Electric heating and transport need more electricity, while using less fuel. Our central sector scenario reaches about 93.4 TWh a year in 2050. We choose it for planning because the lower case relies on strong savings and less growth occurring together.
This is customer electricity before network losses. The supply programme adds a separate data center allowance. The lower and higher cases test other growth and efficiency assumptions; they carry no probabilities.
Planning choice and demand boundariesMore electricity can mean less fuel
Electricity is one form of energy. Replacing an oil boiler or a petrol car changes which energy source we use. It also changes how efficiently we use it. The introductory guide explains this with a heat pump.
Demand model: fuel conversion and efficiencyOur model asks how much electricity reaches homes, businesses and other users. This is called final consumption. The supply calculation must also cover electricity lost in the grid and electricity used to pump water into storage reservoirs.
BFE: electricity balance, table 6In 2024, customers used 57.5 TWh of electricity. Our proposal adds supply for the transition while reducing the fuel needed for heating and road transport.
Observed final consumptionA home and a car
Purchased fuel energy
Heat: 14.1 · Travel: 7.2
Purchased electricity
Heat: 4.0 · Travel: 2.4
Electricity use rises for these services, while purchased energy falls. The heat pump also draws heat from the environment.
Inputs, calculation and limits
Every input is an illustrative assumption: boiler efficiency 85%; heat-pump seasonal performance 3; fuel-car input 0.6 kWh/km; electric-car battery use 0.18 kWh/km; charging efficiency 90%.
Before: heat ÷ boiler efficiency, plus distance × fuel energy per km. After: heat ÷ seasonal performance, plus distance × battery energy per km ÷ charging efficiency.
Assumed home and car with the same useful heat and travel in both cases. Compares purchased fuel with purchased electricity. Household representativeness and equipment performance remain unmeasured. Other household uses, upstream and embodied energy, costs and hourly winter peaks are outside the example. Cold-hour performance needs a separate calculation.
Electrification changes demand
The central case models more electric transport and heating. It also changes population, activity and efficiency. The result shows annual electricity use under those assumptions.
Independent demand model v1Electricity per year, TWh · same scale for every bar: 0–30
The baseline is reconciled to the measured national total. Its sector allocation is model accounting. “Unallocated” preserves the remaining accounting difference.
Transport and heating become larger electricity users as engines and boilers are replaced. The model already includes efficiency gains and charging losses. For heat pumps, it uses performance over a whole heating season.
Model code, assumptions and input provenanceIndustry fuel use and the conversion assumption
16.5 TWh of gas and light fuel oil
- Natural gas
- 11.7 TWh
- Light fuel oil
- 4.8 TWh
This is fuel energy, before conversion into useful heat or other services. Replacing it requires different amounts of electricity for building heating and industrial processes.
Prioritise efficient building heating; assess each industrial process by its temperature and equipment.
Coverage, uncertainty and calculation
The sum uses BFE’s 41 grouped sector estimates once: 42,110 TJ of gas plus 17,380 TJ of light oil, divided by 3,600 TJ per TWh. It includes building and process uses in the covered industry/services establishments. Other fuels and excluded sectors need separate accounting.
BFE labels the finer breakdown experimental statistics. Small-sector estimates can be imprecise; the published coefficients of variation are retained. A joint uncertainty interval and a split between building and process heat have not been established. This total is neither extra electricity to add to the demand model nor a complete industrial electrification target.
BFE: grouped estimates and methods · Fuel totals and conversion
Different assumptions, different demand
All three cases make the same broad switch to electric transport and heating. The difference is how much activity grows and how efficiently it uses electricity. The lower case combines less growth with stronger efficiency gains. The higher case combines more growth with weaker gains.
Scenario inputs checked against the original modelLess growth, stronger efficiency gains
Central sector scenario. The supply programme adds a data center allowance.
More growth, weaker efficiency gains
All bars use the same scale, from 0 to 160 TWh. The cases combine different inputs; probabilities have not been assigned to them.
Original model and results · All case inputs and reconciliation
Population, activity and efficiency: compare the scenario inputs
The main assumptions, side by side
Every value below is an assumed 2050 input. Percentage changes are relative to 2024. “Per unit” separates efficiency from growth.
- ResidentsPeople living in Switzerland.million residents
- Lower9.5
- Central10.5
- Higher11.5
- Service-sector activityAssumed change in service-sector floor area and activity, including agriculture. Expressed as an activity index rather than GDP.% versus 2024
- Lower5% more
- Central25% more
- Higher50% more
- Industrial activityAssumed change in industrial activity, expressed as an activity index rather than GDP.% versus 2024
- Lower5% less
- Central15% more
- Higher40% more
- Road travelTotal distance driven by the covered road vehicles.% versus 2024
- LowerNo change
- Central15% more
- Higher35% more
- Electricity per unit of business activityExisting service and industrial uses, excluding separately modelled building heat and cooling. New electric uses are added separately.% versus 2024
- Lower25% less
- Central10% less
- Higher5% more
- Space heat needed per unit of activityHeat per resident in homes, or per unit of business activity. Insulation and heating controls can reduce it. Weather stays unchanged.% versus 2024
- Lower40% less
- Central25% less
- Higher10% less
- New heat pumps for space heatingHeat delivered per unit of electricity over a season. A larger number means less electricity for the same heat.kWh heat / kWh electricity
- Lower4.0
- Central3.3
- Higher2.5
- Electric-car consumptionEnergy at the battery per 100 km. Charging losses are added separately.kWh / 100 km
- Lower15.0
- Central18.0
- Higher23.0
Heat-pump figures describe performance over a season. Cold-hour performance needs separate analysis. Car figures are battery consumption; each case also has its own charging-loss assumption.
Other differences and the complete inputs
The cases also change home electricity use per person, hot-water demand, cooling, vans and heavy vehicles, charging losses and where vehicles charge.
Industry uses different amounts of process heat and a different mix of heat pumps and direct electric heating. The treatment of old heating equipment and the unallocated accounting remainder also varies.
These changes interact. For example, more driving multiplies the effect of a less efficient car. The model calculates their combined effect.
The network-loss assumption also changes, but affects the separate generation-accounting result. The final-demand totals here are calculated before those losses.
The future inputs are analyst choices; the baseline uses the official sources indexed with the original model.
Download every input for all three cases (CSV) · Download definitions and reproduction details (JSON)
Population assumption compared with the BFE input
BFE’s Energy Perspectives 2060 inputs use an annual-average population of 10.31 million in 2050. Our central case assumes 10.5 million. The official input gives us a useful comparison for that choice. Changing only that population input changes our result by -0.4 TWh, to 93 TWh. That leaves the broad supply requirement close to our chosen reference. Activity, equipment and efficiency remain the larger questions.
Population-only recalculationBFE: population scenario and definition, page 7 · Original workbook: both scenario sheets, cell BC13
We choose the central case because the lower-demand case depends on several favourable conditions occurring together. It combines strong efficiency gains with less industrial activity than in 2024. Committing supply on that basis would leave less room for economic growth or weaker savings. The higher case shows the cost of missing those savings while activity grows. It remains a supply stress case, rather than our expected outcome.
Reason for the planning choiceWhat all three scenarios electrify
What stays the same?
All three cases fully electrify the road-transport modes covered by the model. They replace all modelled oil, gas and coal building heating, plus identified household resistance heating. They also convert 90% of the modelled oil, gas and coal industrial process heat to electricity. Reaching these assumed targets requires equipment replacement and suitable installations.
Shared replacement targets and covered sectorsAll three cases use the same annual heating-weather basis; the differences come from growth, efficiency and the other listed inputs. None includes a separate growth allowance for data centers, hydrogen, e-fuels or electric district-heat production beyond the sector baseline and its activity changes.
Shared inputs and excluded additionsCheck heating performance with meters
Use heat pumps, improve buildings and reward useful control. Supply must cover their cold-hour electricity demand as well as their annual energy use. Measure the whole installed system before counting its savings.
Efficient-electrification recommendation- Winter average
- 140 kW
- Highest recorded hour
- 308 kW
This sample mixes apartment buildings, houses and shared installations. Normal utility controls remain active. It shows why average use and peak demand need separate checks; national sizing needs a representative building mix and more severe-cold measurements.
Sample selection, readings and calculation · Kaiser et al.: meters and control conditions
From these measurements to a national heating estimate
The sample has complete readings for October–March. It measures electricity drawn from the grid, without separately measuring useful heat or all backup-heater use. The associated regional weather series contains only six hours below −10°C. It therefore provides little evidence for a prolonged severe cold spell.
Building counts alone also leave a gap. In the October 2026 federal register, only 13.5% of existing buildings with residential use have an energy-reference area recorded. This is the floor area used for energy calculations. A national estimate also needs building heat loss, heating equipment and controls; missing area requires explicit assumptions.
Federal building register: coverage and definitions · Values and calculation boundaries (JSON)
Measured differences and what they mean
A Swiss study compared the internal readings of 4 heat-pump systems with calibrated field meters. Agreement varied by system and available data. Meter quality and the equipment included in each reading therefore matter.
For January–April 2026, the reported main heat-output series differed from the field totals by -34.7% to +0.1%. Available compressor-electricity comparisons ranged from -10.5% to -4.6%.
For unit F017, the heat difference was -34.7% using minute data, but -7.7% using monthly data. Data selection changes the apparent result.
The electrical comparison concerns the compressor. Whole-system performance also needs pumps, controls and backup heaters. The sample establishes measurement discrepancies; a national seasonal-efficiency estimate requires representative installations and consistent meter boundaries.
Allow separately for growing data centers
Existing electricity uses remain in the sector baseline and change with the activity assumptions. The original central curve has no separate growth addition for data centers, hydrogen, e-fuels or electric district-heat production. We recommend the data center allowance below alongside that curve; the other additions still need defined projects and energy requirements. Aviation conversion to electricity or synthetic fuels is outside the model.
Central inputs and model scopeThe supply programme plans for 94.2 TWh a year. It adds only the data center growth beyond what the sector model already counts. The allowance holds a published 2030 scenario constant afterwards; faster growth needs more supply.
Programme demand reconciliationData center study range and matched supply tests
Estimated use and study scenarios
Plan for data center growth
The BFE-commissioned study expects more electricity use within its defined data center group.
Annual electricity, TWh · common scale 0–4. The shaded range shows alternative study outcomes.
About 0.4–1.1 TWh more by 2030. The separate high case adds about 1.4 TWh and may arrive after 2030. These differences use the rounded 2.1 TWh baseline.
Existing data center electricity is already inside national consumption. We count only the demand above growth already represented in the sector model.
Allowance beyond the existing demand path
Plan for about 1.1 TWh of additional annual demand by 2030 under the study's central case. This uses an explicit assumption that its existing data center demand is part of retained service-sector electricity.
The model already carries 2.15 TWh for that activity by 2030. Subtracting it from the study's 3.2 TWh total gives the extra allowance. The study's lower case adds 0.35 TWh under the same mapping.
We recommend carrying this allowance separately while actual connection and consumption data establish its timing. The hourly 2035 sensitivity holds the study's central total constant after 2030; further growth would need more supply.
Study coverage and the growth calculation
The study covers commercial, dedicated and internal data centers. Small business server rooms and micro data centers are outside its matched comparison. Its 2024 consumption is a survey-based estimate; the report gives a more precise total in its detailed results.
The chart's approximate growth subtracts the rounded 2024 total. The model allowance instead uses the detailed 2.08 TWh baseline, multiplied by service activity and electricity intensity interpolated to 2030. Subtract that embedded demand before adding the study outcome. An all-industry mapping is also available in the download. Both are explicit approximations; the exact sector and cooling allocation remains unresolved. The high case can arrive after 2030. A later plateau is an assumed sensitivity, not a forecast to 2050.
Data center study: summary and scope · Growth reconciliation, mapping sensitivities and page 127 timing
| Data centers: TWh/year | Annual supply gap: TWh | Peak gap: GW |
|---|---|---|
| 3.2 | 4.6 | 3 |
| 5 | 5.4 | 3.2 |
| 10 | 7.7 | 3.8 |
Our central sector scenario of 93.4 TWh already includes data centers. Applying the published 3.2 TWh total adds only growth not already counted, giving 94.2 TWh of customer demand in 2050. The extra allowance changes between years because the sector scenario changes.
The 3.2 TWh figure is a published 2030 scenario held constant thereafter. The 5 and 10 TWh cases replace that total; they are not added on top. Extra demand is spread evenly over the year, and network losses are added once.
The hourly gaps are preliminary model results. Cold-weather heating, access to hydro turbines and internal grid constraints need engineering checks before equipment is sized.
Sector baseline and data center allowance · Matched growth tests
Large new electricity users need identified supply and grid connections. We recommend updating the programme as their projects become credible, while counting efficiency gains only once.
The winter chapter explains when this electricity must arrive.
The 2025 annual check and comparisons with other studies
Compare the path with recorded demand
The unchanged model gives 58.95 TWh for 2025. Recorded national consumption was 57.97 TWh, leaving the model 1.7% higher. That supports retaining it as a working path while checking how quickly equipment is actually replaced.
BFE: national final-consumption balance Unchanged path compared with 2025The model retains its 2024 weather basis and assumed replacement path. The observed year has its own weather, activity and equipment stock. Their differences have not been separated here.
A similar annual total can hide errors that cancel each other. It does not validate future heat-pump performance, industrial conversion or the hourly winter peak. The narrower industry/services survey also has different coverage from the national balance.
Why other studies can give different electricity totals
Separate the heat people need from the electricity used to provide it. The ETH–PSI study illustrates this distinction: some models start from heating and transport needs, while another starts from fixed electricity demand. The choice of equipment then determines how much electricity each service uses.
Demand methods and input definitions, Tables 2–3ETH–PSI demand inputs: keep heat and electricity separate
The study compares models with different demand methods. SES, SES-ETH and STEM choose how to meet energy needs. Nexus-e starts with a fixed electricity-demand input.
| Input | TWh/year | Meaning |
|---|---|---|
| Space heating | 52 | Heat needed in SES, SES-ETH and STEM; electricity depends on how that heat is supplied. |
| Hot water | 9.3 | Heat needed in SES, SES-ETH and STEM; electricity depends on how that heat is supplied. |
| Industrial heat | 18.6 | Heat needed in SES, SES-ETH and STEM; electricity depends on how that heat is supplied. |
| Nexus-e electricity input | 76 | Fixed Nexus-e electricity input, adopted from BFE 2020. |
Heat-service quantities describe useful heat. Converting them to electricity requires the heating mix and equipment performance; adding them directly to electricity consumption would overstate demand.
The other three models calculate electricity consumption within the model. Their base electricity-service inputs also differ from final consumption. Align these definitions before comparing their results with our demand cases.
Table 3 and footnotes 2–3, page 5 Study assumptions and review (JSON)
These different definitions explain why headline demand figures can disagree. We retain our sector-based reference so that the chosen heating, travel and activity assumptions remain explicit.
References
- WattInfo archive-to-model checks: demand, thirteen weather years and heat storage
WattInfo · Additive model evidence review, 18 September 2026 - Schweizerische Gesamtenergiestatistik 2025 - Tabellen
Bundesamt für Energie / named report authors · 2026-07-10 - WattInfo proposed supply path: decisions and supporting calculations
WattInfo · Proposed priorities and fixed-charging supply screen, 18 September 2026 - Chauffez efficacement avec une pompe à chaleur
SwissEnergy / Swiss Federal Office of Energy · Web edition retrieved 2026-09-16 - Visual explanations: seasonal stress, demand and storage
WattInfo · Visual chapter data and explicit teaching assumptions - Swiss Electricity Statistics 2024
Swiss Federal Office of Energy (BFE) · 2024 - Independent demand model v1
Swiss Energy research project · Frozen research output - What changes between the three demand cases
WattInfo · Explanation of frozen demand-model-v1 scenarios; 17 September 2026 - Energy Perspectives 2060: demographic input workbook
BFE · Workbook listed 1 September 2026 - Energy Perspectives 2060: common assumptions
BFE · Listed 1 September 2026; PDF dated 14 September 2026 - Heat-pump readings compared with calibrated meters
OST / ETH Zürich for EnergieSchweiz · Report dated 11 August 2026; listed 15 September - Swiss data-centre electricity use and efficiency potential
TEP Energy / HSLU / Willers for EnergieSchweiz · Report dated 6 July 2026 - Demand and field-trial evidence: selected inputs and calculations
WattInfo · Primary-evidence integration, 17 September 2026 - ETH–PSI: nuclear white paper modelling documentation
ETH Zurich and PSI · 2026-06-29 - ETH–PSI study: conditions, model inputs and reading notes
WattInfo · Study review, 18 September 2026 - WattInfo electricity programme: 66 hourly cases, delivery dates and owner finance
WattInfo · Conditional programme, 19 September 2026 - Matched electricity, peak-power and trade comparisons
WattInfo · 2026-10-02 - Swiss residential load profiles 2023–2024, dataset.zip
Kaiser et al. / Zenodo · 2026-07-21 - Dataset on residential electricity load profiles in Switzerland, v1 21 July 2026
Kaiser et al. / ETH Zürich · 2026-07-21 - National public GWR building extract, selected complete ZIP members
Federal Statistical Office / GWR · 2026-10-04 - Measured heating demand and repository planning boundaries
WattInfo · 2026-10-04 - Energy consumption in Switzerland2025 by NOGA divisions: industry and services survey
BFE · 2026-10-10 - Industry fuel totals, generation price references and delivery gates
WattInfo · 2026-10-14 - Dekarbonisierungspotenzial bei HT-Prozesswärme
BFE / commissioned authors · 2026-10-23
Find definitions, calculations and model inputs in the source register.