How Many Batteries Can France Add Before Batteries Destroy Their Own Profits?

A battery can make money by buying electricity when it is cheap and selling it when it is expensive. But that business has an unusual weakness: when enough batteries follow the same strategy, they make the cheap hours less cheap and the expensive hours less expensive. The electricity system becomes more flexible while the next battery has less of the price difference left to earn.

France makes this tension unusually visible. Solar generation increasingly pushes electricity prices down during the middle of the day, yet evening demand can still produce higher prices. RTE counted 513 hours with negative average hourly wholesale prices in 2025, up from 352 in 2024, against an annual average price of €61.1/MWh. More than half of those negative hours were only marginally below zero, so the count alone does not tell an investor how profitable charging would have been. [1]

The more recent market narrative is tempting. An industry assessment reported by pv magazine France on 1 October described over 800 hours with prices at or below zero so far in 2026 and a September average daily high-to-low spread of about €214/MWh. Those statistics are not directly comparable with RTE's strictly negative 2025-hour count, and a daily peak-to-trough difference is not a spread that a real battery can necessarily capture. Still, they explain why storage is attracting attention. At the end of 2025, RTE reported about 1.6 GW of operational battery capacity and around 14 GW of projects with network access. A grid-access position is neither a construction commitment nor a forecast that every project will trade identically. [2,6]

The investment question is therefore not whether France sometimes has very cheap electricity and sometimes has expensive electricity. It is how much of that trading opportunity remains after additional storage has been built to exploit it.

One battery sees a price; a fleet helps set it

For a small battery, it is often reasonable to treat wholesale prices as given. It charges during low-priced hours, loses some energy through the storage process, and discharges when the sales price is high enough to cover both its electricity purchase and operating costs. The revenue that remains must ultimately contribute to construction, financing, maintenance and equipment replacement. A striking market spread is only the start of that calculation.

Now suppose hundreds of similar batteries act together. Their charging increases demand in the cheapest hours, tending to lift the trough. Their discharging adds supply in the most expensive hours, tending to lower the peak. They can also choose to cycle less often when the difference is no longer worth exploiting. In other words, storage competes away part of its own merchant revenue as it successfully reduces the system's need for extreme price movements.

That is not an argument that batteries have no value. Smoothing electricity prices can benefit consumers and improve the use of existing generation. The difficulty is who receives the benefit. A fall in price volatility may be useful to the electricity system without appearing as cash income for the marginal battery owner. Capacity payments, balancing services and network benefits can provide other sources of value, but each has its own availability requirements and competition.

Testing the feedback without pretending to forecast France

The full French investment threshold cannot be read from the public summary statistics above. It would require rights-cleared interval-price observations and an estimate of how the French market would reprice as a fleet charges and discharges, while hydro, nuclear generation, interconnectors and other participants react. To show how that feedback works, we instead built a transparent, deliberately illustrative market experiment.

The reference is a constructed year of 8,760 hourly electricity prices. It matches RTE's published 2025 annual price mean and selected negative-price statistics but is not an actual EPEX price history and has not been validated against realised French battery dispatch. In the experiment, an optimised 1 MW / 2 MWh battery operates with 86% round-trip efficiency, a maximum of 1.25 full cycles per day and an assumed €5/MWh cost on both charging and discharging throughput. The price-response rule is equally explicit: a further gigawatt of identical batteries charging or discharging at full output shifts the relevant clearing price by €5/MWh. That sensitivity is an assumption, not a French bid-stack estimate. [1,7,8]

We then add identical merchant battery capacity in stages and let the fleet adjust its trading schedule. This produces an internally consistent illustration of the feedback between dispatch and price, rather than assuming every new battery trades against the original unchanged price curve. It does not convert France's roughly 14 GW connection queue into a prediction of 14 GW of new merchant storage.

The first batteries earn the spread; later ones inherit a smaller one

In the constructed reference year, the two-hour battery earns about €34,648 per MW-year of net day-ahead arbitrage margin before fixed operating costs and capital charges, with no additional identical competing fleet. After another 1 GW is introduced, that falls to roughly €27,467. At 4 GW of additions the result is €15,542, or 55% below the starting margin. At 14 GW, only about €2,640 remains. These are outputs of the assumed market and price-response model, not historical French earnings or forecasts.

Figure 1. Illustrative net day-ahead trading margin per MW-year as identical capacity is added. The horizontal axis shows incremental modelled merchant capacity, not France's installed fleet or an expected build-out. The price-impact parameter is assumed at €5/MWh per additional GW at full output.

The pattern is not simply a fixed percentage cut for every new gigawatt. As prices converge, the optimised batteries stop cycling in hours that no longer cover their losses and variable costs. Equivalent annual full cycles decline from about 429 at zero new GW to 328 after 4 GW. That adjustment cushions the fall in trading margin, but it cannot restore the original opportunity. A four-hour battery initially earns more wholesale margin than the two-hour design in this same experiment; it also requires additional energy capacity and does not escape saturation indefinitely.

This is the modelling mistake that matters for portfolios. A developer can make a plausible revenue forecast for one project using a fixed wholesale price curve. It does not follow that a portfolio of projects can all earn that forecast after their combined operation has changed the curve. The next step is to ask how much revenue a new investment needs, which is a different question from the direction of the arbitrage line.

The investment threshold is not the arbitrage threshold

Consider a hypothetical two-hour project costing €500,000 per MW upfront, with €10,000/MW of annual fixed costs, a €30,000/MW augmentation outlay in year eight and a 15-year operating horizon. At an assumed 8% real discount rate, the day-ahead margin above is not sufficient by itself to establish a positive net present value. This is a constant-real, unlevered comparison, not a quotation of French EPC costs or a model of an identified project.

Now add a hypothetical annual payment for providing dependable capacity. At €20,000 per MW-year, the project starts at approximately minus €134,000 NPV per MW, even before we introduce additional merchant competitors. At €40,000 per MW-year, its starting NPV becomes approximately plus €37,000/MW and crosses zero at around 0.6 GW of additional identical batteries in this constructed scenario. Reduce the initial capital cost to €400,000/MW with that same assumed €40,000 payment, and the crossing moves to about 3.1 GW.

Those crossing points are not estimates of how many more gigawatts France can economically support. Neither availability payment is a reported auction award. The useful finding is the contrast: a project can fail the investment test before cannibalisation becomes severe, or remain investable through more competition if costs are lower and additional income is dependable. Arbitrage, total project cash flow and social value should not be treated as one number.

What a French investment decision would still need

France's capacity mechanism illustrates why the rest of the revenue stack matters. On 9 October 2026, regulator CRE published parameters for a 2030-31 capacity auction. The framework can remunerate qualifying dependable availability, including storage, but a battery's nameplate MW are not automatically certified winter capacity, and qualification does not guarantee a particular auction price. Balancing, ancillary services and contracts with renewable generators can also contribute income. Those services require energy or dispatch headroom, so it would be wrong to stack each hypothetical revenue stream on top of the maximum achievable day-ahead trade without allowing for operational conflicts. [3,4]

The most consequential missing variable remains the actual French price response to coordinated charging and discharging. It depends on where batteries connect, the timing of renewable curtailment, hydro flexibility, nuclear output, imports, exports and competing industrial demand. France's shift to 15-minute day-ahead trading in October 2025 also means a proper empirical study must work at a finer resolution than our illustrative hourly year. A defensible investment conclusion would need those market data, credible storage cohort assumptions and commercially grounded capex and revenue contracts. [1,5,9,10]

So how many batteries can France add before they destroy their own profits? There is no single credible national GW threshold in the evidence assembled here. Our controlled experiment shows how self-cannibalisation can be large, with modelled two-hour arbitrage falling by more than half at 4 GW of added identical capacity. The investment break-even, however, moves with the cost of the battery, the price feedback it causes and the value of services it can reliably sell alongside energy. For investors, the error to avoid is simpler than the answer: do not underwrite an entire storage build-out as if every project will inherit the spreads that attracted the first one.

Research note. The interval prices, fleet price effects, arbitrage values, financial examples and break-even crossings in this article are constructed scenarios, not measured French merchant-battery returns or forecasts. Official sources establish selected market and fleet statistics; the technical report documents the assumptions, optimisation, sensitivities and evidence needed for an empirical extension.

Explore the research

The accompanying technical report, editable workbook and reproducible model engine explain how the synthetic prices were constructed, how competing batteries change the modelled clearing price and how project economics vary with capital cost, storage duration and assumed capacity income. Public links to these materials belong with the approved Rokin article.

Source notes

1. RTE, 2025 annual electricity review: spot prices and negative hours. https://analysesetdonnees.rte-france.com/bilan-electrique-2025/prix

2. RTE, 2025 annual electricity review: generation and battery storage. https://analysesetdonnees.rte-france.com/bilan-electrique-2025/production

3. CRE, Parameters for the long-term 2030-31 capacity auction, 9 October 2026. https://www.cre.fr/actualites/toute-lactualite/la-commission-de-regulation-de-lenergie-publie-ses-deliberations-relatives-au-parametrage-de-lenchere-de-long-terme-2030-2031-du-mecanisme-de-capacite.html

4. RTE, Capacity mechanism service portal: principles and certification. https://www.services-rte.eu/en/learn-more-about-our-services/participate-in-the-new-capacity-mechanism.html

5. CRE, Negative-price and support-reference data, updated October 2026. https://www.cre.fr/documents/open-data/donnees-necessaires-au-calcul-des-complements-de-remuneration.html

6. pv magazine France, reporting a Storio market assessment, 1 October 2026. https://www.pv-magazine.fr/2026/10/01/la-france-franchit-le-cap-des-800-heures-a-prix-de-lelectricite-negatifs-ou-nuls/

7. National Laboratory of the Rockies / NREL, Annual Technology Baseline 2025: utility-scale battery storage. https://atb.nlr.gov/electricity/2025/utility-scale_battery_storage

8. National Laboratory of the Rockies, Cost Projections for Utility-Scale Battery Storage: 2025 Update. https://research-hub.nlr.gov/en/publications/cost-projections-for-utility-scale-battery-storage-2025-update/

9. Fraunhofer Energy-Charts, API and market-data documentation. https://api.energy-charts.info/

10. RTE, Electricity market prices and conditions of use. https://analysesetdonnees.rte-france.com/prix/evolution-prix