Overbuild

What it is (precise)

Three different margins hide under the one word, and a sizing conversation goes wrong the moment they are conflated. The first is the conversion margin every plant carries: DC nameplate energy typically sits about 5-15% above the AC Usable energy the offtaker sees at beginning of life, covering one-way conversion losses, auxiliary consumption, and the restricted SOC operating window.

The second is the contract margin at COD — utility-scale projects commonly overbuild DC nameplate by roughly 10-25% above the day-one contract quantity, buying the first several years of fade before any addition is needed. The third is pure overbuild: enough surplus that even the year-15 or year-20 fleet clears the contract with no additions at all. When someone quotes "a 10% overbuild", the first question is which of the three it is measured against.

The pure-overbuild arithmetic is a chain of divisions walked backward from the guarantee point. Take a 400 MWh usable-AC obligation at the POI at year 20. One-way discharge efficiency of about 95% (the square root of a ~90% Round-trip efficiency) means the cells must supply roughly 400 ÷ 0.95 ≈ 421 MWh; a ≈5-95% SOC operating window means about 421 ÷ 0.90 ≈ 468 MWh must still be installed in year 20; and warranted retention of ~70% at end of term means a BOL DC nameplate of about 468 ÷ 0.70 ≈ 670 MWh.

That is a 1.6-1.7× overbuild on standard assumptions, and across the realistic range of assumptions the multiplier runs about 1.3-1.9× — with the end-of-life SOH the warranty is written to as the single strongest lever, and the one buyers negotiate least.

Two adjacent uses of the word do not belong here. Oversizing the PCS and transformer in MVA is apparent-power headroom for reactive obligations at the POI — an AC-side power decision, unrelated to stored energy. And the PV habit of quoting a DC/AC "overbuild" ratio that clips at the inverter does not transfer: BESS overbuild is energy-side, it leaves POI MW and the interconnection position untouched, and the surplus is not clipped or curtailed — it is depth held in reserve for the fleet's old age.

Why it matters in a real grid-scale project

Overbuild is a risk-transfer instrument as much as a sizing choice. Paying for the fade at COD forecloses everything a staged plan leaves open: one cell SKU procured once, one commissioning campaign, one UL 9540 certification basis and one NFPA 855 permit edition, no mid-life construction inside an energized plant, and no mixed-age fleet for the BMS and EMS to manage.

Augmentation carries the opposite ledger — every deferred tranche is exposed to future cell prices, the near-certain discontinuation of the original part, and whatever code edition the AHJ enforces at the time of the addition. Lenders' independent engineers price both plans, but an overbuilt plant is simply easier to underwrite: the capacity that meets year-15 obligations already exists and was proven in the COD capacity test.

The surplus also works from the first cycle, which is the part spreadsheets miss. Spreading the same MW and the same daily MWh over more cells lowers the per-cell C-rate and shallows the daily depth of discharge, and both slow cycle aging — the mechanism runs the other way in a sized-to-fit plant, where a fleet built at 0.25C operates near 0.3C once retention reaches about 80%.

Gentler duty also helps keep the fleet on the near-linear part of the retention curve, away from the late-life knee. Calendar aging, though, is unmoved: the surplus ages with time, temperature, and resting SOC whether it is ever dispatched or not, so on a hot site or a lightly cycled plant — where calendar fade dominates the ~1.5-3%/yr LFP fleet total — overbuild buys noticeably less life than the cycle-life argument suggests.

Against those advantages stands the price history of the thing being pre-bought. Lithium pack prices have fallen roughly 90% since 2010 (BNEF), from around $1,200/kWh to about $108/kWh by 2025, and the time value of money sits on the same side of the ledger, so a deferred tranche has historically cost a fraction of its day-0 price and deferral usually wins on NPV.

That is why pure overbuild is rare and the blend is standard practice: a modest initial overbuild buys the first years, staged additions — cumulatively 15-40% of the original DC build over a 15-20 year term — carry the rest, and the real design variable is how much fade to pre-buy versus defer.

Typical values and standards

The numbers that anchor a review: DC nameplate ~5-15% above BOL AC usable as the conversion margin; ~10-25% above the day-one contract quantity as the common COD overbuild; ~1.6-1.7× the contracted energy for a pure overbuild to a ~70% year-20 retention, moving across ~1.3-1.9× as the SOC window, RTE, and EOL definition vary; End of Life itself defined at 65-70% of BOL nameplate in most warranties. Each figure is meaningless without its boundary conditions — the measurement bus, the reference year, and the duty profile the retention curve is conditioned on.

Every overbuilt megawatt-hour is physical, and the constraints bind at COD rather than mid-life. More DC means more containers, foundations, and inter-container separation under NFPA 855, more collection cabling and DC-bus design work, more land, and more auxiliary load — thermal management runs whether the surplus is dispatched or not, and contracted energy is net of it.

The compensating advantage is certification: the whole fleet is one product configuration under one UL 9540 listing with one set of UL 9540A test data and one permit, where a future addition must match — or formally re-open — the commissioned safety case under whatever NFPA 855 edition is then in force.

How it shows up in specs, studies and contracts

The document overbuild is actually sized against is the Capacity warranty's degradation table: warranted retention versus year, conditional on an operating envelope of cycles per year, depth of discharge, temperature, C-rate, and resting SOC. The tolling or capacity contract fixes the obligation — guaranteed usable MWh at the POI versus year — and the COD capacity test proves the surplus exists and sets the contractual baseline the guarantee is tracked against.

The interconnection study is the one document overbuild barely touches: DC energy behind the same PCS changes neither POI MW nor MVA, so a generously overbuilt plant needs no new study, where AC-coupled augmentation consumes interconnection (POI) and main step-up transformer headroom.

Reviewing a sizing case, pin down five things. The boundary: is the guarantee DC, at the PCS terminals, or net at the POI — only the POI number pays. The year: which contract anniversary the sizing clears, and on whose curve — the vendor's warranted table or the independent engineer's expected case. The duty: whether the dispatch the revenue model assumes stays inside the envelope the retention curve is conditioned on, because a heavier market duty outruns the curve and the overbuild with it.

The warranty interaction: retention floors are written as a percentage of BOL nameplate, so an overbuilt nameplate raises the absolute MWh the warranty defends. And the fallback: whether land, feeder positions, and conversion headroom are still reserved for augmentation anyway — an overbuild plan with no fallback is a bet that the fade model is right for twenty years.

Common pitfalls

The commonest failure is quoting an overbuild percentage without its denominator. A "10% overbuild" over BOL AC usable is a conversion margin, nothing more; the same phrase against the day-one contract buys a few fade years; against the year-20 obligation it is not remotely enough, since the full waterfall demands 1.3-1.9×. The sibling error is sizing at the DC bus while the obligation is AC at the POI — about 95% one-way efficiency compounds into roughly 90% round-trip, so a surplus that looks adequate at the battery terminals can still fail the capacity test at the meter.

Do not read overbuild as a revenue play. The market pays contracted energy at the POI, not installed DC, so the surplus earns nothing directly — its return is avoided liquidated damages, avoided future procurement, and slower fade.

And do not treat overbuild and augmentation as rival camps: every staged plan carries some initial overbuild to buy its first years, every overbuild plan should keep augmentation physically possible as its hedge, and the projects that end up in trouble are the ones that chose a point on the continuum at financial close and never revisited it against measured state-of-health data.