Internal Rate of Return IRR

What IRR actually computes

IRR is the rate r that makes the sum of every cash flow, each divided by (1+r) raised to its year, equal to zero. For a utility-scale battery the inputs are the year-0 capital outlay (containers, racks, PCS, transformers, balance-of-plant, POI interconnection), then 15-20 years of annual net cash flows — arbitrage, capacity and ancillary revenue, minus charging cost, O&M, insurance, land lease, and periodic augmentation to offset capacity fade.

Because the equation is a high-order polynomial, a spreadsheet solves it iteratively — which is why a model can occasionally return a strange or missing answer.

Several variants circulate and must never be mixed. Project (unlevered) IRR is computed on total project cash flows before financing; equity (levered) IRR is computed on the cash left for shareholders after debt service, which Gearing / leverage amplifies. Pre-tax and post-tax IRRs differ, as do nominal and real IRRs — a nominal IRR embeds the inflation assumed in revenue escalators. IRR is an annualized, whole-of-life metric that already accounts for the time value of money, which makes it comparable across projects of different size and duration in a way a simple Payback period is not.

Where IRR sits in the project

IRR is the gate every storage project must clear. Investors compare the equity IRR against a hurdle rate — their Weighted Average Cost of Capital plus a risk premium — while lenders test the same cash flows through the Debt Service Coverage Ratio. If IRR sits below the hurdle the project does not get financed, however good the engineering looks on paper. The same Discounted Cash Flow model that produces IRR also produces NPV and DSCR, so one bad technical input propagates through the whole investment case.

This couples finance directly to engineering. Round-trip efficiency, usable energy at the POI, the degradation curve (hence augmentation cadence), auxiliary load, and availability all flow straight into the cash-flow model. A 2-3 point swing in AC-AC round-trip efficiency, or a steeper LFP fade curve, can move IRR by roughly a percentage point — enough to make or break the decision. Engineers sizing the system are, in effect, sizing the IRR: guaranteeing 100 MW / 400 MWh usable at the POI in year 20 costs oversizing or augmentation capex today, all of it landing in the same equation.

Typical values and how they move

Merchant, fully market-exposed standalone storage has historically been underwritten to equity IRRs on the order of 8-15%, occasionally higher in volatile markets like ERCOT where arbitrage spreads are wide. Contracted assets — a 10-20 year tolling agreement or capacity contract — trade certainty for return and sit several points lower, with unlevered IRRs often in the mid-to-high single digits. These ranges shift with interest rates, market saturation, and incentives, so treat any single figure as indicative, and confirm its basis before comparing it to another.

The levers are mechanical. Adding debt (gearing of roughly 60-80% is common on contracted projects) raises equity IRR whenever the after-tax cost of debt sits below project returns, at the price of DSCR headroom. In the US, the investment tax credit — historically around 30% of eligible capex for standalone storage, though specifics change with legislation — pulls value into early years and can add several points of IRR.

Revenue uncertainty is handled with scenarios: lenders size debt to the conservative P90 case, while equity typically underwrites nearer P50. The gap between the two P50 / P90 cases is itself a measure of merchant risk.

Where you actually meet it

A working engineer meets IRR inside the project's Discounted Cash Flow model — a spreadsheet with a technical-inputs tab someone must certify. Check what it assumes for round-trip efficiency (AC-AC at the POI, not DC-DC at the battery terminals), annual degradation and the augmentation schedule, auxiliary consumption, availability, and cycles per year.

The battery warranty caps cycles or annual energy throughput; a model that cycles harder than the warranty allows overstates IRR by voiding the guarantee. The commissioning RTE test report verifies the efficiency line the whole return rests on, so read it before trusting the number.

In commercial documents, IRR is the underwriting basis of term sheets, the pricing engine behind tolling agreements (the toll is set so the owner clears a target IRR), and the reference point in buyout or refinancing clauses tied to an agreed base-case model.

Ask three questions of any quoted IRR: levered or unlevered, nominal or real, pre-tax or post-tax? A levered post-tax equity IRR of 12% and an unlevered pre-tax project IRR of 12% describe very different projects. Also confirm which revenue case and which Discount rate sit behind the accompanying NPV; the two only agree on shared assumptions.

Common pitfalls

Two mathematical traps matter. First, IRR implicitly assumes interim cash flows are reinvested at the IRR itself, overstating returns for high-IRR projects; modified IRR (MIRR) substitutes an explicit reinvestment rate and is the fairer comparison. Second, a cash-flow stream with more than one sign change — a large mid-life augmentation capex, or end-of-life decommissioning costs — can produce multiple mathematical IRRs or none. In those cases the Net Present Value at the hurdle rate is the reliable decision criterion, and any implausible solver output should be sanity-checked against NPV.

The practical traps are about comparability. Comparing a levered IRR against an unlevered hurdle, mixing nominal cash flows with a real discount rate, or benchmarking a P50-based IRR against a competitor's P90-based one all produce meaningless rankings. And because IRR ignores scale, a 10 MW pilot at 18% can create less absolute value than a 200 MW / 800 MWh project at 11%. The house rule is simple: never report IRR alone — always alongside NPV, DSCR, Payback period, and the assumptions that generated it.