Ask a developer how big a battery is and you'll usually get a number in megawatts. But power is only half the spec. The other half — and the one that increasingly decides whether a project pencils — is duration: how many hours the battery can sustain its rated output before it's empty. A 100 MW battery with two hours of duration (200 MWh) and a 100 MW battery with six hours (600 MWh) look identical on a one-line diagram and behave like completely different assets on the grid.
For most of the last decade, the industry's default answer was four hours, and in many markets it still is. But the question is reopening. As short-duration storage saturates the easiest part of the grid's need, and as battery costs fall faster than almost anyone forecast, the calculus is shifting toward longer systems. This is the framework we use at Solyx to decide — and why, for a solar-paired hybrid plant, we keep landing in the four-to-six-hour band.
In this paper
01What duration actually buys
A battery earns its keep in two fundamentally different ways, and duration affects each differently.
The first is energy arbitrage — buying (charging) when power is cheap and abundant, selling (discharging) when it's scarce and expensive. The wider and longer the daily price spread, the more a battery can earn, and the more hours of duration it can profitably use. The second is capacity — the promise to be available during the few hours each year when the grid is genuinely short of supply. Grid operators pay for this reliability, either through explicit capacity markets or, in energy-only markets like Texas, through the scarcity prices that spike when reserves run thin.
Here's the catch that makes duration interesting: these two value streams reward length on completely different curves. Arbitrage value rises fairly smoothly with duration — more hours, more spread to capture. Capacity value does not. It rises steeply at first and then hits a wall.
| Duration | Primary value | What it covers | Capacity behavior |
|---|---|---|---|
| 1–2 hr | Ancillary services, short spikes | Frequency regulation, brief price peaks | High credit early, erodes fast as more is added |
| 4 hr | Capacity + evening arbitrage | The core evening net-peak window | The market default — but credit now declining |
| 4–6 hr | Firm evening shift + deeper arbitrage | Full evening ramp plus the shoulder hours | Restores capacity credit lost to 4-hr saturation |
| 8+ hr | Multi-hour / overnight shifting | Bridges into the early-morning peak | High credit, but cost per usable hour climbs |
02The four-hour wall
Four hours became the industry standard for a reason. In a grid with a lot of solar, the reliability problem isn't the middle of the day — it's the evening, when the sun drops but demand stays high. A four-hour battery, charged on cheap midday solar, can cover that evening net-peak window almost perfectly. For the first wave of storage, four hours was exactly the right tool.
The problem is what economists call saturation. The capacity value of a battery — formally, its effective load-carrying capability, or ELCC — depends on how much storage is already on the system. When the grid has little storage, a new four-hour battery gets close to full capacity credit, because it's filling a wide-open evening gap. But as more four-hour batteries pile in, they collectively flatten that evening peak. Once the peak is shaved down to four hours wide, the next four-hour battery has nothing left to do at the margin — and its capacity credit falls.
Adding more four-hour batteries to a four-hour-saturated grid is like adding lanes that all merge at the same exit. At some point the bottleneck moves, and length — not width — is what relieves it.
This is no longer theoretical. California's storage fleet is now large enough that regulators have begun reducing the capacity credit assigned to incremental four-hour systems, and analysts increasingly frame the next reliability need in terms of longer duration rather than simply more megawatts. The research is consistent on the mechanism: storage ELCC declines with penetration, and that decline can be partially offset by adding longer-duration storage. In plain terms — once everyone has built the four-hour battery, the grid starts paying a premium for the fifth and sixth hours.
03The arbitrage case for longer
Capacity is half the story. The other half is the daily price spread, and here the trend also favors more hours — for a subtly different reason. As solar floods the middle of the day, midday prices are pushed toward (and sometimes below) zero, while evening prices stay high. That widens the spread a battery can harvest. It also lengthens it: the cheap-power window now stretches across most of the daylight hours, and the expensive window stretches across a long evening ramp.
A two-hour battery can only sip from that spread. A longer battery can drink. Texas is the live experiment: as evening price spreads widened through 2025, the average duration of operating ERCOT batteries crept up from about 1.5 to 1.65 hours, and developers began bringing four-hour systems toward completion. That's a market voting, in real time, for length — because the revenue is increasingly in the hours a short battery can't reach.
How much of the evening does each duration cover?
Illustrative — evening net-peak window ≈ 5 hours (roughly 4–9 PM)
A 4-hour battery covers most of the evening ramp; 6 hours covers the full window plus the shoulder. Beyond that, each added hour does less for the evening problem specifically — though it can reach into the early-morning peak.
04Why solar pairing changes the math
Everything above applies to a standalone battery deciding how many hours to buy. A battery paired with solar — a hybrid plant — solves the problem from a different starting point, and that nudges the optimal duration up.
Standalone batteries have to source their charge from the grid, paying whatever the market asks and competing with every other battery for the same cheap midday hours. A co-located solar array hands its battery a free, predictable charge every single day, peaking exactly when grid power is cheapest. That changes two things. First, the marginal cost of filling more hours of duration is lower, because the energy to fill them is already on site. Second, the value of those hours is higher, because the whole point of the hybrid is to take a midday solar surplus and deliver it as firm power across the entire evening — not just the first two hours of it.
Put differently: a solar-paired battery wants to be long enough to move a meaningful share of the day's generation into the evening peak. Too short, and you're spilling clean energy you already paid to build. The sweet spot — enough to shift the bulk of the evening ramp without overbuilding storage you can't fill — tends to land in the four-to-six-hour range.
For a hybrid plant, duration isn't a storage question. It's a question of how much of your solar you want to still be selling after sunset.
05The cost counterweight
If longer is better, why not build ten hours and be done with it? Because every added hour costs money, and the grid only pays for hours it actually needs. This is the discipline that keeps the answer bounded.
The encouraging news is that the cost ceiling is falling fast. Lazard's 2025 analysis put the levelized cost of storage at roughly $93/MWh, down from $104 in 2024 and $155 in 2023 — a decline driven by cheaper cells, higher energy density, and a global oversupply of battery manufacturing. BloombergNEF's benchmark for a four-hour storage project fell about 27% year-on-year. Falling per-kWh costs make each incremental hour of duration cheaper to add, which is precisely why the industry's center of gravity is drifting from two hours toward four, and from four toward six.
But the cost curve also explains the upper bound. Capacity value flattens past the evening window, while cost keeps accruing linearly with every hour. Somewhere around six hours, for most solar-paired systems on today's grids, the next hour stops paying for itself. That's not a permanent law — as penetration rises and the grid's needs lengthen, the optimal duration will keep creeping up — but it's where the arithmetic sits now.
06So why 4–6, and not 8?
Stack the four forces together and a band emerges rather than a single number. Capacity value says go past four hours, because the four-hour slot is saturating. Arbitrage says go longer, because the daily spread is widening and lengthening. Solar pairing says go longer still, because you have free midday charge and a full evening to sell into. And cost says stop before you're paying for hours the grid won't reward — which, today, means somewhere short of eight.
The intersection of those forces, for a solar-paired hybrid plant in a high-solar market, is four to six hours. It's long enough to firm the evening peak and bank a real share of the day's solar; short enough that every hour of storage still earns its place. It's not a number we inherited from the industry default — it's where the grid's actual physics and economics point, and it will keep moving as both evolve.
What it means for Solyx
We design our hybrid systems for the four-to-six-hour band because that's where a solar-paired battery does the most useful work: shifting midday solar across the full evening ramp, holding its capacity value as shorter systems saturate, and harvesting a widening daily spread — without overbuilding storage the grid won't pay for. Duration isn't a spec we copy from a competitor's data sheet. It's an outcome of matching the battery to the solar it's paired with and the grid it serves.