Fusion in the energy system
Why is it so hard to say whether fusion will ever power our homes?
Prior knowledge
- D–T fusion releases $17.6\ \text{MeV}$, of which $3.5\ \text{MeV}$ goes to the helium nucleus and $14.1\ \text{MeV}$ to the neutronchapter Fusion reactions: from the Sun to deuterium and tritium
- The fusion gain $Q$ compares fusion power with heating power, not with electricitychapter Inside ITER: the machine and fusion gain
Core path: follow the steps in order. Tasks marked Optional give extra practice. Optional tasks: 1.
Think about it
Every few months a headline says that fusion has made a breakthrough. Laboratories report new records, governments announce power plants for the 2040s, and companies promise electricity even sooner. At the same time, the world needs to cut its carbon dioxide emissions now, and wind and solar power are growing fast.
If fusion is so powerful, why can nobody tell you the year when it will power your home?
Review: what Q means
Sentence starters
- I think fusion is still not used because …
- One problem could be …
- I have read / heard that …, but I am not sure if …
- Maybe the difference between … and … is important.
Think about it: when will fusion arrive?
Discuss the opening question with a partner. List two or three reasons why nobody can give a reliable year for fusion electricity. Write two or three sentences. It is fine to be unsure: you will test your ideas today.
Pair work: share your ideas aloud first, then each of you writes your own answer.
Model answer
I think nobody can give a year because no fusion machine has yet produced electricity, so we do not know how long the engineering will take. One problem could be the fuel: tritium is rare and radioactive. Another reason may be money: fusion machines are very expensive, and their costs and schedules have grown before. Maybe it is also important who says the year: a company that needs investors may be more optimistic than a laboratory.
Assess your own answer
Tick the points your own answer already contains.
Review the theory summary.
Predict
Make your predictions before investigating 2 questions
Prediction 1: how much fuel?
Prediction 2: what is radioactive?
Investigate
The world's energy challenge
People use energy for heat, transport, industry and electricity. In 2024 the world used roughly $590\ \text{EJ}$ of primary energy (energy in the form we find it, such as coal, oil, wind or uranium), and more than four fifths of it still came from fossil fuels. Burning them releases carbon dioxide, the main cause of climate change. The IPCC concluded in 2018 that, to limit warming to $1.5\ ^\circ\text{C}$, net carbon dioxide emissions must reach net zero around 2050.
Electricity is a growing part of the answer: electric cars, heat pumps and data centres all use it. In 2025, world electricity demand grew by $2.8\ \%$. An electricity grid must match supply and demand every second. Some sources are dispatchable: operators can switch them up or down when needed (gas, hydro, nuclear). Wind and solar power are variable: their output depends on the weather and the time of day. A system with much wind and solar also needs storage, flexible demand, strong grids or firm low-carbon power that runs whatever the weather.
Where did the world's electricity come from in 2025?
Statistical data · Ember, Global Electricity Review 2026 and Ember world data (shares of 2025, read 2026-11-01); the row other is the remainder computed from the listed shares.
| Source | Share in 2025 | Low-carbon? | Weather-dependent? |
|---|---|---|---|
| coal | $33.0\ \%$ | no | no |
| gas | $21.8\ \%$ | no | no |
| hydro | $14\ \%$ | yes | partly (rain) |
| nuclear (fission) | $8.9\ \%$ | yes | no |
| solar | $8.7\ \%$ | yes | yes |
| wind | $8.5\ \%$ | yes | yes |
| other¹ | $5.1\ \%$ | mixed | no |
| fusion | $0\ \%$ | yes | no |
¹ Bioenergy, oil and other sources: the remainder of the listed shares. Ember gives hydro only to the nearest per cent.
Word bank
demand · supply · dispatchable · variable · firm power · storage · grid · low-carbon · fossil fuel · to balance · to replace · to complement · in addition to
Sentence starters
- The table shows that …
- Wind and solar are useful because …, but …
- A fusion plant could complement … because …
- A limitation of fusion is …
- Fusion cannot help before … because …
Discuss: what could fusion add?
Discuss with a partner, then write four or five sentences. Use at least one number from the table.
What kind of power does a grid with a lot of wind and solar need? What could a fusion power plant add, and why can it not help with today's emissions?
Pair work: one of you argues for a role for fusion, the other looks for problems; then each of you writes.
Model answer
The table shows that coal and gas still gave more than half of the world's electricity in 2025, while wind and solar gave about one sixth, $17.2\ \%$. Wind and solar are useful because they are low-carbon, but their output depends on the weather, so the grid also needs storage and firm power. A fusion plant could complement them, because it would be low-carbon and could run whatever the weather, like a fission plant. A limitation of fusion is that no fusion plant has produced electricity yet, and the UK's public prototype STEP targets 2040, so climate plans for 2030 cannot count on it.
Assess your own answer
Tick the points your own answer already contains.
Review the theory summary.
Results, plans and claims
News about fusion mixes three kinds of statements. Learn to tell them apart.
- An established result has already happened and has been measured, usually reported by a laboratory and checked by other scientists. Example: on 5 December 2022 the National Ignition Facility (NIF) released $3.15\ \text{MJ}$ from a fuel target hit by $2.05\ \text{MJ}$ of laser light.
- An institutional plan is what a government, an international organisation or a public laboratory intends to do. Plans change: ITER's 2024 baseline moved deuterium–tritium operation to 2039.
- A corporate claim is a goal or forecast announced by a private company. Companies need investors, so their announcements are also advertising. A claim is not necessarily false, but it is not yet evidence.
A good reader also asks when a statement was made. Every changing fact in this unit carries its date.
Uncertainties: engineering, money and time
No fusion device has yet delivered electricity to a grid (as of November 2026). The engineering questions include materials that survive years of fast neutrons, breeding enough tritium, removing intense heat, and keeping a machine running reliably for months. Scientific gain $Q$ compares fusion power with heating power only; a power plant must also pay for its magnets, lasers, pumps and conversion losses.
The economic questions are just as open. In 2006 ITER's construction was estimated at $5.9$ billion euros; later estimates of construction and operation together are $18\text{–}22$ billion euros. Nobody can yet give the price of fusion electricity: published figures for future power plants are estimates, not prices. Money comes from two directions: public funding (ITER, national programmes such as the UK's £2.5 billion announced in June 2025) and private investment. In a 2026 survey of 56 fusion companies, the Fusion Industry Association counted USD 14.24 billion raised in total, USD 4.48 billion of it in the year to July 2026.
Evaluate: result, plan or claim?
All nine statements were really reported or announced; decide what kind of statement each one is.
Sentence starters
- … is an established result, because it has already …
- … is only a plan, because the words … show that …
- This is a corporate claim: the company …
- I was not sure about …, because …
- The date matters here because …
Justify two of your choices
Discuss with a partner, then write three or four sentences. Choose one statement you found easy and one you found difficult to sort. For each one, explain which words or facts decided the category.
Why does the date of a statement matter?
Pair work: compare your sorting with a partner and talk about any statement you sorted differently.
Model answer
The NIF statement is an established result, because it has already happened: the energies were measured on 5 December 2022 by a national laboratory. The survey of fusion companies was harder, because the survey itself is real, but it only reports what the companies expect. So it is a corporate claim: the words expect and by the 2030s show that it is about the future. The date matters because plans change; ITER's schedule has moved several times.
Assess your own answer
Tick the points your own answer already contains.
Review the theory summary.
Theory summary
Key idea
Fusion could add firm, low-carbon power to a future energy system. It would emit no carbon dioxide while running, and its fuel is plentiful. But it competes with sources that already work and are getting cheaper: wind, solar, hydro and fission. Whether fusion becomes part of the system depends on results that have not happened yet: reliable machines, enough tritium, durable materials and an affordable price. So the honest answer to the opening question is a conditional one: fusion could help, if these problems are solved. No fusion plant has yet delivered electricity, and the public prototype plans (STEP, 2040) come after the 2030 target, so climate plans cannot count on fusion for the cuts needed by 2050.
Fuel: deuterium, lithium and tritium
A deuterium–tritium (D–T) power plant burns two isotopes of hydrogen: ${}^{2}_{1}\text{H} + {}^{3}_{1}\text{H} \to {}^{4}_{2}\text{He}\ (3.5\ \text{MeV}) + \text{n}\ (14.1\ \text{MeV})$. Deuterium is in all water: about one hydrogen atom in 6420 in sea water is deuterium, about $0.035\ \text{g}$ in every litre. The energy per kilogram is enormous. In its last experiments in 2023, JET released $69.26\ \text{MJ}$ of heat in one pulse from only $0.21\ \text{mg}$ of fuel, about as much energy as $1.5\ \text{kg}$ of petrol gives when it burns.
Tritium is the problem. It is radioactive, with a half-life of $12.3$ years (it decays to helium-3 by beta decay), so almost none exists in nature. The world's civilian stock was only about $25\ \text{kg}$ in 2024, mostly from Canadian fission reactors. A power plant must therefore breed its own tritium. A breeding blanket that contains lithium surrounds the plasma: a fusion neutron and a lithium-6 nucleus give a new tritium nucleus and helium, ${}^{6}_{3}\text{Li} + \text{n} \to {}^{3}_{1}\text{H} + {}^{4}_{2}\text{He}.$ The blanket must make slightly more than one new tritium nucleus for every one burned; this has not yet been shown at power-plant scale, and ITER will test small test blanket modules.
Fusion and fission compared
| Property | Fission | Fusion (D–T) |
|---|---|---|
| reaction | heavy nuclei split | light nuclei join |
| fuel | uranium | deuterium, lithium |
| in electricity (2025) | $8.9\ \%$ | none yet |
| chain reaction | yes, controlled | no |
| main waste | spent fuel, long-lived | activated structure |
In a fission reactor, a chain reaction of neutrons keeps going and must be controlled. A fusion plasma holds only a few grams of fuel at a time, and any disturbance cools it within seconds, so it cannot run away. The product of the reaction, helium, is not radioactive.
Fusion still has real hazards. Tritium can escape into the environment, so it must be contained. The $14.1\ \text{MeV}$ neutrons make the steel and other materials of the machine radioactive. This neutron activation gives large volumes of low- and intermediate-level waste, but no long-lived high-level waste from the fuel. Most of the activity fades within about 100 years. Strong magnets, very cold helium and great heat are further, industrial hazards.
Words of the lesson
Check the key words of the lesson.
Back to your predictions
Compare your predictions with the theory. Then answer in one or two sentences: what makes fusion fuel attractive, and what makes tritium a problem? You can start with Fusion fuel is attractive because … and Tritium is a problem because …
Would you change your prediction?
Apply
Worked example
Problem
How much energy is released by the fusion of one kilogram of D–T fuel? Compare it with petrol, which gives $46.4\ \text{MJ}$ per kilogram when it burns.
Step 1/7
Given: One D–T reaction releases ${E_1 = 17.6\ \text{MeV}}$. The masses of the nuclei are about ${m_\text{D} = 2.014\ \text{u}}$ and ${m_\text{T} = 3.016\ \text{u}}$, with ${1\ \text{u} = 1.661 \cdot 10^{-27}\ \text{kg}}$ and ${1\ \text{MeV} = 1.602 \cdot 10^{-13}\ \text{J}}$. Petrol: ${e_\text{petrol} = 46.4\ \text{MJ/kg}}$. We want the energy per kilogram $e_\text{DT}$ and the ratio $e_\text{DT} : e_\text{petrol}$.
Step 2/7
Principle: Each reaction uses one deuterium and one tritium nucleus, so the fuel mass of one reaction is ${m_1 = m_\text{D} + m_\text{T}}$. The energy per kilogram is the energy of one reaction divided by the fuel mass of one reaction. The electrons' mass is very small and is left out.
Step 3/7
Formula: Solved in symbols:
$$\begin{aligned} e_\text{DT} &= \frac{E_1}{m_\text{D} + m_\text{T}} \\[6pt] \text{ratio} &= \frac{e_\text{DT}}{e_\text{petrol}} \end{aligned}$$
Step 4/7
Unit conversions: The energy of one reaction in joules: ${E_1 = 17.6 \cdot 1.602 \cdot 10^{-13}\ \text{J}}$ ${= 2.820 \cdot 10^{-12}\ \text{J}}$. The fuel mass in kilograms: ${m_1 = 5.030\ \text{u}}$ ${= 5.030 \cdot 1.661 \cdot 10^{-27}\ \text{kg}}$ ${= 8.355 \cdot 10^{-27}\ \text{kg}}$. Petrol: ${46.4\ \text{MJ/kg} = 4.64 \cdot 10^{7}\ \text{J/kg}}$.
Step 5/7
Substitution: Energy per kilogram, then the ratio:
$$\begin{aligned} e_\text{DT} &= \frac{2.820 \cdot 10^{-12}\ \text{J}}{8.355 \cdot 10^{-27}\ \text{kg}} \\ &= 3.3747 \cdot 10^{14}\ \text{J/kg} \\[6pt] \text{ratio} &= \frac{3.3747 \cdot 10^{14}\ \text{J/kg}}{4.64 \cdot 10^{7}\ \text{J/kg}} \\ &= 7.27 \cdot 10^{6} \end{aligned}$$
Step 6/7
Answer: One kilogram of D–T fuel releases about ${3.37 \cdot 10^{14}\ \text{J}}$, about seven million times as much as one kilogram of petrol. Three significant figures fit, because the reaction energy $17.6\ \text{MeV}$ has three.
Step 7/7
Reasonableness: JET's record pulse released $69.26\ \text{MJ}$ from $0.21\ \text{mg}$ of fuel, which is ${69.26 \cdot 10^{6}\ \text{J} : 0.21 \cdot 10^{-6}\ \text{kg}}$ ${\approx 3.3 \cdot 10^{14}\ \text{J/kg}}$, the same as our result. A chemical reaction releases a few electronvolts per molecule, a nuclear reaction millions of electronvolts, so a ratio of millions is expected. (ITER's website says "nearly four million times" compared with coal, oil or gas; such comparisons depend on the fuel chosen and on what mass is counted, but every version gives millions.)
How much tritium does a power plant burn?
A fusion power plant releases a fusion power of $1.0\ \text{GW}$ for one year, $3.156 \cdot 10^{7}\ \text{s}$. Each D–T reaction releases $17.6\ \text{MeV}$ and burns one tritium nucleus of mass $3.016\ \text{u}$.
How many kilograms of tritium does the plant burn in a year? Compare your answer with the world's civilian stock of about $25\ \text{kg}$ (2024).
Hint 1: a guiding question
First find the energy released in a year: $E = Pt$.
Hint 2: what is the connection
The number of reactions is $N = E : E_1$, where $E_1$ is the energy of one reaction in joules.
Hint 3: an intermediate step
$E_1 = 17.6 \cdot 1.602 \cdot 10^{-13}\ \text{J}$ $= 2.820 \cdot 10^{-12}\ \text{J}$ and $E = 3.156 \cdot 10^{16}\ \text{J}$, so $N = 1.119 \cdot 10^{28}$.
Show the solution
Given
Fusion power ${P = 1.0\ \text{GW} = 1.0 \cdot 10^{9}\ \text{W}}$, time ${t = 3.156 \cdot 10^{7}\ \text{s}}$, energy of one reaction ${E_1 = 17.6\ \text{MeV}}$, mass of a tritium nucleus ${m_\text{T} = 3.016\ \text{u}}$. We want the mass of tritium $m$ burned in a year.
Principle
Every reaction burns one tritium nucleus. The number of reactions is the total energy divided by the energy of one reaction, and the mass is that number times the mass of one nucleus.
Formula
$$\begin{aligned} N &= \frac{Pt}{E_1} \\ m &= N m_\text{T} = \frac{Pt\, m_\text{T}}{E_1} \end{aligned}$$
Unit conversions
${E_1 = 17.6 \cdot 1.602 \cdot 10^{-13}\ \text{J}}$ ${= 2.820 \cdot 10^{-12}\ \text{J}}$ and ${m_\text{T} = 3.016 \cdot 1.661 \cdot 10^{-27}\ \text{kg}}$ ${= 5.010 \cdot 10^{-27}\ \text{kg}}$.
Substitution
$$\begin{aligned} N &= \frac{1.0 \cdot 10^{9}\ \text{W} \cdot 3.156 \cdot 10^{7}\ \text{s}}{2.820 \cdot 10^{-12}\ \text{J}} \\ &= 1.119 \cdot 10^{28} \\[6pt] m &= 1.119 \cdot 10^{28} \cdot 5.010 \cdot 10^{-27}\ \text{kg} \\ &= 56.1\ \text{kg} \end{aligned}$$
Answer
The plant burns about ${m \approx 56\ \text{kg}}$ of tritium a year. Two significant figures are enough, because the power $1.0\ \text{GW}$ has two.
Reasonableness
One gigawatt for a year is $3.2 \cdot 10^{16}\ \text{J}$; with $3.4 \cdot 10^{14}\ \text{J}$ per kilogram of D–T fuel this is about $94\ \text{kg}$ of fuel, and tritium is $3.016 : 5.030 \approx 60\ \%$ of that mass, about $56\ \text{kg}$. It is more than twice the world's civilian stock of about $25\ \text{kg}$: a power plant cannot buy its tritium, it must breed it.
Optional Find the error
Common misconception
"Fusion is clean, so it makes no radioactive waste." This is false. The fuel's product, helium, is harmless, but tritium is radioactive and the neutrons activate the machine itself. The waste is short-lived compared with spent fission fuel, not absent.
A second mistake is to read a plan or a company's goal as a result. "Will deliver electricity in 2028" is a claim about the future. Only a measured, checked event is an established result.
Check your understanding
1. Concept: can a fusion plant run away?
2. Calculation: tritium decays
3. Evaluate a source
Word bank
established result · institutional plan · corporate claim · scientific gain · net electricity · tritium · breeding · uncertain · evidence · to overstate
Sentence starters
- The headline is partly right, because …
- However, … is only a plan / a claim, not a result.
- No fusion device has yet …
- A more accurate headline would be …
4. Write: evaluate a headline
Imagine a news site writes in 2026: "Fusion breakthrough: clean energy for everyone by 2035!" Write five to seven sentences that evaluate the headline. Say what is an established result, what is a plan or a claim, and what problems are still unsolved. End with a more accurate headline of your own.
Model answer
The headline is partly right, because fusion experiments have produced real results: NIF released more energy than its laser light delivered in 2022, and JET released $69\ \text{MJ}$ in one pulse in 2023. However, "by 2035" is only a plan or a claim, not a result, and no fusion device has yet delivered electricity to a grid. ITER, for example, plans deuterium–tritium operation only from 2039, and it will not make electricity. Fusion also has unsolved problems: breeding enough tritium, materials that survive the neutrons and an affordable cost. "Clean" is also too simple, because the machine becomes radioactive, although the waste is short-lived. A more accurate headline would be: "Fusion records grow, but no one can yet say when fusion electricity will come."
Assess your own answer
Tick the points your own answer already contains.
Review the theory summary.
Summary
- The world still gets more than four fifths of its primary energy from fossil fuels; climate targets need net zero emissions around 2050.
- A grid needs both variable sources (wind, solar) and dispatchable or firm power; fusion could become a firm, low-carbon source.
- D–T fuel releases about $3.4 \cdot 10^{14}\ \text{J}$ per kilogram, millions of times more than chemical fuels. Deuterium is plentiful; tritium (half-life $12.3$ years) must be bred from lithium in a blanket.
- Fusion cannot run away like a fission chain reaction, and it makes no long-lived high-level waste from its fuel, but tritium and neutron activation are real hazards.
- No fusion device has yet delivered electricity (November 2026). Separate established results, institutional plans and corporate claims, and look at the date.
Le coin français
France hosts ITER, so energy debates there often include la fusion. Listen, repeat, and say the last sentence to your partner. In the hints, capitals show the stressed syllable, and õ is a nasal vowel: the air goes through the nose and no n is heard.