How close do you want to stand?

For anyone who has ever looked up and wondered what starlight is made of.

A tiny star,
built by hand.

Fusion is what makes the Sun shine. Tiny pieces of matter bump together, join up, and give away energy as light and heat. People are trying to make a little piece of that star-story happen safely inside machines on Earth.

Fusion is the work stars do every second: light atoms pushed close enough to become heavier atoms, releasing energy as they settle into something new. On Earth, the dream is simple to say and brutal to build: make a star-hot reaction, hold it for a moment, and harvest the glow.

The race is not one race. It is several incompatible bets on the same rule: heat the fuel, crowd it, and keep it together long enough. Tokamaks ask for patience. Laser fusion asks for an instant. Pulsed machines try to meet in the middle. Blue Laser Fusion sits in the laser corner, carrying Shuji Nakamura's blue-light lineage into a much harder problem.

Every concept is negotiating the Lawson product: density x confinement time x temperature. Magnetic devices trade low density for long confinement. Inertial systems trade nearly no time for enormous density. Magneto-inertial machines try to buy enough of both without needing a steady star in a bottle.

3ingredients decide whether the reaction can run: heat, crowding, time.
2022NIF crossed scientific ignition for a laser shot, not wall-plug power.
0commercial fusion plants have delivered electricity to the grid.
2030sthe decade most public roadmaps still point toward, with risk attached.

choose the path

A long story should still have doors.

Read straight through, or jump by question. The explainer is built as a field guide: first the recipe, then the machine families, then the plant, the games, the claims, and the sources.

Illustrated title card for A Tiny Star, Built by Hand.
Magnets, lasers, pulsed machines, Blue Laser Fusion, and the caveats behind the dream.

the recipe every machine obeys

Three ingredients, and none of them forgive you.

Imagine trying to make two tiny balls stick together while they are racing around. You need them hot, packed close, and kept together for long enough. That is the whole puzzle.

Fusion is not magic. It is a collision problem. Make the fuel hot enough to move fast, dense enough to meet often, and held together long enough for the meetings to matter.

The hard part is that the knobs fight each other. Squeeze harder and the fuel tries to explode. Hold longer and the machine gets bigger, colder, leakier, more expensive. Every design below is a personality built around one sacrifice.

For deuterium-tritium fusion, useful ignition is often framed around a triple product near n tau T ~ 3e21 keV s m^-3, with temperature optimums in the tens of keV. The model here is deliberately qualitative; it teaches tradeoffs, not plant design.

The glow is a teaching instrument. It compresses the physics into intuition: more heat, more crowding, and more time move the fuel closer to ignition.

Low crowding with long time feels like a tokamak. Extreme crowding with almost no time feels like laser fusion.

The labels change because the same recipe can describe machines that live for seconds or only fractions of a nanosecond.

The fuel is close, but not self-sustaining. 62%
temperature x density x confinement time = the gate every concept tries to cross

the field, seen from above

When you plot density against time, the machines reveal their character.

This is an order-of-magnitude map, not a claim of exact device placement. Its job is to make the tradeoff visible: long and thin on the left, brief and crushed on the right, gravity far above every machine humans can build.

fuel density increases
holding time increases

four big families

Every fusion machine is a bet on which pain you can survive.

The approaches below are not skins on the same machine. They are different philosophies of control, repetition, materials, economics, and patience.

magnetic confinement

Tokamaks and stellarators

Use magnets like invisible hands. Keep the hot gas away from the walls and wait for enough collisions.

A magnetic bottle keeps a thin, star-hot gas from touching the walls. The fuel is sparse, so the machine asks for time.

This is the most mature path and the most heavily funded one. ITER and SPARC live here. The promise is controlled burning plasma; the burden is complexity, materials, and cost.

Typical density is around 1e20 m^-3, with confinement measured in seconds. The gain target is plasma gain, not whole-plant electrical gain.

Strengthlong confinement
Weaknessplasma stability and plant complexity
inertial confinement

Laser implosion

Use light as a hammer. Hit a tiny fuel bead so fast that it fuses before it can escape.

A tiny fuel pellet is hit so hard and so evenly that it fuses before it can fly apart.

This is the corner where NIF proved scientific ignition in December 2022. It is also the corner where repetition rate, target cost, laser efficiency, and heat handling become unforgiving.

The fuel can reach roughly solid-density multiples for about 1e-10 s. NIF's famous gain was target gain, not wall-plug gain.

Strengthextreme density
Weaknessefficiency, repetition, target fabrication
magneto-inertial

Pulsed middle paths

Make a small hot ring or cloud, then squeeze it quickly. It is part magnet bottle, part fast squeeze.

Make a self-contained magnetic plasma shape, then compress it quickly.

Helion, TAE-style field-reversed configurations, General Fusion-style compression, and some pinch concepts sit between the extremes. They want useful pulses without building ITER-scale magnets or NIF-scale lasers.

These concepts often target higher density than tokamaks and longer time than laser ICF, with physics dominated by pulsed stability and energy recovery.

Strengthcompact pulse logic
Weaknessproof of repeatable net electricity
for scale

The Sun

The Sun wins because it is huge. Its own weight squeezes the fuel for billions of years.

The Sun does not need our machinery because gravity is its machinery.

It runs cooler than lab fusion plasmas because it has a luxury no lab owns: a star's mass and billions of years of confinement.

Solar fusion proceeds through quantum tunneling in a gravitationally confined plasma. It is useful on the map as a limit case, not as an engineering template.

Strengthgravity
Weaknessrequires being a star

field atlas

The names change, but the bargain stays the same.

Fusion is a continent, not a lane. Some teams build giant steady magnets. Some fire lasers. Some squeeze metal, plasma, or magnetic bubbles in pulses. Some chase cleaner fuels that are much harder to burn. The atlas below is the shortest honest tour of the territory.

tokamak

The workhorse donut

A torus with plasma current and external magnets. It has the deepest operating history and the biggest public projects.

Betlong confinement
Riskdisruptions, neutrons, plant cost
spherical tokamak

The compact donut

A tighter tokamak shape that can reach high pressure for its size, but leaves less room for shielding and magnets.

Betsmaller high-field plants
Riskcenter-column heat and neutron load
stellarator

The twisted bottle

Complex coils do much of the confining work, reducing the need for a large plasma current and aiming at steady operation.

Betsteady-state stability
Riskcoil design and fabrication
mirror / dipole

The open-field bottle

Magnetic mirrors and dipoles use simpler open or planet-like fields. They trade elegance for leakage and end-loss problems.

Betsimpler magnetic geometry
Riskconfinement losses
laser ICF

The tiny crushed star

Lasers compress a pellet so quickly that fusion starts before the fuel can fly apart. NIF proved scientific ignition here.

Betextreme density
Risklaser efficiency, targets, repetition
beam / impact

The projectile idea

Heavy ions, pulsed drivers, or projectiles try to deliver compression more cheaply than giant laser halls.

Betcheaper implosion drivers
Riskprecision at power-plant rhythm
magnetized target

The fast squeeze

A magnetized plasma is compressed by liquid metal, liners, or other pulsed machinery before it has time to cool.

Betmiddle ground between magnets and lasers
Riskrepeatable compression and chamber wear
FRC

The smoke ring plasma

Field-reversed configurations are compact magnetic plasma structures that can be formed, merged, accelerated, and compressed.

Betcompact pulsed systems
Riskstability, losses, energy recovery
Z-pinch

The self-pinching current

Drive current through plasma and its own magnetic field squeezes it. The simplicity is beautiful; the instabilities are not.

Betfew moving parts
Riskviolent plasma modes
electrostatic / DPF

The tabletop temptation

Fusors, dense plasma focus devices, and related ideas can make fusion reactions, but useful power gain is another mountain.

Betsmall, direct, experimental
Risklosses overwhelm output
fuels

The fuel changes the mountain

D-T is easiest to burn but brings tritium and neutron damage. D-D, D-He3, and p-B11 are cleaner on paper and far harder in practice.

Betless activation or direct conversion
Riskmuch higher temperature and lower reactivity
plant layer

The reactor is not the plant

Blankets, tritium breeding, neutron-resistant walls, heat extraction, direct conversion, maintenance, and regulation decide whether plasma becomes power.

Betengineering turns pulses into electricity
Riskmaterials and economics

build-a-plant challenge

Choose a reactor and watch the bottleneck move.

There is no perfect button. Pick a machine, fuel, mission, and main constraint; the scorecard shows why fusion is so hard to declare won. The numbers are qualitative teaching scores, not forecasts.

Machine

Fuel

Mission

Main pressure

the machine after the plasma

The tiny star is only the first room in the power plant.

A fusion plant is not just a beautiful plasma. It is a chain of surfaces, shields, pumps, blankets, coolants, magnets, fuel systems, robots, turbines, direct converters, permits, and maintenance schedules. Touch the cutaway to see where the hard problems move after the plasma starts behaving.

1. BurnFuel fuses and releases energy as charged particles and, for D-T, fast neutrons.
2. CatchWalls, blankets, or direct converters try to collect that energy without being destroyed.
3. BreedD-T plants must make or recover enough tritium to keep feeding themselves.
4. ConvertMost concepts still turn heat into electricity; some pulsed concepts chase direct conversion.
5. MaintainThe plant only matters if it can be repaired, inspected, licensed, and run often enough.

plasma keeper

Now play the bargain instead of reading about it.

Each round gives you a different fusion concept. Move the sliders until your machine matches the target personality. This is intentionally qualitative: the point is to feel why laser fusion, tokamaks, pinches, and advanced fuels ask for different kinds of perfection.

Round 1 of 5

Nakamura's particular wager

What if the hammer could gather itself before it struck?

Blue Laser Fusion is a laser-fusion idea. It tries to make useful blue light, build that light into a stronger pulse, and aim it at a tiny target. The big question is whether the whole chain can work again and again like a power plant.

Blue Laser Fusion belongs to the laser family. The idea is to use diode-laser technology descended from the blue-light revolution, store the beam in an optical cavity, and release a much stronger strike onto a small fuel target.

The reported promise is not that Blue Laser Fusion has already matched NIF. It is that a different laser architecture might change the economics of inertial fusion: smaller, more efficient light sources, amplified in a cavity, repeated often enough to matter.

The central uncertainty is system gain. A diode-pumped, cavity-enhanced source could help laser efficiency, but a power plant still needs target gain, repetition rate, debris handling, precision injection, tritium or alternative-fuel strategy, and conversion to useful electricity.

where the story stands in July 2026

The physics milestone happened. The power plant has not.

This is the emotional center of the subject: fusion is no longer only a dream, but it is not yet a utility. The honest wonder lives between those two sentences.

DEC 2022

NIF achieves scientific ignition.

The laser energy delivered to the target was lower than the fusion energy released. That was historic. It did not mean the building produced net electricity.

2024-2026

Private fusion moves from pitch decks to hardware.

CFS, Helion, TAE, Zap, Tokamak Energy and others are now building machines, factories, magnets, targets and plant sites. The private race is real; so are the unknowns.

2028

Helion targets initial operation for Orion.

Helion says Orion is designed to meet its Microsoft power-purchase agreement. This is a target, not delivered grid power.

2030s

CFS, ITER and Blue Laser Fusion point to different futures.

SPARC aims to show net-energy plasma before ARC. ITER is an experiment, not a power plant. Blue Laser Fusion's claimed plant timeline remains a claim until independently demonstrated.

readiness ledger

What would prove fusion is no longer just promising?

The cleanest way to avoid hype is to name the gates. A concept does not need to pass all of them in one experiment, but a power plant eventually has to pass all of them in the same story.

physics gate

Burn repeatedly at useful gain.

The plasma or target has to make more useful fusion output than the relevant input boundary, not once but as a controlled operating mode.

Evidence
independent shots, pulses, or steady operation with clear gain accounting
Not enough
a single record shot with unclear facility energy
driver gate

Repeat at plant rhythm.

Lasers, magnets, pulsed power, pellets, injectors, switches, and controls must work at the cadence a plant needs.

Evidence
high-duty operation with measured degradation and downtime
Not enough
a machine that needs heroic reset time between runs
fuel gate

Close the fuel cycle.

D-T plants need tritium breeding, recovery, inventory control, and safety systems. Advanced fuels need credible supply and credible burn conditions.

Evidence
fuel balance, breeding ratio, recovery losses, and operating inventory
Not enough
assuming tritium or helium-3 appears because the physics is attractive
materials gate

Survive the chamber.

The first wall, blankets, magnets, optics, and structural components have to survive heat, radiation, erosion, activation, and replacement.

Evidence
reactor-relevant exposure, inspection, replacement, and lifetime data
Not enough
plasma performance with the chamber problem offstage
power gate

Deliver net electricity.

The heat or charged-particle energy has to become usable electricity after pumps, cryogenics, magnets, targets, controls, and conversion losses.

Evidence
measured electrical output against whole-plant electrical input
Not enough
thermal fusion energy reported as if it were grid electricity
economics gate

Be maintainable and financeable.

A plant must be licensed, repaired, insured, staffed, fueled, connected, and priced against real alternatives.

Evidence
uptime model, maintenance plan, component supply, permitting path, and credible cost
Not enough
a beautiful reactor core with no plant operations story

the warm hug also tells the truth

Fusion headlines are brightest exactly where caveats matter most.

"Fusion is here."
No commercial fusion plant is on the grid.

Every commercial date in this piece is a target. A target can be serious and still slip.

"The laser made more energy than it used."
NIF's gain was measured at the target.

The facility consumed far more energy than the capsule released. That does not erase the achievement; it defines what remains.

"ITER will power homes."
ITER is built to learn, not to sell electricity.

Its job is plasma physics and technology validation at scale. Power plants are supposed to follow.

"One concept obviously wins."
The field is hedging because nobody knows yet.

Magnets, lasers, pulsed systems and pinches are not redundant. They are different answers to different bottlenecks.

claim decoder

When a fusion headline glows, ask what kind of glow it is.

The same sentence can mean very different things depending on whether it refers to target gain, plasma gain, wall-plug gain, a company milestone, or a real plant. Use this decoder as a field guide for press releases, investor decks, and breathless headlines.

human glossary

The words are small doors. Open the right one.

Fusion conversations often fail because one person is talking about the plasma while another is talking about the building. These are the terms that keep the room honest.

Ignition

The fuel heats itself enough that fusion reactions help drive more fusion. It is a physics condition, not the same as selling electricity.

ask: ignition of what boundary?

Q

Gain ratio. The important question is which inputs and outputs are counted: target, plasma, facility, or full plant.

Q_plasma != wall-plug gain

Burning plasma

A plasma where fusion products, especially alpha particles in D-T fuel, provide a meaningful share of the heating.

self-heating starts to matter

Breakeven

A slippery word. Scientific breakeven can be real while engineering breakeven and commercial breakeven remain far away.

define the accounting fence

Tritium

A radioactive hydrogen isotope used in the easiest practical fusion fuel. It is scarce and must be bred, recovered, and carefully tracked.

D + T -> helium + neutron

Blanket

The reactor layer that catches neutron energy, shields components, and may breed tritium from lithium in future D-T power plants.

wall, shield, fuel factory

First wall

The plasma-facing surface. It has to tolerate heat, particles, erosion, neutron damage, and replacement realities.

the hardest front door

Aneutronic

Fusion that sends more energy into charged particles and less into neutrons. The catch is usually much harder burn conditions.

cleaner claim, harder mountain

Direct conversion

Turning charged-particle motion into electricity without first making steam. Attractive for some concepts; not a shortcut around proof.

works only after losses are counted

sources and uncertainty

Beautiful, but not credulous.

The page treats company roadmaps as claims, not outcomes. The density-time map is schematic. The facts should be rechecked before publication because fusion status changes quickly.

DOE
NIF ignition, December 2022 2.05 MJ delivered to target, 3.15 MJ fusion output, and why this was a science milestone.
 ITER
ITER goals and caveats Q=10 plasma goal, 500 MW thermal fusion power target, and no conversion to electricity.
 ITER
Blanket and first-wall systems ITER's official context for neutron shielding, heat removal, plasma-facing materials, and tritium breeding tests.
 IAEA
Fusion FAQ Safety, radioactive waste, tritium, commercial timelines, and why fusion is not a chain reaction.
 DOE FES
Fusion Energy Sciences program The official research frame: burning plasma, materials, blankets, theory, simulation, and closing the fuel cycle.
 Helion
Polaris and direct electricity claims Helion's own description of Polaris, fuels, diagnostics, and its clarification of "net electricity."
 Helion
FRC and D-He3 technology claims Helion's description of pulsed field-reversed configurations and direct electricity recovery.
 Helion
Orion and Microsoft PPA Helion's claimed 50 MW power-purchase path and initial operations target.
 CFS
SPARC and ARC roadmap Commonwealth Fusion Systems' public description of SPARC, ARC, and HTS magnets.
 IPP
Wendelstein 7-X stellarator Official context for the largest stellarator and the steady-operation goal.
 TAE
Beam-driven FRC and p-B11 claims TAE's public description of its field-reversed configuration program and advanced-fuel goal.
 General
Magnetized target fusion General Fusion's public description of plasma plus mechanical compression and liquid metal.
 Zap
Z-pinch development Zap Energy's public program context for sheared-flow-stabilized Z-pinch fusion.
 TAE
TAE and Trump Media deal Press-reported merger context for one private fusion financing path.
 Nobel
Nakamura's blue LED lineage The 2014 Nobel Prize context for efficient blue LEDs.