Why GB grid balancing costs hit £2.5bn — and where the money actually went

The number that did the rounds in March was £2,512 million. That is what the National Grid Electricity System Operator (ESO) settled in Balancing Services Use of System (BSUoS) and constraint-management payments across the 2024–25 financial year — up from £1.97 bn the year before, and roughly £200 m above ESO's own December forecast.

Ofgem's first reaction was a terse acknowledgement that "system-management costs are running materially above the trajectory implied by the RIIO-ESO settlement." Octopus chief executive Greg Jackson put it less politely on LinkedIn: "We are paying wind farms not to spin and gas plants to spin instead, in the same half-hour, across the Scottish border."

Both readings are correct. Neither is a full explanation. Here is where the £2.5 bn actually went.

Constraint payments: 47% of the total

The single biggest line item — about £1.18 bn — was constraint management. That is what ESO pays when it cannot move power from where it is generated to where it is needed, because the transmission network does not have the capacity. In GB the bottleneck is overwhelmingly north-to-south: Scottish wind is plentiful and cheap, English demand is bigger, and the high-voltage corridor between them is only being upgraded now.

When the corridor saturates, ESO does two things in the same half-hour. It pays a Scottish wind farm to curtail — to stop generating — typically at something close to the Contracts for Difference strike price plus loss-of-revenue compensation. And it pays a gas plant in the south to start up instead, often through a Balancing Mechanism bid that prices in start-up costs and minimum-run constraints.

The arithmetic of that swap is unflattering. SSE's Beatrice offshore wind farm was curtailed for roughly 6.4% of its potential output in 2025. The compensating gas was sometimes priced above £200/MWh in the same window. The difference — paid by consumers via BSUoS — is the headline waste.

Frequency response: 22%

The next big bucket, about £550 m, was frequency response. The grid runs at 50.00 Hz; physics requires that supply and demand balance second-by-second, or frequency drifts, and at extreme deviations protection relays start tripping plant offline.

Historically frequency response came from spinning gas turbines kept part-loaded. Increasingly it comes from batteries, which respond in under 200 ms — faster than any thermal asset and crucially without the carbon. The newer Dynamic Containment (DC) and Dynamic Regulation (DR) markets, introduced in 2020 and expanded through 2024, are now dominated by battery operators. Zenobē, Gresham House Energy Storage, and Harmony Energy collectively cleared more than 1.6 GW of DC last year at average prices of £8–15/MW/h.

That sounds expensive. Compared to what it replaced — gas turbines burning fuel just to be available — it is roughly 35% cheaper per MW of response capability, per ESO's own Frequency Response Market Review. The £550 m is not a problem to be solved so much as a market doing its job.

Reserve and reactive power: 18%

About £450 m went on operating reserve (warm capacity available within minutes if a large plant trips) and reactive power services (voltage support, especially as transmission lines load up).

Reactive power has become quietly expensive. Older synchronous condensers — essentially big spinning machines that absorb or supply reactive power — are being retired faster than they are being replaced, and the alternative procurement via voltage-support contracts has not kept pace with prices. Statera Energy and Drax both signed new long-duration reactive contracts in late 2025, but the inflation in that line is material.

Black-start and inertia: 8%

About £200 m bought ESO the ability to restart the grid from cold if a national blackout occurred, and to maintain enough rotational inertia in the system that frequency cannot collapse before action can be taken. As gas retires, both costs go up; battery-based black-start pilots, including the EFCC trial with Drax and Statkraft at Cruachan, are credible but not yet cleared as headline procurement.

The rest

The residual ~£130 m is a mix of trading losses, system testing, and the recovery of legacy contract liabilities. None of it is interesting except in aggregate, and even there the trend is benign.

So is it £2.5bn of waste?

No. About £1.2 bn of it is a transmission-build problem — the cost of running an under-built network while the Eastern Green Link HVDC bootstraps and the SSEN/SPT Beauly–Denny reinforcement come online. Once those land in 2028–30, constraint payments should drop sharply.

The other ~£1.3 bn is mostly the price of buying ancillary services from an increasingly low-carbon stack, and it is broadly priced fairly relative to its (very real) value to system security.

The honest critique is not that the money is wasted. It is that the cost of waiting for transmission is being socialised onto consumer bills today, while the upside — cheaper Scottish wind, displaced gas — accrues quietly tomorrow. Which is exactly the sort of intertemporal trade-off energy policy has been bad at communicating, and which the Citizens Advice Energy Bills Support team keeps quietly raising at Ofgem stakeholder meetings.

If you want to compress this to a sentence: balancing costs are high because the network is too small for the generation we built, and the bill will fall when the network catches up.

Where to watch next

Three numbers will move in the next twelve months:

1. Eastern Green Link 1 energisation (target Q4 2027) — if it slips, constraint payments stay sticky around £1.0–1.2 bn for another year.
2. The Strategic Spatial Energy Plan (NESO consultation closes June 2026) — the proposed regional pricing zones would change the constraint accounting fundamentally.
3. Battery share in Dynamic Containment — already 71% in March; if it hits 85% by year-end, frequency-response costs fall another 8–10% in nominal terms.

None of these are speculative. They are line items with known timelines, and they are the lens through which to read the next ESO annual report.