Skip to content

New Technology Characterization

What the model may build, what it costs, and where it can go

Watch first: New Technology Costs (2:08, narrated). Generated from a real build of the German model; every number on screen traces to a file. The sections below follow the film in the same order, with the rules and file names behind each step. Also: interactive version · India · all films.

Existing-stock characterization tells the model what is on the ground today. This page covers what can be built tomorrow: the list of options, their cost from now to 2050, the cost of the money to build them, and the nodes where they are offered.

The numbers come from one module, 4technology_characterization/, which is shared by VerveStacks and the KiNESYS global model.

Note

Status, October 2026. The cost table holds everything described here. The country builds read the central reference path from it. The high, low and learning paths, and the country cost of capital, are emitted today for the KiNESYS models and are being brought into the country builds.

1. The menu

A country model is offered these new plant types (CONV_TECH_CATALOG in excel_manager.py):

Group Technologies
Power only Gas combined cycle, gas turbine, gas with carbon capture; coal subcritical, supercritical, ultra-supercritical and gasification; nuclear, small modular reactor; biomass, biomass with carbon capture
Heat and power Gas combined cycle (extraction), biomass chips (extraction); gas with carbon capture, biomass pellets with carbon capture
Heat only Gas boiler, biomass boiler, electric boiler, large air-source heat pump
Storage Lithium-ion battery, 4-hour and 8-hour

Three kinds of new capacity are handled elsewhere:

2. Many sources, one table

No single source covers every technology, region and year. The module reads each source into a standard table, converts everything to one currency, and assembles one long file, unified_reference_costs.csv: 63 technologies, 9 regions, about 19,400 rows.

Source Role
IEA World Energy Outlook 2025, power generation assumptions Regional cost levels, fixed operating cost, efficiency and capacity factor for power-only technologies, in 9 regions
NREL Annual Technology Baseline 2024 v3 The high and low spread around the central path; batteries; small modular reactors; technical lifetimes
Danish Energy Agency technology catalogues Combined heat and power, heat-only plants, heat pumps, storage, and carbon capture (June 2026 datasheet for electricity and district heating)
IRENA Renewable Power Generation Costs Observed installed costs: the anchors and history for the learning paths
US EIA Annual Energy Outlook, Lazard Variable operating cost where the main sources give none
Damodaran (NYU Stern), World Bank indicators Cost of capital (section 5)

Sources and what each supplies

One currency. Sources report in different price years and currencies. Every value is converted to US dollars of 2022, using the US GDP deflator and, for the Danish catalogues, the European Central Bank's annual euro–dollar rates.

Every number keeps its source. source_versions.csv records, for each input, the publisher, version, file hash and extraction date. A value in the cost table can be traced to the file it came from.

Costs are overnight. Interest during construction is not included; construction time is carried as its own parameter.

3. Not one future

Nobody knows what a plant will cost in 2050, so each technology carries a range. There are two families of paths.

The reference family (capex_mid, capex_hi, capex_lo) follows institutional projections. The central path is the IEA regional value. High and low come from the ratio of NREL's Conservative and Advanced cases to its Moderate case:

capex_hi(region, year) = capex_mid(region, year) × ATB Conservative / ATB Moderate
capex_lo(region, year) = capex_mid(region, year) × ATB Advanced     / ATB Moderate

Technologies with no NREL counterpart have a central path only.

The learning family (capex_l0, capex_l1, capex_l2) exists for solar, onshore wind, offshore wind and batteries. Cost falls by a fixed percentage with every doubling of what the world has built (Wright's law). The three paths are slow, central and fast learning, anchored on observed costs. Regions sit at a premium above a global frontier.

Technology and family Low / fast Central High / slow
Nuclear, reference 2,401 4,234 7,770
Solar, learning 303 467 616

Examples, European Union, capital cost in 2050 (USD 2022 per kW)

Learning is an input, not a result. The model does not lower costs because it builds more: learning depends on world deployment, which one country's model does not decide.

4. Same plant, different place

Costs come in the nine regions of the IEA dataset: United States, European Union, Japan, Russia, China, India, Middle East, Africa and Brazil.

Region Capital cost
China 496
India 620
European Union 885
United States, Japan 974

Gas combined cycle built in 2030, central path (USD 2022 per kW)

Each country takes the costs of its region, through the ep_regionmap sheet. The labels are the IEA's and are broader than they sound: "United States" also serves Canada, Mexico and Australia; "Brazil" serves Latin America.

Technologies from the Danish catalogues are priced for Denmark. They are moved to other regions with the regional ratio of a comparable IEA technology. In the reference family, battery costs are the same in every region.

5. The cost of money

A plant is paid for over decades, so the rate at which money is borrowed matters as much as the price. The module builds a cost of capital for each country and technology (source_tables/std_wacc.csv):

cost of capital = cost for the industry, worldwide   (Damodaran, global industry file)
                + premium for the country            (Damodaran, country risk premiums)

Rates are real: the nominal rate less 2.3 points of inflation.

Country Gas plant Coal plant
Germany 3.5% 6.2%
India 5.4% 9.5%

Real cost of capital

Coal plants carry the rate of the coal industry, not of power generation, on the view that the risk being priced is exposure to the fuel. This is an interpretation, and it adds 2.7 to 4.1 points over gas depending on the country.

An alternative file takes renewable rates from IRENA's survey of financing costs.

The rate enters the model as a technology-specific discount rate, separate from the capital cost.

6. Where it can go

A new thermal plant is not offered everywhere. It is offered at every node that already hosts a coal, gas or nuclear unit above the country's size threshold for that fuel: places where fuel supply, cooling and grid connection exist.

Design Nodes open to new thermal plants Of
Germany kan20 27 45
India kan50 40 50

The rule is not fuel-specific. A node that qualifies is offered every thermal option. Batteries are offered at every node that hosts a unit of any fuel above its threshold. See Grid Representation.

Lifetime and construction time

These are constants by technology.

Technology Lifetime (years) Construction (years)
Gas combined cycle 30 3
Gas turbine 30 2
Gas with carbon capture 30 4
Coal, all types 40 4
Nuclear 60 7
Small modular reactor 40 3
Biomass 45 3
Biomass with carbon capture 30 3
Heat and power plants 25 2 to 5
Heat-only plants and heat pumps 20 to 25 1 to 3
Batteries 20 1

The cost table

unified_reference_costs.csv has one row per technology, region, year and parameter.

Column Description
variable Technology, for example en_gas_ccgt, en_battery_4h
region One of nine, for example ieapg_european_union
year The source's own years between 2020 and 2050. Values are not interpolated; the model interpolates between its own milestone years.
parameter Capital cost paths (capex_mid, _hi, _lo, _l0, _l1, _l2), fixed_om, variable_om, efficiency, heat_efficiency, the heat-and-power coefficients, capture_rate, roundtrip_efficiency, lifetime, construction_time
value, unit The number and its unit, for example USD2022/kW

Columns

File Role
extractors/ One reader per source, each writing a standard table to source_tables/
compile_unified_reference.py Assembles the cost table from the standard tables
learning_curves.py Builds the learning family
source_versions.csv Version, hash and extraction date of every input
source_tables/std_wacc.csv Cost of capital by country and technology
excel_manager.py Reads the cost table into each country model

Files

Limits to know about

  • The country builds use the central reference path. Scenario files for the other paths exist for the KiNESYS models and are being connected to the country builds.
  • Cost of capital is not yet read by the country builds.
  • Region labels are coarse. Nine regions cover the world.
  • Source terms. The cost table is derived from published sources, some of which restrict redistribution of their data. The table is used inside the models; it is not offered as a dataset.

See also

Existing Stock Characterization

Costs and performance of plants that already exist

Grid Representation

Nodes, and where new plants are offered