On 8 October 2026, the Canadian company General Fusion announced a "world first": its machine had heated plasma past 1 kiloelectronvolt — roughly 12 million degrees. Two days later, the Princeton Plasma Physics Laboratory announced a paper by its physicists in Physical Review Letters that headlines quickly called a "shortcut to fusion energy".
Both stories are true. Neither means a fusion power plant is a few years away. To see why this keeps happening, it helps to picture the road to a power plant as a staircase — and to ask which step each headline actually stands on.
Why the world wants fusion
Electricity demand is rising. In its 2026 mid-year update, the International Energy Agency (IEA) expects global consumption to grow by 3.6% this year and 3.8% in 2027. The drivers include industry, household appliances, air conditioning, electric vehicles and new data centres.
Fusion promises a source that emits no carbon dioxide while running, does not depend on the weather, and runs on fuel with enormous reserves: deuterium comes from water. Unlike a fission reactor, the reaction cannot run away. Disturb the conditions and it simply goes out.
That promise is why every step in the field draws so much attention. It is also why it matters to know how big each step really is.
How fusion works
The Sun shines because hydrogen nuclei fuse into helium and release energy. On Earth, the easiest reaction fuses two heavy forms of hydrogen: deuterium and tritium.
Three conditions have to be met at once. The fuel must be extremely hot — around 100 million degrees, several times hotter than the Sun's core. It must be dense enough for nuclei to collide often. And it must be held long enough that the energy doesn't escape before the reactions can replace it.
Their product, the triple product, has to pass a threshold known as the Lawson criterion. It is the field's main yardstick: a machine can set a temperature record and still fall short on density or time.
There are two main ways to hold such a plasma:
- magnetic — trapping it in strong magnetic fields (tokamaks and stellarators);
- inertial — using lasers to crush a tiny fuel capsule in billionths of a second, so it ignites before it flies apart.
There are hybrids too. General Fusion uses magnetized target fusion: plasma held in a magnetic field is compressed from the outside.
The staircase to the socket
Here is a simplified sequence of steps every approach has to climb. Each one is much harder than the last.
| Step | What it means |
|---|---|
| 1. Temperature | The plasma is hot enough |
| 2. Lawson | Temperature, density and time are enough together |
| 3. Scientific gain | Reactions release more energy than went into the fuel |
| 4. Engineering gain | The whole plant gives back more than it draws from the grid |
| 5. Steady operation | For months, without stops, breeding its own tritium |
| 6. Power plant | Reliable and at an acceptable cost |
As of October 2026, nobody has climbed above step three — and only one facility has reached step three, in a rather special sense.
What NIF actually showed
On 5 December 2022, the National Ignition Facility (NIF) at Lawrence Livermore National Laboratory in the US became the first to get more energy out of fusion than its lasers delivered to the fuel: 3.15 megajoules out against 2.05 in. It was called "ignition", and it was a genuine scientific achievement.
The result has improved since. The record is 8.6 megajoules from 2.08 megajoules of laser energy — about four times more — set on 7 April 2025. By June 2026, NIF had achieved ignition 11 times.
But there is a large catch. To produce that first 2.05-megajoule pulse, NIF's lasers drew about 300 megajoules from the grid. Even the record shot returns only a few per cent of the energy the whole facility consumes. NIF also fires such shots one at a time, with long preparation in between, while a laser power plant would need — in common designs — around ten shots every second.
NIF was not designed as a prototype power plant: besides ignition, much of its work serves research for the US nuclear stockpile. It proved that step three can be reached. Step four is a different problem.
October's three headlines on the staircase
General Fusion: step one. The company reported an electron temperature of 1.1 keV on its LM26 machine. The measurement was made together with the UK Atomic Energy Authority (UKAEA), which built a key part of the diagnostic — so this is not a fully independent check. A technical paper has been submitted for peer review but not yet published.
"World first" is narrower than the headline suggests. Many machines have passed 1 keV before: large tokamaks routinely reach temperatures ten times higher. General Fusion is first only among machines that compress plasma slowly. A reactor needs roughly ten times more, which is the company's next target, with no date attached. It plans a first plant around 2035 — the company's own plan, not an independent estimate.
Princeton: theory, not experiment. The paper by Luis Delgado-Aparicio and colleagues suggests that in a tokamak it may pay to heat the plasma first and raise its density afterwards, rather than the other way round. That route could reach ignition with less energy.
But it is a calculation, and the authors list its limits. The approach has not been tested experimentally; current machines don't reach the temperatures involved; and tungsten impurities of one part in ten thousand could roughly double the pressure needed. The paper refines the map of the staircase. It doesn't move anyone up it.
The US Department of Energy programme: engineering bricks. Under a programme that pays companies only for independently verified milestones, Zap Energy had two capacitor milestones approved — parts that must keep up with a power plant's pulse rate; the production design is projected to last 6 billion pulses. It isn't plasma news, but it is exactly these "boring" components that decide whether anyone reaches step five.
Big machines and real timetables
The world's largest project is ITER, an international tokamak in southern France. Its schedule was revised in 2024: research operations are now due to begin in 2034, and deuterium–tritium operation in 2039. ITER aims to produce 500 megawatts of fusion power from 50 megawatts of heating.
But ITER won't produce electricity either. It is an experiment, not a power plant.
The private company Commonwealth Fusion Systems is assembling a compact tokamak, SPARC, in Massachusetts. It hopes for first plasma in 2027 and scientific gain soon after. The original schedule said 2025.
If it works, SPARC would be the first magnetic machine on step three. For now, that is the company's forecast.
Why 'breakthroughs' arrive every year
What follows is interpretation, not measurement.
First, the field has many measurable intermediate records: temperature, duration, density, magnet strength, the best result for a given machine type. Each record is real, and each can honestly be called "the first" if the category is drawn narrowly enough.
Second, different approaches race in parallel. A laser fusion record says nothing about tokamaks, and vice versa. "First to achieve X" often means "first to achieve X this way".
Third, private companies have to show progress. Funding programmes like the American one are explicitly built around milestones. That is a useful discipline, but it also rewards loud wording.
None of this means the field is standing still. It is moving — the staircase is simply longer than the headlines make it look.
What we still don't know
- Materials. A reactor's walls will be bombarded by fast neutrons for years. Which materials can survive that, and for how long, has not been fully tested.
- Tritium. There is almost none in nature. A plant must breed its own from lithium, and nobody has yet shown this at the scale needed.
- Continuous operation. Every record so far is a pulse, lasting from billionths of a second to about 22 minutes, the record set by the WEST tokamak in France in February 2025. A power plant must run for months.
- Cost. Even if the technology works, it's not known whether fusion electricity can compete with other sources.
Three plausible scenarios
Slow but real. One or two machines reach scientific gain by the end of the 2020s, demonstration plants appear in the 2030s, and fusion makes up a noticeable share of power systems no earlier than the second half of the century. Condition: the engineering problems with materials and tritium are solved without major surprises.
It works, but it's expensive. Fusion becomes technically possible but stays a niche source for countries and regions that especially need steady generation.
Delays. Timetables slip again, and the energy transition of the coming decades rests on what already exists: renewables, storage, grids and fission plants. Fusion remains science's long bet.
In every scenario, one thing is clear: fusion will not solve the rise in electricity demand over the next 10–15 years.
How to read the next fusion headline
Four questions that show how big the step really is:
- Which step is it? Temperature, Lawson, fuel gain, or gain for the whole plant?
- Who measured it? The company, a partner, or an independent group? Is there a peer-reviewed paper?
- Is it an experiment or a calculation?
- "First" in which category? In the world, or among one type of machine?
Questions for discussion
- Why do you think "breakthrough" stories are so appealing, even when the practical result is decades away?
- What matters more for society: one big bet on fusion, or many smaller solutions that work today? Can the two even be set against each other?
- If you had to explain the difference between "the fuel gave more energy" and "the plant gave more energy" to a child, what analogy would you use?
Log in or Sign up to join the discussion.