Educational content, not investment advice.
Unlike a conventional generator, a battery is both a power plant and a load — pumped hydro and CAES share that trait, but a battery can flip between the two in milliseconds. That single fact creates more revenue streams — and more spreadsheet errors — than almost any other asset class in the power sector. Here’s how the money actually works, with 2025–2026 numbers.
The trap is thinking of the streams below as separate income lines. They aren’t — a battery co-optimizes across all of them in the same second, and every commitment it makes to one forecloses part of another. This revenue-stacking visual shows a single asset’s day play out hour by hour.
The core stack
1. Energy arbitrage. Buy low, sell high. Charge at midday solar-depressed prices, discharge into the evening ramp. This is now the workhorse revenue in mature markets, and it’s the stream that scales with volatility — which renewables keep adding.
2. Ancillary services. Frequency regulation, responsive/contingency reserves, and newer fast-response products. The two payment shapes matter: reserves pay you to stand by — an availability price on the MW of headroom you hold in reserve — while regulation adds a performance component (mileage, or pay-for-performance) for how far you actually move to follow the signal. Batteries are exceptional at both — millisecond response, precise control — which is exactly why these markets saturate: they’re shallow. Public market analyses of ERCOT show the sequence plainly: early storage earned most of its money in ancillary services, then fleet growth crushed clearing prices, and by 2024–2025 energy arbitrage had become the dominant earner. ERCOT ran this sequence first; CAISO and Great Britain are running it now.
3. Capacity / resource adequacy. Payment for being reliably available at peak — CAISO resource adequacy contracts, capacity markets, capacity payments in tenders and mechanisms from Great Britain to Australia. This is the stream lenders love, because it’s contracted and predictable — the backbone of bankability.
4. Everything else. Transmission and distribution deferral, congestion plays, behind-the-meter demand-charge management, black start, grid-forming services like synthetic/virtual inertia (a battery’s inverter mimics inertia in software; it has no spinning mass), and fast frequency response, which even a plain grid-following inverter can deliver by reacting to measured frequency. Black start pays in a handful of island and weak grids; T&D deferral pays where a substation upgrade is looming. Everywhere else, a rounding error.
A worked example (simplified)
Take a 100 MW / 200 MWh (2-hour) system doing one cycle per day on arbitrage:
- Discharge price: $80/MWh; charge price: $35/MWh — a $45 spread on paper
- Round-trip efficiency: 88%
The classic spreadsheet error is to write: 200 MWh × $45 ≈ $9,000/day. Wrong — efficiency losses apply to the charging energy, so the calculation is:
- Revenue: 200 MWh × $80 = $16,000
- Charging cost: (200 ÷ 0.88) MWh × $35 ≈ 227 MWh × $35 ≈ $7,955
- Gross margin ≈ $8,045/day ≈ $2.9M/year ≈ $29/kW-year from arbitrage alone
That figure assumes one cycle a day, and cycles-per-year is the lever the slogan hides. Revenue scales roughly linearly with cycles — but so does life consumption, since each cycle burns one of your warranted equivalent full cycles. So raw cycle count is the wrong thing to chase. What you actually optimize is margin-per-cycle against the degradation cost-per-cycle we size in the risks section (~$7/MWh discharged).
Notice the effective margin is about $40 per discharged MWh, not $45; the 12% round-trip loss eats the rest. On top of that, auxiliary loads — thermal management, controls, standby draw — typically take another 2–4% of throughput, roughly $1.5–3/MWh against the ~$80/MWh discharge value here, and a serious model carries it as an explicit line item. So the realized arbitrage margin is closer to $38/MWh than $45 before you have spent a dollar on degradation.
Ranking the streams. The $29/kW-year above is arbitrage alone. To see why “revenue stacking” is worth the trouble, put rough magnitudes on the other three (all wildly market- and year-dependent — treat as order-of-magnitude, not forecasts):
- Capacity / resource adequacy is the steadiest. CAISO RA has historically cleared in the low-to-mid single digits of dollars per kW-month, i.e. very roughly $30–90/kW-year for a qualifying 4-hour resource — often the largest single line, and the one lenders underwrite. Critically, the payment scales with accredited capacity — set by an ELCC or duration-based derating, not by nameplate MW. A 4-hour battery typically earns near-full credit, while the 2-hour system in the worked example above is derated well below its nameplate; that’s why duration is a qualifying threshold, not a footnote. And the credit itself erodes as more storage saturates the peak — the same cannibalization dynamic that hits arbitrage, applied to capacity.
- Ancillary services can be the biggest earner early and the fastest to collapse. Frequency regulation and reserves have paid anywhere from tens to over $100/kW-year in a young, shallow market, then decayed toward arbitrage levels within a few years as batteries flood in. Never underwrite on today’s number.
- Energy arbitrage (the $29/kW-year here) scales with volatility and is the durable floor in mature markets.
| Stream | Rough magnitude | Stability | Who counts on it |
|---|---|---|---|
| Capacity / RA | ~$30–90/kW-yr (accredited, 4-h) | Steadiest — but credit erodes as storage saturates the peak | Lenders underwrite it |
| Ancillary services | Tens to $100+/kW-yr early, decaying | Biggest earner in young markets, fastest to collapse | Equity upside, never debt |
| Energy arbitrage | ~$29/kW-yr in the worked example | Durable floor — scales with volatility | The base case everyone models |
| Everything else | Market-specific | Congestion, black start, deferral — niche by design | Site-specific deals |
Forget the exact figures; the shape is what matters. In a young market ancillary can dwarf arbitrage, in a mature one capacity and arbitrage carry the project, and the mix inverts over a single asset’s life. Efficiency, auxiliaries, and degradation belong inside the dispatch math, because they decide whether a stream that looks profitable on the spreadsheet still clears once you net them out.
Why the streams aren’t additive. Here’s the co-optimization mechanic the slogan hides. Our 200 MWh battery can do one full arbitrage cycle a day. Suppose a reserve product pays you to hold 50 MWh of headroom available every evening. That reserved quarter of the pack can’t be sold into the arbitrage discharge — so you’ve traded roughly a quarter-cycle of arbitrage margin (here the pre-auxiliary ~$40/MWh × 50 MWh ≈ $2,000/day of foregone spread, using the round $40 for simplicity) for the reserve payment. If the reserve payment clears below that, stacking loses you money. That is co-optimization: the software solves, second by second, which use of each stored MWh pays most, netted against what that commitment forecloses everywhere else. There is no menu to add up.
The cost side finally cooperated
Turnkey costs fell 31% in a single year. That number is most of the 2025 story:
- Turnkey systems: BNEF’s Energy Storage System Cost Survey put the 2025 global average at $117/kWh, down 31% year-over-year, with 4-hour systems averaging $110/kWh. Regional reality check: about $73/kWh in China, $177/kWh in Europe, and $219/kWh in the tariff-burdened US.
- Packs: stationary-storage pack prices averaged $70/kWh in 2025, down 45% — the cheapest battery segment for the first time (BNEF).
- Levelized cost: BNEF’s LCOE 2026 report benchmarks a 4-hour battery project at $78/MWh globally, down 27% in a year and already below $100/MWh in six markets — squarely competitive with gas peakers for capacity-style duty.
Cheap hardware cuts both ways: your project gets cheaper, and so does your competitor’s, which compresses the very spreads you’re chasing.
Contracted vs merchant: who holds the risk
- Tolling agreements: an offtaker pays a fixed capacity payment and takes dispatch rights. The developer gets bond-like cash flow; the offtaker takes market risk. Dominant in CAISO resource adequacy deals and spreading elsewhere.
- Merchant: you keep the upside of volatility and eat the downside. ERCOT is the flagship merchant market.
- Hybrids: revenue floors with upside sharing, capacity contracts plus merchant energy, hedges and insurance products. Most real portfolios blend these.
Underneath all of it, what the market pays a battery for is optionality — the freedom to choose, every interval, where the next stored MWh goes. A tolling contract sells that freedom for a fixed check; a merchant book keeps it and rides the variance. How much you can afford to give away comes down to your debt: lenders want the fixed check, so the more leverage you carry, the more of your optionality you end up selling.
Policy tailwinds and headwinds (US-centric snapshot)
- Standalone storage remains eligible for the federal investment tax credit, but 2025 legislation attached foreign-entity-of-concern sourcing restrictions that begin to bite in 2026 — supply-chain origin is now a tax question, not just a procurement one. Get current counsel; this area moves.
- Tariffs on Chinese batteries have kept US turnkey costs at roughly three times China’s ($219 vs $73/kWh in BNEF’s 2025 survey). BNEF scenario work suggests further tariff escalation of around 60% would push US turnkey costs back up toward 2024 levels — a reminder that US storage economics are one policy decision away from reversal.
- China removed its mandate requiring storage on new renewable projects in 2025, pushing the world’s largest storage market toward merchant-style, price-driven economics.
The risks nobody puts in the pitch deck
- Spread cannibalization. Every battery built flattens the price curve you planned to arbitrage. Model fleet growth, not just your project.
- Ancillary saturation. Shallow markets, fast saturation. Never underwrite on today’s regulation prices.
- Warranty vs dispatch conflict. The most profitable dispatch is often outside your throughput warranty. Someone must own that trade-off daily.
- Degradation cost per cycle. Every cycle consumes battery life, and you can size the cost yourself: marginal cycle cost ≈ replacement $/kWh ÷ warranted equivalent full cycles. (One equivalent full cycle already moves a full nameplate’s worth of energy, so lifetime throughput is just cycles × nameplate — don’t also divide by usable depth, or you double-count it; depth only belongs in the denominator when the cycles are rated at a partial depth of discharge.) Plug in a ~$70/kWh future pack (the 2025 pack price above) and ~10,000 warranted equivalent full cycles for a modern LFP pack, and you get about $7/MWh discharged. If a cycle costs that $7/MWh in future capacity loss and the spread is only $12, your “profit” is thinner than the trader thinks. Swap in your own pack price and warranty and the number is yours. See how battery degradation enters the dispatch math.
- Basis and shape risk. You settle at a node, not at the hub average in your model.
FAQ
What’s the biggest BESS revenue stream? Market-dependent and shifting: ancillary services early in a market’s life; energy arbitrage plus capacity/resource adequacy as it matures.
How much does a grid battery cost in 2026? Global average turnkey ~$117/kWh (2025, BNEF), but regionally anywhere from ~$73/kWh (China) to ~$219/kWh (US). Full installed project costs add EPC, interconnection, and land on top of turnkey equipment.
How many revenue streams can one battery stack? Physically many; contractually fewer. Every commitment consumes state of charge and headroom — revenue stacking is a co-optimization problem, not a shopping list.
Revenue modeling, market saturation dynamics, tolling structures, and how lenders actually stress-test BESS cash flows — covered in depth in my Grid-Scale BESS: Complete Guide.