Sizing the Power Rating: What the Megawatts Are Actually For
Ask what size a battery is and you will get one number in megawatts. That number describes the rate at which it can move energy, and it is only half the specification — but it is the half that a large share of the revenue is paid against.
On the worked assumptions below:
1-hour battery 72% of revenue linked to MW 28% linked to MWh
2-hour battery 58% 42%
4-hour battery 44% 56%
6-hour battery 39% 61%
The mix flips. A short battery is mostly a capacity and ancillary services asset that happens to store a little energy. A long one is mostly an arbitrage asset that happens to have a power rating. They are different businesses sharing a name.
That matters for sizing because the two dimensions are sold to different buyers, respond to different constraints, and should be solved separately before being checked against each other. This post takes the power rating on its own terms.
ℹ️ Note: The price shape used throughout is a single representative volatile day, used to demonstrate method. Real analysis requires the full 8,760-hour series for the node in question, and annual average spreads are narrower than a good day.
What Does the Megawatt Actually Sell?
Three things, and only one of them cares about duration.
| Revenue | Paid on | Cares about duration? |
|---|---|---|
| Capacity / accreditation | Accredited MW | Yes — via the ELCC ladder, but as a step not a slope |
| Ancillary services | MW held available | Barely — above the product's maximum delivery period |
| Arbitrage | MWh delivered | No — MW only sets the rate |
The third row is the one that confuses sizing conversations. Arbitrage revenue is paid per megawatt-hour, so the power rating contributes to it only by determining how quickly the stored energy can be moved. Once the rate is sufficient to discharge the available energy into the hours worth discharging into, additional megawatts earn nothing further from arbitrage.
The Three Power-Linked Revenues
Assumptions, labelled as such:
Capacity, 4-hour at 55% ELCC × $118,625/accredited MW-yr = $65,244/MW/yr
Ancillary services, ERCOT index at the AS share = $15,840/MW/yr
----------
Power-linked revenue = $81,084/MW/yr
Capacity is the largest and it is duration-sensitive in steps rather than continuously. The accreditation ladder from the Series C post runs 55% at four hours, 65% at six, 67% at eight and 75% at ten — so duration buys accreditation, but in jumps, and the jumps get smaller.
Ancillary services are close to pure power products. A battery holding a frequency response award is paid for the megawatts it stands ready to deliver, and the energy actually delivered is small. As the Series C battery post established, this was once most of a battery's revenue in ERCOT and is now under half, having fallen roughly ninety percent since 2023 as capacity arrived against a fixed requirement.
The rate contribution to arbitrage is real but bounded, and the bound is the subject of the next section.
Energy-Linked Revenue, and Where the Mix Flips
Arbitrage on the same representative day, one cycle per day:
| Duration | Per cycle | Per year | Power-linked | Energy-linked |
|---|---|---|---|---|
| 1 hour | $85.91/MW | $31,357/MW | 72% | 28% |
| 2 hours | $163.68/MW | $59,744/MW | 58% | 42% |
| 4 hours | $281.82/MW | $102,864/MW | 44% | 56% |
| 6 hours | $340.73/MW | $124,365/MW | 39% | 61% |
Two things to take from that table.
The flip happens between two and four hours. Below it, a battery's economics are dominated by revenues paid on its power rating. Above it, by revenues paid on its energy. That is a structural break, not a gradual shift, and it explains why short and long duration assets are underwritten so differently.
Power-linked revenue is constant across the rows. The $81,084 does not change with duration except through the accreditation step. All of the growth down the table is energy. So the question "should this battery be longer" is entirely a question about arbitrage, while the question "should it have more megawatts" is mostly about capacity and ancillary.
The Rate That Is Enough
Here is the bound that decides how much power is actually useful for arbitrage.
To discharge E megawatt-hours into the D most valuable hours of the day, the required rate is E / D. Megawatts beyond that cannot be used, because there is no more stored energy to move and no more valuable hours to move it into.
400 MWh discharged over the top 4 hours requires 100 MW
400 MWh discharged over the top 2 hours requires 200 MW
The second line is not obviously worse. Discharging the same energy into the best two hours rather than the best four captures a higher average price — on the representative day, the top two hours average $91.50 against $81.25 for the top four. So doubling the power rating to concentrate delivery into fewer, better hours raises arbitrage revenue by about 12.6%.
Whether that is worth doubling the power rating depends on what the extra megawatts cost, and on whether the point of interconnection permits them at all. On a power-constrained site the question does not arise. On a land-constrained site with a large connection, it is the central decision.
The general rule that falls out: power beyond E / D earns only capacity and ancillary revenue. If those products are unavailable or unattractive at the site, the rate that is enough is the rate to build.
What Actually Sets the Power Rating?
Four things, in the order they usually bind.
The point of interconnection, as the previous post established. This is a hard cap and it is not purchasable in any reasonable timeframe.
The power conversion cost. The inverters, transformers and balance of plant scale with megawatts and are the fixed component of the cost structure — roughly $380/kW on the assumptions this series uses. Unlike cells, this cost does not fall with duration; it is incurred once per megawatt.
The products available. A market with a deep ancillary services requirement and a capacity construct rewards megawatts. An energy-only market with thin ancillary procurement does not, and the rate that is enough becomes the rate to build.
The accreditation floor. Where capacity revenue matters, the accreditation ladder imposes a minimum duration to qualify at a worthwhile percentage — which is a constraint on energy expressed as a condition on power revenue.
Those four interact in a way worth stating explicitly, because the interaction is where sizing errors come from.
The interconnection sets a ceiling. The power conversion cost sets a price for approaching it. The available products decide whether approaching it is worth paying for. And the accreditation floor means that chasing the largest of those products — capacity — obliges a minimum energy build that the land may not permit.
So a site can fail in a way that looks contradictory: a large interconnection, adequate land for a short-duration asset, a deep capacity market, and no viable project — because the accreditation floor requires four hours, four hours requires more land than the parcel has after setbacks, and without accreditation the megawatts do not clear their cost.
That is not an exotic case. It is the ordinary shape of an urban or constrained site, and it is invisible to any analysis that treats the four factors as a checklist rather than as a system. The constraint-stack post's instruction applies directly: identify which one binds, then test whether the others are still satisfiable at that point.
The Strategy That Worked and Then Stopped
Worth setting out, because a great deal of installed capacity was sized on it and the sizing logic no longer holds.
When ancillary services paid well, the optimal configuration was obvious: maximise power, minimise energy. Ancillary products pay per megawatt held available, have short maximum delivery periods, and rarely call for sustained output. A one-hour battery could hold the same award as a four-hour one at a quarter of the cell cost.
That produced a generation of short-duration assets whose economics were almost entirely power-linked — the 72% row in the table above, and often higher.
Then the requirement filled. As the Series C battery post documented, ancillary service revenues in ERCOT fell roughly 90% since 2023 as installed storage reached scale against a requirement that is fixed by system need rather than by how much storage wants to provide it. Per-megawatt returns compressed 71% between 2023 and 2024 alone.
The assets built on that logic did not become bad assets. They became differently constrained ones: high power, low energy, and now dependent on an arbitrage market their configuration is poorly suited to. A one-hour battery captures only the single best hour of the day, and the table above shows what that is worth relative to four.
Three lessons for sizing that generalise beyond this episode.
A configuration optimised for a saturating market has a short shelf life. Ancillary requirements are fixed quantities. Any product whose procurement does not grow with supply will saturate, and the assets specialised to it will be stranded in configuration terms.
Power is the harder dimension to change later. Adding energy to an existing site is an augmentation project. Adding power means new inverters, a larger connection and possibly a new interconnection study. A configuration decision that errs toward more power is harder to unwind than one that errs toward more energy.
The revenue mix at the time of sizing is not the mix over the asset's life. A fifteen-year asset sized on a two-year-old revenue split has assumed away exactly the change that has already happened once in this market.
The Power-to-Energy Ratio Is Also a Degradation Decision
The link that sizing conversations usually miss, because power is discussed with the commercial team and degradation with the engineering team.
The ratio of power to energy is the C-rate — the rate at which the battery charges or discharges relative to its capacity. A 100 MW / 100 MWh system runs at 1C. A 100 MW / 400 MWh system runs at 0.25C.
C-rate matters because it drives heat, and heat drives degradation. A cell cycled hard at a high C-rate loses capacity faster than the same cell cycled gently, for the same energy throughput. So two batteries delivering identical megawatt-hours over their lives can reach end of warranty at different times purely because of their power-to-energy ratio.
Three consequences for sizing.
A short-duration battery is a harder-worked battery. The one-hour asset in the table above is not merely earning a different revenue mix; it is running at four times the C-rate of the four-hour asset, with the thermal management and degradation profile that implies.
Warranty terms are usually written against a maximum C-rate, not just a cycle count. A configuration that exceeds it either voids cover or requires a derate, and discovering that after the equipment is ordered is expensive.
The auxiliary load rises with C-rate. Thermal management consumes energy, and it consumes more of it at high rates. That is a direct deduction from round-trip efficiency, which the next posts in this series treat as a fixed parameter and which is in fact partly a consequence of how the asset was configured.
The practical instruction is to carry the C-rate as an explicit output of the sizing calculation rather than as an implied consequence of two numbers chosen separately — and to check it against the warranty before the configuration is fixed rather than after.
The C-Rate Rises Over the Asset's Life
A second-order effect that follows directly, and which is genuinely under-modelled.
Degradation reduces energy capacity. It does not reduce the power rating, which is set by the inverters and remains what it always was. So as the battery ages:
Year 1 100 MW / 400 MWh = 4.00 hours 0.250C
Year 10 100 MW / 340 MWh = 3.40 hours 0.294C
Year 15 100 MW / 320 MWh = 3.20 hours 0.313C
The asset becomes shorter in duration and higher in C-rate every year, at constant power. Three things follow that a flat-configuration model will miss entirely.
The revenue mix drifts. A four-hour battery accredited at 55% becomes a 3.2-hour battery over its life, and where the accreditation ladder has a step at four hours, the capacity revenue steps down with it. That is a revenue cliff arriving on a schedule the degradation curve already predicts.
The stress increases as the asset weakens. The C-rate rises exactly as the cells become less able to tolerate it, which is the opposite of how most equipment ages.
Augmentation restores energy, not power. Adding cells returns the duration and lowers the C-rate back toward its original level — which is another reason augmentation is a configuration decision rather than a maintenance one, and why the reserve funding it should be tied to throughput rather than to elapsed time.
Modelling the configuration as fixed at its year-one specification therefore overstates late-life capacity revenue and understates late-life degradation. Both errors run the same way.
How Do You Size the Power Rating in Excel?
By attributing revenue to the variable that pays it, then testing megawatts against the marginal cost of adding them.
The revenue attribution
Power_Linked_Revenue = Capacity_MW × ELCC% × $/acc.MW-yr
+ AS_MW × AS_$/MW-yr
Energy_Linked_Revenue = Σ over cycles of (Σ sell prices − Σ buy prices / RTE)
PF_PowerLinkedShare = Power_Linked / (Power_Linked + Energy_Linked)
Publish that share. It tells a reader in one number which kind of asset this is, and therefore which sizing question is worth spending time on.
The marginal megawatt test
Marginal revenue of +1 MW:
if MW < Energy_MWh / Discharge_Hours → capacity + AS + arbitrage uplift
if MW ≥ Energy_MWh / Discharge_Hours → capacity + AS only
Marginal cost of +1 MW = BOS_$/kW × 1,000 ÷ annuity factor
= 380,000 ÷ 6.8109 = $55,793/MW/yr
Against $81,084 of power-linked revenue per megawatt-year, an additional megawatt clears its cost on capacity and ancillary alone — provided those products are actually available and the accreditation holds. Strip out the capacity revenue and the same megawatt earns $15,840 against a $55,793 cost, and fails badly.
That contrast is the whole power-sizing decision in two lines, and it explains why battery power ratings differ so sharply between markets with a capacity construct and markets without one.
The concentration test
Discharge over top 4 hours: average captured price $81.25/MWh
Discharge over top 2 hours: average captured price $91.50/MWh
Uplift from concentrating 12.6%
Extra MW required to concentrate 400 MWh into 2 hours +100 MW
Extra cost 100 × $55,793 = $5,579,300/yr
Arbitrage uplift 400 × 10.25 × 365 = $1,496,500/yr
On these assumptions, concentrating is not worth it on arbitrage alone — the extra megawatts cost nearly four times what they add. They would only clear if the additional 100 MW also earned capacity and ancillary revenue, which returns the decision to whether those products exist at this node.
ℹ️ Note: Size power and energy separately first, then check the interaction. Solving them jointly in one optimisation is more elegant and much harder to interrogate — and the separate answers usually reveal which constraint was actually binding, which the joint answer conceals.
To build the revenue attribution, the marginal megawatt test and the concentration comparison, prompt Dezzmond with your price series, product prices and interconnection limit.
What Do Developers Actually Check?
- What share of revenue is power-linked versus energy-linked?
- Is there a capacity construct at this node, and what does it accredit?
- What is
E / D— the rate that is enough for the arbitrage strategy? - Does a marginal megawatt clear its cost on capacity and ancillary alone?
- Is the ancillary services market saturated? ERCOT's fell roughly 90% from 2023.
- Does the accreditation ladder impose a duration floor on the capacity revenue being assumed?
- Is the power rating the interconnection limit, or a choice? Both are common; they are different situations.
Frequently Asked Questions
What does the megawatt rating of a battery sell?
Capacity accreditation and ancillary services, both paid per megawatt, plus the rate at which stored energy can be moved. Arbitrage itself is paid per megawatt-hour, so power contributes to it only up to the rate needed to discharge the available energy.
When does more power stop adding arbitrage revenue?
Above energy ÷ discharge hours. Beyond that rate there is no more stored energy to move and no more valuable hours to move it into, so additional megawatts earn only capacity and ancillary revenue.
Why does the revenue mix flip with duration?
Because power-linked revenue is roughly constant across durations while energy-linked revenue grows with them. On the worked assumptions the crossover falls between two and four hours — 72% power-linked at one hour, 61% energy-linked at six.
Should power and energy be sized together or separately?
Separately first, then checked. A joint optimisation gives a tidier answer and conceals which constraint was binding; separate answers reveal it.
What is the C-rate and why does it matter for sizing?
The ratio of power to energy — 100 MW / 400 MWh is 0.25C. Higher C-rates generate more heat and degrade cells faster for the same throughput, and warranties are usually written against a maximum C-rate as well as a cycle count.
Does the C-rate change over the asset's life?
Yes, and upward. Degradation reduces energy capacity while the inverters keep the power rating, so a 4.00-hour asset at 0.250C becomes roughly a 3.20-hour asset at 0.313C by year fifteen — more stressed as it becomes less able to tolerate stress.
Does a bigger power rating let you capture better prices?
Yes, by concentrating discharge into fewer, better hours — on the worked day, the top two hours average 12.6% more than the top four. Whether that pays depends on the cost of the extra megawatts, which on these assumptions it does not on arbitrage alone.
Closing: The Rate, Not the Size
"How big is the battery" is a question with two answers, and the megawatt figure that usually gets quoted is the one describing how fast rather than how much.
That distinction has a practical consequence. The megawatts are sold to capacity and ancillary markets, which pay for availability and are largely indifferent to how long the asset can sustain output. The megawatt-hours are sold to the energy market, which pays for nothing else. Two different products, two different buyers, two different sizing questions — and a single number that conceals which one is being discussed.
The most useful thing a sizing analysis can produce at this stage is therefore not a megawatt figure. It is the power-linked revenue share: one number that says whether this project is a capacity asset with storage attached or a storage asset with a power rating, because the answer determines whether the next post's question is the important one or a footnote.
The next post is that question, and it is the one this series exists for: how many hours, and why four rather than five.
Sources: NREL — Utility-Scale Battery Storage, Annual Technology Baseline · Modo Energy — Battery Dispatch Model · Timera Energy — What Battery Durations Are Investable? · Berkeley Lab — Value of Adding Up to 4-Hour Duration Batteries to Solar or Wind