Pick a plan and a demand pathway, then step through the years to 2050. Every figure comes from the model’s hourly runs.
Compare generation build plans for Britain from 2030 to 2050 against 4 demand pathways. The model tests whether electricity is available when needed and estimates generation, storage and trade costs.
The pathways come from the National Energy System Operator (NESO), Future Energy Scenarios 2025: Falling Behind (lower demand), Holistic Transition (central) and Electric Engagement (higher). A fourth case, AI expansion, is NESO’s higher pathway, mandates included, with an extra 60 TWh of data-centre demand phased in from 2030 to 2035: about 101 TWh a year of data centres in total (the higher pathway’s 41 TWh plus 60 TWh). It is the most demand the model plans for. That is a planning choice, not a forecast. It sits above the roughly 40–50 TWh implied by the government’s forecast that Britain will need at least 6 GW of AI-capable capacity by 2030 (UK Compute Roadmap, 2025) and below LSE’s 20 GW high case in evidence to Parliament. NESO’s own figures are lower: 20–41 TWh in 2035 in its 2025 scenarios, and about 22 TWh in its 2026 Ten Year Outlook. The three NESO pathways do not include it.
Builds hourly demand profiles using annual and peak-demand assumptions, then tests each modelled year against 7 historical-weather profiles (2018–2024).
Tracks construction, retirements, plant availability, wind and solar output, fuel use, storage and limited electricity trade. Electrolysis can move to lower-demand or surplus-generation hours within a bounded window, preserving its energy requirement. Calculates construction spending and annual capital, fuel and operating costs.
Checks hourly shortages, reserve capacity and an approximate inertia screen. 2 additional cold, low-wind and nuclear-outage stress cases are retained in the supporting audit; this page shows the historical-weather results.
Main sources: NESO demand and generation records; Department for Energy Security and Net Zero fleet, project-pipeline and generation-cost datasets; supplier and engineering estimates. All source references are linked at the bottom, alongside the calculation methods and assumptions.
Grid connection and network upgrade costs for new projects are estimated separately and shown alongside the cost totals (central estimate, with a range). Inherited network, constraint payments and retail costs are not included.
Choose a plan and demand pathway, then edit what gets built. Preset results are precomputed; edits update the build schedule immediately. Use Recalculate to update hourly supply and cost results.
1. Choose a starting plan
2. Choose a demand pathway
The nuclear buildout can be built in different ways. Official standard (the default) sizes the gas fleet to the 3-hour reliability standard within limits on how fast equipment can be delivered. Tight equipment market does the same with scarce, more expensive turbines and engines. Planning case, Stronger retention and Faster delivery keep existing gas for five or ten years and vary how much new equipment can arrive each year.
Current policy adds no firm power beyond like-for-like gas replacement, so there is nothing for the setting to change and it is greyed out. That is not because it falls short; it does, but that is the result it is there to show. Current policy delivered is sized once, with the Official standard costs and delivery limits. Programme demand offers only Official and Tight, because only the settings sized to the standard were run for it.
These alternatives retain suitable existing gas capacity, reduce new equipment throughput and let coal’s operating role change during the nuclear build-out. The five delivery settings above select the main planning and sensitivity cases. This register retains all detailed experiments; supply gaps remain visible.
Headline cost covers generation, storage and net electricity trade. Wider network and retail costs are not fully included; carbon and renewable-support payments are shown separately.
Normal-weather mean · annual electricity, TWh
Domestic output by source
Normal-weather mean, before exports
Delivery warnings cover orders placed from 2029 onward. Inherited pipeline delivery risks are excluded from these warnings; supply tests still include the whole fleet.
Seven historical-weather profiles against the selected demand pathway. Annual shortfall is unserved TWh divided by demand TWh, taking the worst profile for each year. Peak hourly gaps and reserve/inertia failures are shown separately. Extreme-weather tests remain in the supporting model and audit. The last column is a proxy for the GB reliability standard (3 hours a year expected loss of load): random plant outages across the seven weather years with NESO interconnector credit. It is shown only for tested presets and is not NESO's own modelling.
| Year | Normal-weather screens | Worst annual shortfall % | Unserved TWh | Peak hourly gap GW | Reserve-gap hours | Inertia-gap hours | Expected shortfall hours/yr (3 h std) |
|---|
The clean-ambition case and its gas-firmed alternative share wind, solar, storage and named nuclear projects. The latter adds analyst gas orders. These are modelled sensitivities, not complete official 2050 deployment plans. Low-carbon firming, demand flexibility and interconnector expansion are not fully represented. The two supply-first columns use the selected delivery setting. The 95% clean column remains the frozen output-ceiling experiment, with its original retirement schedule and no retention extension. The supply status here covers normal-weather profiles; the supporting audit retains the full stress-test results.
The 2030–2050 total sums annual capital recovery and operating costs, including fuel and net trade, across all 21 years. It uses real 2024 pounds without discounting; construction cash is not added again. Costs remain subject to the coverage note above. Read them alongside supply shortfalls: unserved electricity has no monetary penalty in this model, so a shortage case is not an equal-service cost comparison.
What each type of power station costs to build and to run, using the model’s reference assumptions.
Compare generation units
Construction cost divided by potential lifetime electricity, using the model’s reference utilisation and operating life.
| Generation unit | Unit size GW | Construction £bn / unit | Grid connection £bn / unit | Build years | Order to online years | Potential utilisation | Service life years | Annual output TWh | Lifetime output TWh | Construction only £/MWh · p/kWh | With grid connection £/MWh |
|---|
Gas and coal use potential output when available, not an assumed policy dispatch share. Availability/utilisation includes the stated maintenance allowance. The table’s solar comparison uses constant potential output; the hourly model separately applies solar ageing and inverter provisions. Build years exclude the common one-year planning assumption; order-to-online includes it. These are model assumptions under the proposed reforms. Repeat SMRs also require pilot and factory readiness: a five-year lead does not permit orders before that gate. Batteries store electricity and do not have a generation lifetime-output ratio.
Grid connection is the cost of connecting the unit to the transmission or distribution network, which the construction references exclude. Wind and solar figures are sourced: DESNZ/Arup generation-cost infrastructure estimates (2023 prices), plus offshore transmission assets valued by Ofgem for offshore wind. Gas plant figures are DESNZ’s own (Electricity Generation Costs 2025, 2024 prices). Nuclear is benchmarked on Ofgem’s £656m allowance for the line connecting Hinkley Point C, about £200 per kW. Coal, batteries and pumped hydro use a model allowance of £0–150m per GW (£50m central), because no published UK figure was found; coal reuses existing station sites. Wider network reinforcement is estimated at plan level, not per unit.
Compare running costs
Standard new units operating at their design efficiency, with the same potential utilisation and maintenance allowances as the capital table. Figures are in real 2024 pounds.
| Generation unit | Unit size GW | Annual output TWh | Fixed running £m per year | Variable running £/MWh | Fuel at design efficiency £/MWh | Annual running £m per year | Operating only £/MWh · p/kWh |
|---|
Fixed costs include maintenance and the model’s lifecycle provisions. Solar inverter replacement is funded within its existing operating allowance. Borrowing costs, including interest on coal stocks, are excluded from this reference table and remain in the hourly model accounts.
These figures describe intended performance under steady operation at design efficiency. Actual operation, site conditions and grid decisions may prevent this being achieved: starts, part-load running and lower utilisation can increase operating cost per MWh. The hourly model calculates actual plan outcomes using its separate fleet and dispatch assumptions. Wind and solar retain weather-based potential yields; solar ageing is applied in the hourly model. SMR costs are forward estimates, not evidence from an operating Rolls-Royce fleet.
The build plan
Order whole stations or reference wind and solar projects. Delivery time, capacity and construction cost follow automatically.
Choose the year once, then add several technologies together.
In this plan for 2029
| Generation or storage | Units to add | Online | Added capacity | New-build cash £bn | Refurbishment cash £bn |
|---|
Delivery time includes one planning year. New orders begin in 2029. The Wylfa pilot is an inherited programme, shown with inherited projects; its assumed delivery starts are not new order decisions. Existing-fleet assumptions remain in the model. Changing the build plan updates capacity and construction spending immediately.
| Technology / project | Units | Order / delivery start | Online | Added GW | New-build cash £bn | Refurbishment cash £bn |
|---|
Separate planning and grid-connection costs for new projects are excluded from the headline estimate unless explicitly added in Details.
Construction cash, finance and annual capital recovery are separate. The policy selector chooses whether the ceiling is a diagnostic or a binding annual gas-output budget. The constrained case retains the gas fleet but leaves demand unserved when its output allowance is exhausted.
Foreign availability and prices are assumptions. No automatic interconnector expansion or foreign reserve credit is included.
The same annual electrolysis energy is required. The default shifts production within each day at no more than twice average electrical load. A hydrogen buffer of twice the window (48 hours by default), initially half full and ending unchanged, supports steady offtake if electricity is served. Capacity and buffer are estimates, with day-ahead weather knowledge; there is no seasonal shifting or hydrogen-to-power generation. Electrolyser, buffer, compression and cycling costs are excluded unless represented by the additional-cost allowance. Zero means excluded, not free. Recalculate after changing these settings.
Baseload targets full available output, with the existing 85% maintenance/availability allowance. Fuel interruptions can reduce output; any surplus is curtailed and fuel costs remain charged.
Domestic production and fuel delivery are assumptions, not secured supplies. Delivered prices include mining and transport.
Hydro, biomass, waste and other renewable generation remain visible as existing sources. No new hydro or tidal projects are inserted. Pumped hydro is storage.
| Source | Baseline | Retain % | From year |
|---|
Installed capacity is before weather and outage adjustments. Construction spending is cash in that year, not an annual electricity cost.
| Year | Nuclear GW | Gas GW | Coal GW | Offshore GW | Onshore GW | Solar GWp* | Storage GW / GWh | New-build cash £bn | Refurbishment cash £bn |
|---|
*Inherited solar retains its documented capacity basis; new solar uses DC panel capacity.
Normal-weather means for the inspected year. Wider-network and fuel-infrastructure allowances are partial sensitivities. Carbon and renewable-support payments are separate financial transfers.
| Generation / storage cost | £bn / year |
|---|
| Allowance | Gas delivery TWh/day | Working stock TWh | Additional resource £bn/year | Separate stability payment £bn |
|---|
| Separate renewable support | Reference £/MWh | Support £bn | Treasury £bn | Consumer levy £bn |
|---|
The support ledger uses financial years and excludes nuclear Contracts for Difference and future unawarded rounds. It is not added to the resource-cost subtotal.
Transparent inputs, explicit assumptions
The charts use the audited model. All source data, estimates and methods remain available below.
Each project enters capacity after its planning and construction lead time and leaves at its recorded retirement date. Existing assets retain their estimated vintages. Optional gas life-extension cases include inspection success, refurbishment spending and outages, reduced availability and lower efficiency.
Online year = order year + lead timeEach year uses seven historical-weather profiles and two synthetic cold/low-wind and nuclear-outage stresses. Thermal starts, part-load losses, storage, reserves and bounded imports are explicit. Electrolysis uses a separate capacity-limited schedule; its annual energy is unchanged. Preset gas orders are sized in whole stations against the retained supply screens and delivery limits; edited plans receive no automatic backup additions.
Supply + storage discharge + shortage = demand + charging + exportsResource costs combine capital recovery, operations and fuel; trade revenue is shown separately. Construction cash, support payments and carbon transfers are not added again.
Annual net subtotal = capital recovery + operations − export receiptsAnnual energy totals can balance while particular hours fail. Passing the tests means satisfying the model’s energy, reserve and scalar inertia screens; it does not establish dynamic frequency stability or a probabilistic reliability standard.
Costs are partial generation/storage costs in real 2024 pounds, not household bills or full system costs. Connections, network geography, retail and complete fiscal settlement remain incomplete. Normal-weather costs are averaged across seven years; synthetic stress cases have no assigned annual probability.
The proposed SMR delivery rate and repeat-unit price remain assumptions. Repeat SMR orders start conditionally in 2034, after a pilot design/construction-readiness review; first repeat output is 2039. Assumed pilot operation is 2035 for all three units. Four 2 GW coal stations open in 2035–38; one additional 1.7 GW large reactor opens in 2037 under the seven-year-build sensitivity. Every preset includes the same bounded electrolysis flexibility. Other official demand flexibility, hydrogen-to-power/CCS, long-duration storage and interconnector expansion remain incomplete. The 95% target includes nuclear and retains gas backup. The ambition case removes additional adequacy gas; the alternative keeps it. The ambition case enforces the annual gas-output ceiling; the gas-firmed alternative prioritises supply and can exceed it. Shortages are conditional on the modelled resources and quota controller.
Preset gas orders start in 2029. Retaining existing generation, future fuel availability and modelled foreign supply remain conditions. Renewable support-contract expiry does not automatically retire a physical asset.
| Method | Calculation | Limits and estimates |
|---|
| Area | Assumption | Value | Basis and status |
|---|
| Ref. | Source and use |
|---|
The audit report sets out which values come directly from sources, which are analyst estimates and which are policy assumptions.