduit and spare MV feeder positions; PCS and transformer headroom; a DC voltage window future blocks can match; interconnection filed for the full build-out.
3. At COD: the commissioning capacity test that becomes the baseline; the augmentation reserve named in the financial model.
4. Every operating year: state-of-health telemetry against the warranted curve; re-forecast the crossing; check the code edition and tariff landscape.
5. At each crossing minus lead time: the order — this is the decide-by date the planner computes, and after it passes, every remaining option is worse.

Every default, and where it comes from

Table is wider than the screen — scroll it sideways →

Assumptions last reviewed 2026-08-11.