Fixed demand, unlimited supply

The standard account of the Swedish FCR-D market goes like this. Prices were extraordinary in 2022 and 2023. Batteries rushed in. The market saturated, prices fell, and operators now need to diversify into aFRR, mFRR and energy arbitrage. Every market commentary published in the last eighteen months says a version of this, and none of it is wrong.

It is also not the interesting part. Saturation implies a market that filled up, which implies it could empty again. That framing is what keeps ancillary revenue lines flat or gently recovering in BESS financial models three and four years out. The framing is wrong, and the reason it is wrong sits in plain sight in Svenska kraftnät's own dimensioning method.

Demand is a physical constant

FCR-D upward exists to hold system frequency when the largest single unit in the Nordic synchronous area trips. That unit is Oskarshamn 3, at 1 450 MW. The Nordic requirement is dimensioned against it. Sweden takes a share determined by a key factor based on generation and consumption, which moved from 37.4 percent to 37.75 percent for 2026. FCR-D downward is dimensioned the same way against full export on Nordlink or North Sea Link, a reference incident of 1 400 MW.

Svenska kraftnät states plainly that it forecasts no change to either reference incident before 2030.

So the demand curve is not a curve. It is a horizontal line at roughly 550 MW per direction, drawn by a physical fact about a nuclear reactor and a subsea cable, and it does not respond to price, to electrification, or to how many batteries get built. No amount of demand growth elsewhere in the power system moves it. Load can double and the FCR-D requirement stays where it is.

A market where demand cannot grow is not a market that saturates. It is a market that divides.

Supply has no such constraint

Against that flat line, here is what has been prequalified. All figures are Svenska kraftnät's, as of Q4 2025.

Source: Svenska kraftnät, Balancing Market Outlook 2030, June 2026 update, data to December 2025. Requirement marks are the Swedish volume requirement or, for FFR, the annual procured volume of 113 MW. aFRR upward is shown at the maximum national hourly demand of 97 MW. mFRR is omitted because at 17 010 MW prequalified it does not fit the scale.

Batteries alone hold 2 710 MW of prequalified FCR-D upward capacity. The entire Swedish requirement is around 550 MW. Ignore hydro completely, ignore demand response, ignore every other technology on the register, and battery capacity by itself is close to five times the whole market.

One caveat that matters, and it cuts against the starkest reading of this chart. Prequalified capacity is not the same as capacity that bids. A great deal of the hydro and mFRR volume on the register is production capacity that participates opportunistically, when spot conditions make the opportunity cost low. Those megawatts are not permanently pressing on the price.

Battery megawatts are. A grid-scale battery has no alternative use for its capacity and close to zero opportunity cost in offering it. Svenska kraftnät makes this point itself, noting that storage and other new resources show less dependency on spot price, so supply volumes and pricing stay stable even in hours when low spot prices and thin dispatchable availability used to lift the clearing price. That is precisely the mechanism that used to produce the price spikes. It has been structurally removed.

The exit is smaller than the room

The consensus response to FCR-D compression is diversification into aFRR, mFRR and energy trading. Two of those three have the same architecture as the market being escaped.

• aFRR maximum national hourly demand currently sits at 97 MW upward and 124 MW downward. Svenska kraftnät's own estimate for future demand ranges from 120 to 350 MW. That is a market an order of magnitude smaller than FCR-D, against 2 630 MW already prequalified upward.
• mFRR capacity market maximum volume is 1 300 MW upward against 17 010 MW prequalified. It is the largest of the reserve markets and still a fixed volume set by the TSO.

Both are dimensioned on system need, not on how many assets want in. The escape route from a capped-demand market is two more capped-demand markets, one of which is considerably smaller, and the batteries arriving there are already prequalified and ready to bid.

Energy arbitrage is different in kind. It is the one revenue line whose size scales with volatility rather than with a TSO's dimensioning decision. It is also the one that requires trading capability rather than asset availability, which is a different business with different people in it. That distinction is where the sector is actually heading, and it is under-appreciated in how BESS teams are staffed.

What this means for underwriting

The practical conclusion is narrow and I think defensible.

Any BESS model that carries an ancillary services revenue line flat or recovering past 2027 is wrong by construction rather than by forecast error. The error is not an optimistic price assumption. It is a category mistake: treating a fixed-volume procurement as though it had a demand curve. Prices in a capped market with growing zero-opportunity-cost supply have one direction available to them until either the reference incident changes or entry stops. Svenska kraftnät forecasts no change to the first. Nothing in the buildout data suggests the second.

There is one live variable worth watching, and it arrives shortly. From September 2026 the Nordic TSOs introduce a limit on how much static FCR-D upward may be procured, initially set at 50 percent of the total Nordic requirement, with a dynamic component determined hourly on the share of rotational energy in the system. That reshuffles which providers can access the volume. It does not change how much volume exists.

Why this matters beyond storage

The same arithmetic reaches into charging infrastructure, which is where I work.

The sites with genuine charging demand sit in dense urban and corridor locations where the local grid is already constrained. Connection capacity is the binding limit and effekttariffer make the peaks expensive. Storage solves both problems technically. Whether it solves them commercially has depended on stacking grid-service income on top of peak shaving, because peak shaving alone rarely carries the capex at current integration costs.

That stack is thinner every quarter, and for structural reasons rather than cyclical ones. The consequence is a growing set of charging sites that are technically viable and commercially marginal. Underwriting them on a recovering FCR-D line does not fix that. It just moves the moment someone notices.