13 SLIDES · THE SCIENCE · THE PROJECTS · THE TIMELINE
Pressed close enough together, light nuclei fuse — the strong force snaps them into a heavier nucleus, and the mass deficit becomes kinetic energy. The easiest reaction we know:
¹H + 1n · abundant in seawater, ~33 g/m³
¹H + 2n · radioactive, t½ ≈ 12.3 yr · must be bred
17.6 MeV per fusion · ~4×10⁸ J / g of fuel
For comparison: 1 g of D-T fuel ≈ 8 tonnes of oil equivalent.
Two positive nuclei repel via the Coulomb force. To get them close enough for the strong force to take over, you have to fling them at each other — thermally.
More than 6× hotter than the core of the Sun. The Sun cheats by being absurdly massive (gravity does the confining); we don’t have that option.
density × temperature × confinement time
(keV · s · m⁻³)
You can trade among the three. Tokamaks run hot & long; inertial schemes run hot & dense for nanoseconds.
No solid material survives contact with a 100-million-degree plasma. The plasma must be held away from every wall — by magnetic fields, by inertia, or both.
They are different reactions with very different consequences. The marketing conflation costs fusion politically.
| Fission | Fusion | |
|---|---|---|
| Reaction | ²³⁵U + n → fragments | D + T → ⁴He + n |
| Fuel | uranium / plutonium | hydrogen isotopes |
| Energy / kg fuel | ~8×10¹³ J | ~3.4×10¹⁴ J |
| Long-lived waste | 10⁵-yr actinides | none from the reaction itself |
| Runaway risk | chain reaction; needs control | plasma quenches if disturbed |
| Weaponizable byproduct | plutonium | tritium (limited; not a bomb fuel) |
Fusion does produce activated structural materials from neutron flux — a real engineering problem, but on decade half-lives, not millennia.
Charged particles spiral along magnetic field lines. Bend the lines into a closed torus and the plasma stays trapped — in principle, forever.
The word tokamak is Russian shorthand for “toroidal chamber with magnetic coils.”
Don’t hold the plasma — crush it. A peppercorn-sized capsule of D-T is hit symmetrically by a converging shock; for a few hundred picoseconds the fuel is denser than lead and hotter than the Sun’s core. By the time it blows apart, the reaction has run.
192 lasers delivered 2.05 MJ to a capsule. The fusion reaction released 3.15 MJ. First net energy gain in a controlled fusion experiment in human history.
~2 mm diameter, frozen D-T layer inside a diamond shell, suspended in a gold cylinder (hohlraum) that converts laser light to a uniform X-ray bath.
A working power plant would need to do this ~10 times per second, every second, for years. NIF currently fires a few shots a day.
For 60 years fusion meant national labs and ITER. Since ~2018, $7+ billion of private capital has flowed into ~40 startups, each betting on a different shortcut.
MIT spinout. Compact tokamak using high-temperature superconducting (HTS) tape. Targets Q > 2 in 2027. Sited in Devens, MA.
Pulsed field-reversed configuration. Burns D-³He. Direct electric conversion (no steam). Microsoft signed a 50 MW PPA for 2028 — aggressive.
Aneutronic p-¹¹B fuel. Hardest fuel cycle (needs ~10⁹ K) but cleanest output. Backed by Google.
Magnetized target fusion: pistons crash a liquid lithium liner around a plasma. Building demo in Oxford, UK.
UK firm. Compact spherical tokamaks + HTS magnets. Reached 100 M K in ST40 (2022).
That faster iteration + new magnet tech beats one giant 30-year intergovernmental megaproject. Verdict pending.
Tokamak performance scales steeply with magnetic field strength — roughly as B⁴. Doubling B shrinks the machine by ~16× for the same fusion power.
~5–6 T · cooled to 4 K with liquid helium · brittle, expensive
20+ T · works at 20 K · thin, robust, manufacturable
ITER-class performance in a building you can fit on a campus. This is why SPARC, Tokamak Energy, and others suddenly look credible.
CFS’s 2021 demonstration of a 20 T HTS toroidal-field coil at full scale was the moment serious people stopped dismissing private fusion. It happened. The magnet works.
Headlines say “NIF achieved net energy.” True — but only in a specific, narrow sense. Three different Qs matter:
| What | Definition | NIF Dec ’22 |
|---|---|---|
| Qscientific (target gain) | fusion energy out / laser energy on target | 1.54 |
| Qengineering | fusion energy out / total wall-plug electricity in | ~0.01 |
| Qcommercial | net electricity out / wall-plug in, after capture losses | ~0 |
NIF’s lasers drew about 300 MJ from the wall to deposit 2 MJ on target. So to get from a scientific milestone to a power plant we still need ~50–100× more. It is not a small gap. It is a real gap. Both things are true.
Beyond ignition, four engineering problems must be solved simultaneously:
World tritium stockpile is ~25 kg. A 1 GW plant burns ~56 kg/year. Plants must breed their own tritium from a lithium blanket struck by fusion neutrons. Required ratio > 1.0; nobody has demonstrated this in a real machine yet.
14 MeV neutrons embrittle steel and create activation. First-wall tiles facing the plasma erode. New alloys (RAFM steels, tungsten) must survive ~150 dpa of damage over a plant lifetime.
NIF: a few shots a day. ITER: 400 s pulses, then cool down. A plant needs months of continuous operation. Every component must work for that.
Even if it works, capital cost per kW must compete with solar+storage and fission. The unit economics are unproven; the regulatory path doesn’t exist yet.
The old joke — “fusion is 30 years away, and always will be” — has been roughly right since the 1950s. Here’s where reasonable people now disagree:
HTS magnets + private capital + iteration. Helion, CFS, Tokamak Energy all guide to a demo plant by ~2030 and pilot grid power by mid-decade.
ITER scientific results late 2030s → DEMO designs → first commercial plant ~2045. The IAEA roadmap. The boring answer.
Tritium breeding may not close. Materials may not survive 30 years. By then, solar + batteries + advanced fission may have eaten the niche.
If fusion works at scale, it is the closest thing physics offers to a generic energy abundance:
Cheap, clean, abundant electricity is the upstream input to nearly every problem we want to solve — desalination, direct air capture, fertiliser, heat for industry, AI compute. It is not a niche fix. It is a civilisation-scale lever, if we can build it.
It is also not guaranteed, not soon, and not a reason to slow anything else down. Build solar. Build fission. Build fusion. All of them.
YouTube · “fusion energy explained”
Kurzgesagt, Real Engineering, Sabine Hossenfelder, and Veritasium all have solid intro tracks.
YouTube · “NIF ignition breakthrough”
LLNL’s own announcement, plus deep-dives on what gain Q=1.5 actually buys.
⟡ END · 13 / 13 ⟡