Nuclear fusion, explained: three ways to bottle a star

Nuclear fusion is the reaction that combines light atomic nuclei into heavier ones, releasing about four times more energy per kilogram of fuel than nuclear fission — the same process that powers the sun and every other star. It matters because the fuel is effectively inexhaustible, the reaction cannot run away, and it leaves no long-lived high-level waste. The physics has been understood for a century; the unsolved problem is engineering a machine that holds a star's conditions long enough, and cheaply enough, to sell the electricity.

Key facts

• Fusion combines light atoms into heavier ones, releasing roughly 4x more energy per kilogram than fission and nearly 4 million times more than burning coal
• The fuel — usually the hydrogen isotopes deuterium and tritium — must be heated above 100 million degrees Celsius and confined long enough to sustain the reaction
• Three machine families compete: tokamaks, stellarators and laser-driven inertial confinement
• Laser fusion holds the gain record: 8.6 MJ from a 2.08 MJ shot — target gain above 4 — at the National Ignition Facility in 2025
• Private fusion companies have raised $14.24 billion cumulatively, including a record $4.48 billion in the year to mid-2026
• No fusion machine has yet delivered sustained net electricity to a grid

How it works

Two light nuclei carry positive charges, so they repel each other. Fusion happens only when they collide hard enough to overcome that repulsion — which means heating the fuel into a plasma at over 100 million degrees Celsius, roughly ten times hotter than the core of the sun. Stars cheat: their gravity squeezes fuel at pressures no machine can match, so they burn at lower temperatures. On Earth, the whole game is confinement — holding a plasma hotter than anything in the solar system away from every material surface, long enough for fusion output to exceed the power spent heating it.

Three families of machines solve that puzzle in very different ways.

Tokamaks confine the plasma in a donut-shaped magnetic cage, driving a strong current through the plasma itself to complete the confining field. They are the most studied approach and hold most of the field's performance records for sustained fusion power — ITER, the largest science experiment on Earth, is a tokamak, and so is Commonwealth Fusion Systems' SPARC, the leading private machine.

Stellarators twist the chamber and the external magnetic coils themselves so the plasma stays confined without a driven current. That trades enormous engineering complexity at construction time for calmer, more stable plasma at run time — no current means no sudden disruptions. Germany's Wendelstein 7-X, the flagship stellarator, set a world-record triple product over a 43-second plasma in 2025, the key metric on the road to a self-sustaining burn. The approach also holds a regulatory first: in August 2026, Tennessee issued Type One Energy the first state fusion-specific license in the U.S. for its stellarator-based Project Infinity at TVA's Bull Run site.

Inertial confinement abandons sustained confinement entirely: laser pulses compress a peppercorn-sized fuel pellet to fusion conditions for a fraction of a second. A power plant would repeat the shot several times per second. This is the approach that achieved ignition first, at the National Ignition Facility in December 2022.

The numbers

Common misconceptions

"Fusion produced net energy, so power plants are close." NIF's ignition shots measure fusion output against laser energy delivered to the target — not against the roughly 300 MJ the facility pulls from the grid to fire those lasers. A power plant needs engineering gain, wall-plug to wire, plus materials, tritium breeding and a shot rate no laser system has demonstrated.

"Fusion is just another kind of nuclear power." Today's reactors split heavy atoms in a chain reaction that must be actively controlled. Fusion is the opposite trade: no chain reaction, no meltdown mechanism, no long-lived high-level waste — but a reaction so hard to sustain that any fault simply extinguishes it. Fusion's safety case is easier; its physics case is far harder.

"The tokamak has already won." Tokamaks lead on performance records and capital raised, but laser fusion is the only approach to demonstrate gain above 1, and stellarators may prove the better power-plant topology precisely because they avoid the tokamak's disruption problem. The $14.24 billion private field is still spread across all three families — and several more exotic ones.

Who is building fusion

The private fusion industry counted 56 companies employing more than 16,000 people in the Fusion Industry Association's 2026 survey, up from 23 companies in 2021, and 71% of them expect the first fusion plant to deliver grid electricity in the 2030s. Capital is concentrated at the top: Commonwealth Fusion Systems alone has raised $4 billion, Helion reached a $15.5 billion valuation with its $465 million Series G, and Europe's largest round to date is Proxima Fusion's $518 million. The leading developers by approach:

The list runs longer — the FIA counts stellarator, mirror, levitated-dipole and electrostatic ventures from New Zealand to Sweden — but the table above holds most of the $14.24 billion invested to date.

Current state (September 2026)

The field's center of gravity is shifting from laboratories to companies. Commonwealth Fusion Systems raised another $1 billion in July 2026, bringing its total to $4 billion, with CEO Bob Mumgaard putting SPARC assembly at roughly 80% complete and holding to grid power from the follow-on ARC plant in the early 2030s. Industry-wide, the Fusion Industry Association counted a record $4.48 billion raised in a single year. Regulators are moving too: on August 31, 2026, Tennessee issued Type One Energy the first fusion-specific license granted by any U.S. state — a byproduct-material license, the pathway the NRC settled on for fusion in 2023 — for its stellarator-based Project Infinity at TVA's retired Bull Run coal site in Clinton, Tennessee. The company initiated licensing in January 2026 and targets its Infinity One prototype in 2029 and a roughly 400 MWe Infinity Two plant in the early-to-mid 2030s. On the public side, ITER's rebaselined schedule targets research operations in 2034, while NIF keeps raising the gain ceiling and Wendelstein 7-X extends the stellarator's long-pulse records. The scoreboard that matters — sustained net electricity on a grid — still reads zero for every approach.

Related reading

For the money and hardware behind the leading private tokamak, see NNN's coverage of Commonwealth Fusion Systems' $1 billion raise. Fusion is also one thread in the U.S. government's AI-for-science push tracked in DOE's Genesis Mission project portfolio, and the test-bed model for advanced machines has a fission cousin in the DOME microreactor test bed.