Electricity Storage is Getting Quite Cheap—But Firmer Storage Isn’t as Cheap as Sometimes Believed

Understanding India’s solar-plus-storage bids by considering what they don’t cover

In September 2025, a solar park tender in Madhya Pradesh cleared at ₹2.70 per unit (kilowatt-hour [kWh]) of electricity. Eight months later, Uttar Pradesh procured storage-backed evening supply at ₹6.45 per unit after accounting for a viability grant, which translates to roughly ₹6 per unit once solar is blended in for supply. Yet, this contract does not demand 100% assured supply but rather 95% availability, settled monthly. Both are recent, competitive Indian prices for “renewables plus storage.” What then explains the factor-of-two gap between them, within months of each other? It is the price of the obligation: what each tender actually requires the developer to deliver, in which hours, and how the shortfalls are counted. A related question follows: what risks are left behind?

That gap is what a forthcoming paper measures; some results are given in Key Insights: Storage is Getting Quite Cheap – But Measuring its System-level Cost is Complex. The claim that solar plus batteries now beats new coal for round-the-clock supply rests on comparing a tender price with a much more firm power cost, and the two are not fully comparable. The widely shared storage tender price is the cost of meeting a contractual obligation, which is lowered through a range of techniques. A blended system with solar, for a “4-hour battery” (for 10 hours supply = 6 hours solar plus 4 hours battery), will always be cheaper than 4 hours of supply from storage alone; the direct or “vanilla” solar portion only costs perhaps ₹2.5/kWh. Some bids cover only the battery hardware, priced like a monthly rental, and ignore the cost of charging the battery – batteries also need a “fuel”! Some bids assume two cycles per day, which lowers the per-unit cost. But how will you charge the battery a second time? Solar will only manage one cycle, at best.

Solar and storage are cheap and India needs far more of both, but current designs aren’t truly “firm” in output. The low ₹6.45 storage bid charged by solar doesn’t match output to demand. 

Standalone battery tariffs are down 86% in three years, which is a great achievement. But the differences in bid asks described above are one reason costs come in lower than what a firm system requires. For example, the Madhya Pradesh bid assumed two cycles of use per day, where the state has to charge the battery a second time overnight, at no cost to the developer.  In contrast, truly firm power is the cost of meeting demand in every hour, and this demand won’t be equal across the year. The two prices coincide only when the obligation is written and settled in hourly terms aligned with demand. Current solar obligations rely on annual capacity utilisation factors with tolerance bands, two-to-four-hour discharge windows, and monthly averaging of output.

A Bottom-Up Model to Understand the Portfolio-Level Costs of Storage

To find the number the tenders never ask for, we built a counterfactual: an optimisation model, built from first principles in GAMS using actual Indian data, that solves for the least-cost combination of solar, battery power (kW), and battery energy (hours of storage, translating to kWh) needed to meets a stated share of different demand shapes in every one of the year’s 8,760 hours. Battery physics, efficiency, depth of discharge, charging rates, degradation, and mid-life replacement are all represented explicitly.

A demand-flexibility engine then lets us relax that obligation the way real tender documents do: if a tender allows, say, a 10% relaxation (meaning it only asks for 90% output, that too on average), we model when that relaxation would occur.  Whatever demand is left unserved, we value it in two ways: at observed power exchange market prices and based on the cost of new conventional capacity operated at the utilisation rate and shape (profile) the storage’s leftover profile permits.

A subtle difference between current and firm tenders is the relaxation most present bids allow: they ask for a semi-firm output, e.g., 90% of the output, but apply penalties after averaging output over time. Thus, a 90% output system could, in theory, supply 100% for much of the year while providing zero output for over a month.

Simple spreadsheet arithmetic on today’s battery rentals produces “24×7 for ₹3.74 per unit,” including the solar needed to charge the battery. But this is only the cost of amortising the hardware plus notional solar input.  Counting the costs that are already the developer’s contractual burden, like losses from round-trip efficiency, moves the floor to ₹5.75, based on first-principles modelling and aggressive capex prices, close to recent bids.

A spreadsheet shows 24×7 solar-plus-storage as low as ₹3.74/unit based on recent bids. But modelling India’s real hourly demand and solar, true firmness can vary from ₹8.30–11.82.

We go further. Our analysis focuses on what these linear extrapolations from bid prices leave out—the variability and shape of both supply and demand. Neither is the demand flat, nor is supply equal each day.  Because of this, one has to oversize both the battery and the solar capacity compared to an ideal (linear) extrapolation from today’s bids. This can only be done via an optimisation model.  Adding what a spreadsheet cannot represent, namely, time-of-day matching against India’s actual demand shape, real day-to-day and seasonal solar variability, and feasible charging, pushes the cost of truly firm supply to ₹8.30 with a good solar site.  Instead, using average national solar profiles pushes this up further to ₹9.07/kWh. Sizing the system to hold up across more difficult years (poor solar output and/or high demand), which a 25-year contract implicitly promises, can raise the cost for a national-average source to ₹11.82. Nothing in that climb is a technology cost. It is assumptions being replaced by the Indian reality as it actually occurs.

Isn’t Partial Cheap Decarbonisation Good Enough?

Most RE developers and supporters agree that 100% RE systems are expensive, and so they naturally ask: why not decarbonise, say, 80%, which is substantial and much cheaper? The problem is that it isn’t really cheaper. We use 24×7 output as a long-term objective that can be met in stages. Early decarbonisation is the low-hanging fruit India should aggressively pursue, but it’s not the same as 100% decarbonisation, even in the power sector.

Consider two systems with an identical 95% annual capacity utilisation factor (CUF, also called plant load factor, or PLF), indistinguishable on every metric a tender settles against. Require one to supply at least 90% of demand coverage in every (say, hourly) time block as a floor, and the price comes to ₹7.88. Let the other skip its worst 5% of blocks, and the price falls to just ₹6.41, an 18.7% discount. Both offer the same total output. Where does the discount come from? The exempted hours cluster exactly where supply is hardest: monthly CUF collapses to 74% in August and 83% in July, and the battery’s utilisation drops. The cheaper asset is doing less work in the hours that matter and handing what can be the worst part back to the distribution company buying the power.

That handed-back remainder is the expensive leftover of our study’s title. Valued at 2023 exchange prices, it costs ₹8.28–8.79 per unit, more than the project’s own levelized cost of energy (LCOE). We’re now seeing far worse price spreads on the power exchange, approaching what is called a necklace curve: prices at the ceiling for most of the day, except midday, coming down in an inverted U shape or necklace thanks to solar.

Say we only want 95% annual output (cheaper than 100%, but more than current bids offer): supplying every hour @95% costs ₹7.88; skipping the worst 5% of hours, when demand is high and solar low, costs ₹6.41.

If we instead consider new alternative generation for covering the residual unserved demand, perhaps because the market isn’t deep enough for the volumes required by a high-RE system, the cost is even higher.  The relaxed design leaves a residual so bursty that a dedicated coal plant would run at just 4.7% PLF and cost ₹51.6 per unit; even open-cycle gas would cost ₹36. Add the project and residual costs together for a system or portfolio cost, and the ranking inverts: the “cheap” flexible storage-based procurement produces a portfolio cost of ₹8.01–8.91 per unit against ₹6.51–7.88 for its firm twin, at every coverage level, unless cheap, flexible gas is available. The right level of firmness to demand in a tender depends on what will balance it. No Indian tender document currently asks that question.

Two popular shortcuts for firm power fail along the way. Replicating the cheap two-hour bids to “reach 24 hours” adds power, not duration: stacked units sit underutilised, over-deployment grows, and the surplus ends as curtailment that today’s grid may absorb, but a future high-RE grid will not value. Stated another way, today’s storage bids can be utilised almost in full, but as storage capacity grows, these units will also face diminishing value, analogous to solar’s diminishing value today.  And coal’s 85% availability is not comparable to an 85%-coverage solar-storage product: coal schedules its maintenance outages and dispatches on demand, whereas renewable shortfalls are correlated, seasonal, and cannot be summoned. A separate study on coal’s availability is underway at CSEP.

DISCOM would need to buy expensive power to cover the skipped hours at ₹8.28–8.79 on the exchange (historical prices), making the “cheap” contract costlier overall.

None of this is an argument against renewables or storage; India needs far more of both. The argument is about visibility. The roughly ₹1.5 per unit (across all units!) that relaxed obligations shave off a 95%-coverage bid does not disappear; it reappears as monsoon procurement at exchange prices, as stranded-cost peaking plant, or as load shedding. We should not compare tender prices with coal tariffs without adjusting for the obligation behind each. We should write and settle firmness in time blocks, because averaging is the transfer mechanism. We should procure and price the residual explicitly instead of discovering it in the worst hours and treat portfolio-level analysis as a precondition of tender design. At the minimum, upcoming storage + RE should aim for some level of hourly firmness; future bids can tighten the requirement.

Our study includes analysis of India’s standalone BESS and solar-plus-storage tenders, the full cost ladder, twenty-five years of solar-demand pairings, the firm-versus-flexible twin experiment, and the residual valuations. India is not choosing between cheap power and expensive power. It is choosing where the cost of reliability gets recorded, and who pays for it. This study shows how to read that choice off the tender document itself.

We recognise that this is a supply-side design study, not grid planning. The model excludes transmission, reserves, stability services, unit commitment and inter-state coordination, and it does not represent the rest of the grid, especially the existing coal fleet. This cuts both ways: the coal fleet may hold slack capacity that could serve part of the leftover demand, though rarely in the stress hours that matter, while additional midday solar pushes those same plants toward their technical minimum, which for older units sits above the notified 55% in practice today. Nor do we model demand-side response, which is a genuine alternative to some of the firmness priced here. Most public storage analyses share this same supply-side lens but assume a flat, year-round duty cycle—the very assumption this paper tests and rejects. Readers should therefore treat every cost in the study the way we do: not as a forecast of system outcomes, but as a lower bound on the true cost of firmness.