AN INTERACTIVE GUIDE TO FUSION ENERGY
Reactor designs, experimental results, and the engineering of fusion power
01 / FROM NUCLEI TO A POWER PLANT
Fusion joins light atomic nuclei to form heavier ones, releasing energy. This example follows deuterium and tritium (D–T), two forms of hydrogen, from the reaction to electricity exported by a power plant
Fully fusing one gram of D–T releases roughly as much energy as burning eight tonnes of oil. Yet no fusion power plant has produced electricity, and building one remains a major scientific and engineering challenge. Researchers and companies are pursuing it because it could help meet the world's growing energy needs with electricity day and night, helping replace our use of fossil fuels.
Deuterium–tritium (D–T) needs less demanding plasma conditions than other candidate fuels, making it a leading choice for proposed fusion power plants. However, its energetic neutrons damage materials, and a plant must produce fresh tritium from lithium.
Compare D–T, deuterium–deuterium, deuterium–helium-3, and proton–boron-11: what each fuel offers and what it takes to use it
Today's nuclear power stations use fission. Fusion is a different nuclear reaction, still being developed for electricity generation. Both release energy from atomic nuclei — the centres of atoms
A neutron can split a heavy nucleus such as uranium, releasing energy and more neutrons. Those neutrons can split other nuclei, sustaining a chain reaction that a power station controls
Spent fuel remains radioactive and needs long-term management
In D–T fusion, deuterium and tritium combine into helium and a neutron, releasing energy. Keeping the fuel hot and confined enough to sustain fusion is one of the central challenges
Tritium is radioactive, and fusion neutrons can make reactor materials radioactive too
Different reactions, a familiar route to electricity: fission plants and most proposed fusion plants use the released energy as heat to drive a turbine and generator. Reaction sketches above are schematic, not to scale
More energy from each gram. Fully reacting D–T fuel releases about four times as much energy per gram as fissioning uranium-235, and millions of times more than burning fossil fuels. These are fuel-level comparisons; electricity output would be lower after conversion losses and plant power use
Potentially shorter-lived waste. D–T fusion needs no uranium or plutonium fuel and produces no spent fission fuel. Choosing suitable reactor materials could make its radioactive waste shorter-lived and easier to manage. It is not waste-free: radioactive tritium and reactor components made radioactive by neutrons still need careful handling and disposal
2025 · Approximate share of global electricity generation
Inside renewables · Share of all electricity generation, 2025
Together ≈34% of world electricity. Figures rounded independently. Other renewables includes bioenergy and waste, geothermal, concentrating solar power and marine energy. IEA · Renewable generation by source ↗
Fusion: 0% No fusion power station supplies electricity to the grid yet
IEA · Global Energy Review 2026 ↗ · Rounded estimates. Electricity generation only; direct fuel use for transport and heating is outside this chart
The IEA projects electricity use by all data centres to nearly double by 2030, reaching about 3% of global electricity demand. AI-focused facilities are growing faster
AI's appetite for electricity has helped renew interest in fusion. Google, for example, has invested in fusion and agreed to buy electricity from a future plant. Fusion could eventually serve computing, industry and electrified transport, but today's data centres rely on existing energy sources
IEA · Key Questions on Energy and AI, 2026 ↗
Includes AI and other computing. One TWh = one billion kWh
Coal, gas, oil and nuclear fuels store energy in matter. Solar, wind and hydro collect natural energy flows, so energy per gram of fuel is not a useful comparison across all eight
| Source | Electricity supply | Main constraints |
|---|---|---|
| Fusion · D–T | Experimental today; no grid electricity yet. Proposed plants would use heat from joining light nuclei | Net electricity, tritium breeding, durable materials and economic operation remain to be demonstrated together |
| Fission | Operating power stations use heat from splitting heavy nuclei, without needing sunlight or wind | Construction time and cost, cooling, and long-term spent-fuel management |
| Coal | Burns mined fuel to make steam; generation can be scheduled | CO₂ emissions, air pollution, ash and mining impacts |
| Natural gas | Burns gas in a turbine; combined-cycle plants also use the exhaust heat to make steam and generate more electricity | CO₂ emissions, methane leaks along the supply chain, and dependence on fuel supply and prices |
| Oil | Burns liquid fuel in engines, turbines or boilers; stored fuel supports generation on demand | CO₂ emissions, air pollution and a continuing supply of fuel |
| Solar · PV | Converts sunlight directly into electricity; output follows daylight and weather | Variable output; storage, transmission and other generators help balance supply |
| Wind | Moving air turns a generator; output varies with wind conditions | Grid balancing, suitable sites and wildlife impacts |
| Hydro | Moving water turns a turbine; reservoirs can store water and adjust output | River geography, water availability and effects on river ecosystems |
The fuel comparison assumes a D–T mixture with equal numbers of deuterium and tritium nuclei, fully reacted. At 17.6 MeV per reaction, one gram releases about 340 GJ, or 94,000 kWh of heat. Electricity output would be lower after conversion losses and the plant's own power use. Lithium for tritium breeding and the plant's materials are additional inputs.
02 / REACTOR CONFIGURATIONS
Select a configuration, rotate the model, and inspect its components
03 / ENERGY ACCOUNTING
A fusion power plant must generate more electricity than it consumes. Each measurement below counts a different part of the process; only the whole-plant balance tells us what can reach the grid
ITER is an international experimental fusion reactor being built in southern France by seven partners, including the European Union. It aims for 500 megawatts (MW) of fusion power from 50 MW of heating delivered to the plasma.
500 MW ÷ 50 MW = Q of 10
MW measures power: the rate of energy production or use.
Q = 10 is a research goal. ITER has no system for generating electricity.
The National Ignition Facility (NIF) is a laser research facility at Lawrence Livermore National Laboratory in California. On 7 April 2025, it released 8.6 ± 0.45 megajoules (MJ) of fusion energy from 2.08 MJ of laser light delivered to a target.
8.6 MJ ÷ 2.08 MJ ≈ 4.13 target gain
This is the highest target gain in this snapshot. MJ measures an amount of energy.
NIF does not generate electricity. Its lasers and other systems consume electricity outside this ratio.
Net electricity has not been demonstrated.
In this illustrative example, a pulse uses 100 MJ of electricity and the recovery system returns 102 MJ. Pumps, cooling and other systems consume another 3 MJ, leaving a net loss of 1 MJ.
102 − 100 − 3 = −1 MJ
The whole plant must finish with a surplus to supply the grid, even when the pulse recovers more than its own input.
The model excludes useful recovery of supplied heating energy, blanket multiplication, transient storage, maintenance and capital costs. Auxiliary demand excludes the driver counted separately. Finite Q with zero external heating does not model ignition.
Compare results within each measurement and operating regime
04 / THE FUSION ATLAS
Research programs, reactor developers, and suppliers, with reported experiments, design studies, and future targets identified separately
Each company is counted once by its primary approach. Public programmes and supplier-only companies are excluded. Select a bar to see the companies
05 / ENGINEERING CHALLENGES
Explore what can stop a fusion experiment from becoming a working power plant—and what engineers have to solve
Research snapshot · 22 October 2026
Experiments report observations, models describe simulations or design studies, and targets describe announced plans. Passing a target date does not establish that it was met. Component tests and predecessor experiments are identified separately.
Sources include scientific publications, public research programs, and company announcements. Company-reported results are attributed and are not treated as independent verification. The label beside each evidence badge identifies the cited source, such as a company source, preprint, or peer-reviewed paper. A company summary of a paper is labeled as a company source.
This selection covers 42 organizations and 15 reactor configurations. It is not a census of every fusion project. Profiles link to the sources behind each claim; the downloadable dataset preserves dates, measurement boundaries, and qualifications.
Map positions are approximate city or area centers; unknown or unselected sites are left unmapped. Financing records preserve currencies, dates, revisions, and conditions.