The subsidy that slowed the transition

How the Renewables Obligation became a brake on the electrification it was meant to enable.

Take a trip back to 2018; we can see which elements of the energy transition have exceeded (or been exceeded) by expectations. We can do this using the FES (Future Energy Scenarios) projections, the annual scenario set National Grid ESO (now NESO) has been publishing since 2011.

The forecast error you can see from space is the overprediction of demand; something I’ve written about before. This is nowhere less evident than heat, where successive governments have baulked at gas-boiler bans, preferred the purity of small volumes of heavily subsidised pure heatpump retrofit over more flexible and scalable hybrid models. Even challenges like fixing EPC (Energy Performance Certificates) and smart metering/settlement have dragged on years and years.

These failures have broader consequences. Much of the investment we’ve made into renewable generation would have paid off better; had we seen the growth in demand that policy makers projected at the time the decisions to invest in renewables were made. Most of the capital costs of renewables are fixed; and much of the recovery is from volume based energy charges, so in the same way that a growing economy provides a tax base to support public investment, so a growing demand for electricity could have diluted the burden of the energy transition. Furthermore; demand growth in Scotland could have directly avoided much of the curtailment costs.

One small upside is the flattening of demand since 2024 — now nearly two years of broadly stable consumption — but consistent growth remains to be seen.

It’s worth being explicit about the feedback loop this creates. Most of the costs of the GB grid — the renewable subsidies, the network capex to connect them, the standing cost of the gas plants that back them up, the curtailment bill — are essentially fixed each year. They get amortised across whatever electricity the country actually uses. When demand sits ~25% below where 2018 policy expected, those fixed costs land that much heavier on every remaining MWh. And the higher the per-MWh cost gets, the harder it is for electricity to undercut gas at the boiler, the petrol pump or the factory floor — so the electrification growth that would have diluted the burden stalls further. The 2018 demand miss isn’t just a sunk cost; it’s a feedback loop that’s still tightening today.

You can see policy reaching for the sticking plaster. The Boiler Upgrade Scheme now pays £7,500 to replace a gas boiler with a heat pump (rising to £9,000 for off-gas-grid oil-heated homes) — a capital subsidy designed to bridge an operating-cost gap that wouldn’t exist if electricity weren’t loaded with policy costs in the first place. Bigger upfront grants don’t fix the loop; they just smooth its visible edges, and the smoothing itself becomes another bill borne by the same taxpayers we’re nominally trying to relieve.

To be fair, the BUS period hasn’t been pure waste. Nesta’s analysis of MCS install data shows the median air-source heat pump install fell about 11% in real terms over the four years following BUS launch in 2022 — roughly 3%/yr, driven by a mix of installer scale, supply-chain maturity and competitive entry (Octopus, British Gas, and others pushing down headline prices). The trend has stalled in the most recent year (a 0.4% real-terms uplift), which suggests the easy gains have been captured. It’s a meaningful learning curve but vastly slower than the technologies heat pumps depend on: solar PV fell ~15%/yr through the 2010s (~80% per decade), and lithium-ion battery packs have fallen by more than two-thirds since 2015 as the EV and grid-storage markets actually began to scale — ~12-13%/yr CAGR and still tracking ~10%/yr. Heat pumps are running at maybe a fifth of that rate. That’s because the cost stack is mostly labour and on-site work — installer time, plumbing modifications, hot-water cylinder, radiator upsizing — not a commoditised product where global manufacturing scale compresses unit cost. The heat pump unit itself might be getting cheaper, but it’s only ~⅓ of the all-in install bill; the rest is local, skilled labour, and that’s what bounds the scope for cost reductions.

Interest rates compound the problem from the other direction. The MCS Data Dashboard puts the average UK air-source heat pump install at ~£13,300 in 2025; after the £7,500 BUS grant, the household is left at ~£5,800 — versus a like-for-like gas boiler replacement at around £3,000. If that money is financed at a mortgage rate of 5%, even significant energy bill savings could be swallowed up by higher interest and capital payments.

Carbon policy bakes the asymmetry in even more deeply. The gas burned in a CCGT to generate electricity carries two carbon levies — the UK ETS (~£40-50/tCO2 of late) and the Carbon Price Support top-up (~£18/tCO2, frozen since 2015). Both flow through CCGT short-run marginal cost into the wholesale electricity price and onto consumer bills. The same gas burned in a domestic boiler down the road sits outside both regimes and pays no equivalent carbon price. So the carbon in a kWh of electricity (already mostly renewable) carries a price tag that the carbon in a kWh of heating gas does not — and the higher the marginal CCGT runs in the merit order, the wider the wedge. Recent policy nibbles at the edge — the Autumn 2024 Budget confirmed the CPS top-up will be abolished from April 2028, lifting ~£18/tCO2 off the power side — but the bigger ETS levy remains, and the change does nothing for the underlying asymmetry between gas-for-electricity and gas-for-heat. And if the UK relinks with the EU ETS (where carbon has been trading much higher than the UK price), the wedge against electricity widens further, exactly the wrong direction if the goal is to make electrification competitive. The structural fix would be a carbon tax on heating gas, of the kind Ireland has progressively introduced (€56/tCO2 by 2024, with a glide path to €100/tCO2 by 2030). In the UK that idea has been politically toxic every time it’s been seriously proposed.

There’s a second, subtler effect worth flagging. Because CCGTs typically set the price in GB wholesale markets, those carbon levies on power-sector gas flow directly into the wholesale clearing price — roughly £17/MWh from the ETS plus ~£7/MWh from CPS at recent levels, or about £24/MWh of carbon-driven wholesale inflation that wouldn’t be there in an unlevied counterfactual. CfDs settle against that inflated wholesale, so the headline top-up cost (the LCCC bill we see reported as “CfD support to renewables”) looks artificially small. A £45/MWh CfD strike against an ETS-inflated £65/MWh wholesale looks like the consumer is making money from the contract; against the ~£40/MWh wholesale you’d get without the carbon adders, the contract is roughly a wash. The cost hasn’t disappeared — it has just moved from the visible LCCC line to the less-visible wholesale-driven retail tariff. Both are paid by the same consumer; the optics flatter renewables policy.

Political emphasis amplifies the loop too. Clean Power 2030 — the headline target to fully decarbonise GB electricity by the end of the decade — concentrates yet more policy bandwidth, capex and consumer levies onto the sector that’s already by far the cleanest. Domestic heat (~78% still gas at the meter) and transport (majority ICE on the road) each emit more annual CO2 than power, but sit largely outside the CP30 framework. Doubling down on the cleanest sector while demand stalls makes the per-MWh maths worse, not better — and pushes the cheapest remaining abatement (heat and transport electrification, exactly the demand FES 2018 expected to materialise) further out of reach.

The deepest irony sits here. The Renewables Obligation — designed to decarbonise electricity — has, via the per-MWh premium it bakes into every consumer kilowatt-hour, quietly become a brake on the electrification needed to decarbonise heat and transport. The bill lands on households and small businesses; the receipts flow to the institutional investors (pension funds, infrastructure funds, listed utilities) who own most of the older offshore wind portfolios. A subsidy designed to accelerate the transition has, on net, slowed it. The same dynamic falls on every renewable, network upgrade and storage project we’re trying to amortise into a stalled demand base: the cost of capital has effectively doubled since the policy designs that underpin this build-out were drafted in the late 2010s, which flows through to higher CfD strike prices in recent rounds and higher TNUoS charges down the line. The vicious circle pulls in financing too: low demand keeps per-MWh costs high, which keeps electrification slow, which keeps demand low, while the cost of capital to escape any of it has roughly trebled in nominal terms.

Projections for generation technology have by and large been better than demand; though this is not that surprising given Government has more or less directly controlled the growth of the technologies through control of the CfD budget allocations and more recently limiting the growth of grid connections for new battery projects. Only PV, much of which is installed behind the meter and without either a Government CfD could be seen as a less regulated market; and the growth here has been the most impressive, though much of the returns on behind the meter solar investment actually stem from the high levies, meaning it actually exacerbates the problem of high electricity prices.

Batteries are the outright winner; and PV follows close behind. Wind, the bedrock of our system, has tracked close to the bottom of the FES 2018 range in capacity terms — broadly in line with Steady Progression, well below the Two Degrees and Community Renewables paths the policy debate at the time tended to assume.

It’s not just demand and capacity the projections got wrong — fossil costs did too. The 2021–23 gas crisis sent wholesale gas to 3–4× the FES High case, and the UK ETS launched in 2021 already running well above the EU ETS path FES had penciled in. Both fuel and carbon went far beyond the modelled range, which dramatically changed the economics of every CfD struck on the back of those projections.

How gas and wind don’t mix

A one sentence summary of 2010s electricity policy was to retire coal, replace it with wind power, retain enough gas as a backup and attempt a half-arsed gold-plated revival of nuclear power.

By providing a fixed CfD price for wind electricity and a capacity market payment for gas, the idea was that the cost of capital (finance) to duplicate our generation could be minimised. In the short-term this worked because the economy was growing (just), there was a cross-party political consensus RE decarbonisation, interest-rates were at record lows and the prospect of massive growth in demand from electrification in the future meant that the cost to consumers could be deferred. Unrealistic expectations about the speed of transmission infrastructure build out played a supporting role too.

Why gas is a growing cost for wind generation

You wouldn’t normally think of wind generation requiring lots of natural gas. But, the way the GB electricity market is structured, that’s increasingly what happens for the grid operator, NESO.

One silver lining of wind capacity added in recent years and the fall in costs that ended with the great inflation of 2022 has been the reduction in the cost of compensating curtailed wind farms. Offshore CfD strike prices fell from ~£120/MWh in AR1 to ~£37/MWh by AR4 before the 2022 inflation reversed the trend (AR7 cleared at ~£91/MWh). This has created the strange dynamic where the most recent wind farms, with cheaper CfD prices tend to be curtailed more than older vintages. You can see this trend clearly with the outward shifts of the wind bid stack in the animated chart below:

Plus the subsidy bill on top

The bid stack shows what the system pays to balance wind. But consumers pay a separate, much larger bill: the RO and CfD subsidies on every MWh delivered. Recent political debate has fixated on CfD strike prices — they’re public, banded by allocation round, and easy to cite. But CfDs still are the smaller scheme. The chart below splits the two, and RO (in purple) dominates throughout — roughly 90% of the subsidy bill in the late-2010s, falling to ~70% by 2024-25 as more of the fleet rolls onto CfDs. Offshore wind accredited before 2015 was banded at 2 ROCs/MWh (≈£145/MWh at today’s ~£73 ROC value); onshore is 0.9–1.0 ROCs (≈£65–73/MWh). Both rates lock in at accreditation and run for 20 years, RPI-indexed all the way through to the final April 2026 uplift — so the early offshore fleet earns that £145/MWh on top of wholesale into the late 2030s, regardless of where the spot price sits. CfDs work differently, topping up the difference between a fixed strike and the market reference, so they’re at least somewhat counter-cyclical (the 2022 wholesale spike triggered a £290m clawback). Across both schemes, ~£6.7bn was paid to GB wind in 2024, and as a share of baseload the total has exceeded the wholesale value of the energy itself for most of the past decade.

It's worth pausing on the indexation choice baked into the RO. The buyout price was uplifted every obligation year by RPI, not CPI — and the two are not the same animal. CPI is the conventional benchmark for inflation (it's the Bank of England target, and what CfDs are indexed to); RPI consistently runs about a percentage point a year hotter, mostly because of methodological quirks the ONS has long acknowledged but governments have been slow to fix.

From April 2010 (the heart of the offshore-wind RO accreditation window) to April 2026 (when the final RPI uplift was applied, after 16 obligation years of compounding), RPI rose 86% against CPI’s 59% — a 27 percentage point gap, with RPI compounding at 4.0%/yr versus CPI’s 3.0%/yr. That one extra percentage point a year stretched the nominal value of every ROC by ~17% more than it would have been under CPI. Applied to a 2010-era ROC worth ~£37/MWh, that’s the difference between £59/MWh and £69/MWh by April 2026 — about £10/MWh of indexation premium on top of every MWh, for 20 years from accreditation, across ~70 TWh of RO-era wind. Quiet, technocratic, and worth roughly £700m a year by the end of the indexation period.

The RPI premium is the slow leak. The bigger leak — and one the policy debate did partly notice — was the 2021-23 wholesale crisis. RO is structurally different from CfD here: when wholesale prices spiked, CfD generators paid back the upside above strike (a £290m clawback in 2022 alone), but RO generators kept all of it. The ROC is a flat per-MWh top-up paid on top of whatever wholesale price the generator captures, so a pre-2015 offshore wind farm in 2022 earned roughly £180/MWh wholesale + £145/MWh ROC ≈ £325/MWh — versus a comparable CfD generator capped near strike (~£40-130/MWh depending on allocation round). Across 2021-2024 the wholesale windfall to GB wind, against a 2017-2020 baseline, came to ~£22bn — of which ~£16bn went to RO-era assets. The Electricity Generator Levy (Nov 2022 - Mar 2028; 45% rate on revenues above a £75/MWh threshold) recovered roughly half of this — but crucially, EGL receipts flow to HM Treasury as general revenue, not back to billpayers. CfD clawback, by contrast, is netted off the supplier reference price and so does feed through to lower bills. So from a consumer perspective the levy doesn’t undo the RO windfall; it just transfers part of it from generator balance sheets to general government revenue. The other half stayed with the generators. Put plainly: the EGL is effectively a tax on energy consumers collected via generators — the consumer pays the windfall either way; the only choice is whether the generator or the Treasury banks it.

To be fair, HMT has redirected general revenue back into bill support — but selectively. The 2022-23 Energy Price Guarantee and Energy Bills Support Scheme bridged the worst of the crisis for households via direct payments and price caps; a recent reform has moved ~75% of RO costs off domestic electricity bills and onto general taxation, on the grounds that loading policy costs onto electricity makes electrification (heat pumps, EVs) artificially uncompetitive against gas; and energy-intensive industries have separate exemption schemes that remove most policy costs for them. But these are discretionary choices, not mechanical: which billpayer gets relief, how much, and for how long is set by the Treasury on a case-by-case basis — nothing like the automatic per-MWh netting CfD clawback already does for everyone. So the headline “RO is locked in to the late 2030s” understates the bill: it isn’t just the index choice that made the RO generous — it’s that the scheme has no wholesale-side correction when prices run four times the forecast.

Worth noting that we haven’t attempted to unpick the extra network costs of wind farms, which are mostly built in remote parts of the UK that previously lacked substantial transmission capacity. Ofgem’s ASTI (Accelerated Strategic Transmission Investment) framework alone fast-tracked ~£20bn of accelerated transmission projects, and NESO’s Beyond 2030 / Clean Power 2030 plans imply a network capex bill running into many tens of billions across the decade — borne by consumers through TNUoS and supplier pass-through, on top of everything in the chart above.

Why gas is no longer the biggest cost for gas plants

Commentators frequently conflate the SRMC (short run marginal cost) of gas generation with the total cost, which factors in broader startup, maintenance, capital (inc finance) costs that are amortised (spread) across the running hours. Since 2018, renewable displacement of British CCGT generation has roughly halved their generation, meaning the typical CCGT generates but 1/4 of its potential:

As a result, the non fuel costs represent a growing share of the overall CCGT cost envelope per remaining MWh generated, outside of course the short-term 2022 Ukraine spike in gas prices. Not only that, but a “perfect storm” is driving up each non-fuel element of CCGTs, including:

Start Costs CCGTs typically need to generate at 40-50% of their nameplate capacity (Stable Export Limit), and displacement by zero marginal cost renewable generators (wind and solar) in the merit order means they are forced to stop and start far more frequently. For turbines, starting is the biggest single cause of wear and tear, on top of the additional fuel needed to “warm” the turbine before each run. The average running time of a CCGT has fallen about 50% from above 30 to just above 20 hours, with 25% of starts only lasting 15 hours or so. As I’ve written about recently, they’re also being ramped more aggressively to minimise overall generation time, another penalty for wear and tear:

Capital Costs

For new build CCGTs, the great inflation of 2022 lifted the capital cost of construction and gas turbine costs, along with the better known escalation of costs for wind and solar. However, this accelerated with the onset of the AI boom in 2023-4, as demand, primarily outside Europe for gas turbines for power generation for data centers pinched the limited global supply chains for manufacture that had been on the wane for many years, and severe backlogs mean that wait times now extend several years into the future at major manufacturers. Of course, these have been complemented by higher interest rates.

Putting these together, and amortising to ever shrinking generation volumes, the non-fuel costs of gas fired power in the UK are now pushing an estimated £60/MWh, close to both the CfD price paid to cheaper renewables and the normally-dominant fuel cost.

The brake, made visible

Pull the threads together — wind cannibalising its own capture price, the rising cost of balancing it with increasingly part-time gas plants, and a subsidy bill that exceeded the wholesale value of the energy itself for most of the period — and one observation lands: each new GW of wind has delivered less to the system than the last. The cruel irony is that gas and carbon costs ran far higher than FES 2018 projected; that should have lifted baseload prices and rewarded wind handsomely. Instead the demand-side failure dominates: the electrification growth that 2018 policy assumed — heat pumps, EVs, industrial demand — simply hasn’t materialised, and the new wind capacity is scaled to a demand level that hasn’t grown. This isn’t a transient COVID effect: the COVID period is excluded from the fit below, and the trend through the non-pandemic years is unmistakable — broadly flat through the first ~22 GW, then a sharp accelerating decline past 28 GW. Diminishing returns to scale — and the root cause is the demand growth that never showed up.