David Kirtley on Nuclear Fusion, Plasma Physics, and the Future of Energy

David Kirtley is a nuclear engineer and the CEO of Helion Energy, a company building fusion generators by a route most of the field abandoned in the 1950s. The conversation runs from the bare physics — why fusing light atoms releases energy, why it cannot melt down, why a fusion plant cannot be turned into a bomb — through the plasma engineering of confining matter at 100 million degrees, to Helion’s distinctive bet: a pulsed, linear machine that makes electricity straight from the reaction rather than by boiling water. It ends on the stakes: a Microsoft power-plant deal due to deliver first electricity in 2028, the energy appetite of AI data centres, and where unlimited clean power would leave a civilisation.

Key ideas

1. Fission splits heavy atoms; fusion joins light ones — and that difference is why fusion is inherently safe. Fission breaks apart uranium or plutonium and runs as a self-sustaining chain reaction, hard to stop. Fusion forces light hydrogen isotopes together; in both cases the products weigh slightly less than the inputs, and that missing mass becomes energy by Einstein’s E=mc². But fusion is so hard to start that it stops itself: take away the fuel and it simply ceases. Kirtley insists on the word generator, not reactor — a reactor self-sustains, a fusion machine, like a gas turbine, runs only while you feed it.

2. The hard part is confining a plasma at 100 million degrees. Fusing two positively charged nuclei means overcoming their mutual repulsion, which takes enormous speed — temperature is just velocity, around a million miles an hour. At that heat, matter is a plasma: electrons stripped from nuclei, a soup of charged particles that would destroy any material it touched. The sun confines its plasma by sheer gravity; on Earth the choices are to crush it fast (inertial fusion, lasers) or hold it in magnetic fields (the doughnut-shaped tokamak and stellarator). The physics is century-old; assembling it into a working power plant is what remains brutally hard.

3. Helion’s machine makes its own magnetic bottle and gives back electricity directly. Helion uses pulsed magneto-inertial fusion in a straight (not ring) chamber, exploiting the field-reversed configuration (FRC) — a self-organised plasma that carries its own electrical current and so generates the magnetic field that traps it. As the fuel fuses and expands, it pushes back on that field, inducing a current Helion harvests straight off its capacitors. No steam, no turbine: where a tokamak boils water at ~30% efficiency, the FRC piston-stroke promises 80% or more, plus recovery of the input energy at 95%.

4. The path is manufacturing, not a single mega-experiment — and the target is a plant by 2028. Helion has built seven machines (early ones named after beer and Starbucks cup sizes), iterating fast on cheap, mass-producible parts — buying vacuum pumps off eBay to dodge nine-month supply chains, building 100 small magnets instead of one large one. Kirtley’s lesson: relentless cheap, fast manufacturing accelerates the science, because you learn by building. A 2023 deal commits Helion to a grid-connected plant supplying Microsoft, with first electricity targeted for 2028.

5. Fusion is the natural energy source for an AI-hungry, possibly post-expansion civilisation. The cost of AI computation bottoms out at the cost of electricity, concentrated on-site — a near-perfect match for a dense, sitable fusion generator. Kirtley extends the thought to the Kardashev scale and the Fermi Paradox: rather than a great filter, he favours the Matrioshka brain — advanced civilisations that stop planting flags in space and instead pour fusion power into growing their own cognition, going quiet rather than dying out.

Content

Fission and fusion: opposite ends of the same curve

Kirtley opens on the single picture that organises nuclear energy: the binding-energy curve across the periodic table, from light hydrogen up to heavy uranium. Iron sits at the bottom. Elements lighter than iron release energy when fused together; elements heavier than iron release energy when split apart. Both moves exploit the same accounting trick — the products weigh fractionally less than the inputs, and Einstein’s E=mc² converts that lost mass into a torrent of energy. The mass change is real even in ordinary chemistry, Kirtley notes — burning hydrogen and oxygen into water loses a sliver of mass too — but it is so tiny there that we measure it in electron-volts and never notice.

The fuels differ as sharply as the reactions. Fission needs uranium and plutonium, heavy nuclei forged in supernovae and dug out of the ground, so unstable that adding a single neutron cracks them open. Fusion needs deuterium — a heavy isotope of hydrogen, one proton and one neutron — which is everywhere: in seawater, in Coca-Cola, in the body. Kirtley’s estimate is that the deuterium in Earth’s oceans holds between 100 million and a billion years of fuel at humanity’s current electricity use. Nobody can corner that market or cut that pipeline.

Why fusion cannot run away, and cannot be made into a bomb

The chain reaction is the heart of fission’s danger and fusion’s safety. In fission, each split nucleus throws off neutrons that split the next — get the balance wrong and it accelerates uncontrollably. Fusion has no such chain: it is so hard to ignite that it can only be forced, never sustained. Stop feeding fuel and it stops instantly. Helion’s machines hold only about one second of fuel at any moment; the rest is inert deuterium in a tank that cannot fuse by itself.

Kirtley walks through the safety case Helion built for the Nuclear Regulatory Commission. Their worst-case analysis assumes a meteor vaporises an operating plant — and concludes you would not need to evacuate the neighbourhood, because there is simply no large store of energy waiting to be released, unlike the years of fuel sitting in a fission core or a burning coal pile. Fusion still produces ionising radiation while running — X-rays and neutrons — so the machines are shielded in borated polyethylene and concrete, and regulated not as reactors (NRC Part 50) but as particle accelerators (Part 30), the same rules that cover hospital irradiators. Helion licensed its first fusion system as a particle accelerator in 2020; an early regulator’s standard form asked, baffled, where the patients go.

On weapons, Kirtley is categorical: a fusion plant cannot be turned into a bomb. The hydrogen bomb is misnamed — about 90% of its energy still comes from a fission reaction; the fusion fuel only boosts it, and remove the fissile core and the fusion does not happen at all. An all-fusion bomb has never been built and, by present physics, cannot be. The proliferation experts Helion approached did not warn them off; they urged them to deploy fusion as fast as possible, precisely so the world is not driven to build enriched-uranium plants and the centrifuges that come with them.

Confining a plasma: magnets, mirrors, and the doughnut

To fuse, fuel must be hot, dense, and held together long enough. Kirtley uses three quantities throughout: N (density), T (temperature), and tau (confinement time). The sun supplies confinement with gravity. On Earth, the two mainline public programmes are inertial fusion — crush the fuel with lasers in nanoseconds, as the National Ignition Facility demonstrated — and magnetic fusion, which traps charged particles on magnetic field lines and tries to hold them as long as possible.

A charged particle in a magnetic field becomes magnetised: trapped on a field line, spiralling around it, exactly as charged particles spiral down Earth’s field to make the aurora. A straight magnetic tube (a solenoid) leaks particles out its ends. Two historical fixes followed. One bent the tube into a ring so particles circle endlessly — this became the tokamak and the stellarator, the field’s dominant designs. The other closed the ends with stronger fields — the magnetic mirror — but the hottest, most valuable particles always squirted out, like hot gas escaping a balloon.

The field-reversed configuration: a plasma that bottles itself

Helion descends from a different 1950s lineage: the theta pinch, which did not just hold the plasma but crushed it by rapidly ramping the magnetic field, driving density up until fusion really began. Those pioneers hit a hardware wall in 1958 — switching millions of amps in microseconds before the transistor existed — and the field turned to lasers instead. Decades later researchers found that under the right conditions a theta-pinch plasma stopped squirting out the ends and instead held itself together. The accident had a name: the field-reversed configuration.

The mechanism is the conversation’s centrepiece. Reverse the external magnetic field faster than the plasma can respond — in about a millionth of a second, possible only now thanks to semiconductor switching — and the plasma reconnects internally into a closed, self-organised structure. By Lenz’s law it behaves like the secondary of a transformer: huge electrical current flows in the plasma itself, and that current generates the very magnetic field that traps it. Kirtley’s line: in a tokamak you build the magnets and trap the plasma; in an FRC the plasma makes the magnets and traps itself. Nature does it constantly — a solar flare that pinches off the sun is a plasmoid of exactly this kind.

Keeping it stable: the spinning-top trick

A self-bottling plasma is unstable. Its plasma beta — the ratio of particle pressure to magnetic pressure — sits at one, the highest possible, which makes it want to tilt and turn over, because nothing mechanical can hold it (everything inside is 100 million degrees). The fix is the physics of a spinning top: a top falls over, but spin it fast enough and it stays upright through sheer angular momentum. Helion drives the plasma’s ions to very high velocity to stabilise it, captured in a design parameter Kirtley calls S* over E — a stability number against the elongation of the machine. Fortuitously, FRCs form naturally long and thin, so the geometry helps. Plasmas that theory says should survive a few microseconds have been held for thousands. The catch is circular: stability needs heat, and the whole challenge is heating a cold plasma fast enough before it tilts — which is why, Kirtley jokes, Helion is on some days more an electrical-engineering company than a fusion company.

Electricity straight from the reaction

This is Helion’s commercial wager. A tokamak is a campfire: light the match, grow the fire, stand beside it and capture the heat, then boil water to spin a steam turbine at roughly 30–35% efficiency — the same century-old machinery as a coal plant. The FRC is a piston engine. As fusion fires, the new charged particles raise the plasma’s pressure, which pushes back on the magnetic field, which forces electrical current back out into the capacitors the process started from. The expansion does work directly, like a piston turning a crankshaft. Theoretical efficiency for the fusion energy is 80–85%; recovery of the input electrical energy has been demonstrated above 95%.

This choice reaches back into the fuel. Standard fusion runs on deuterium–tritium, whose energy comes out mostly in neutrons — uncharged, so they fly out and can only boil water. Helion targets deuterium and helium-3 (a helion — the company’s namesake), which yields a charged proton instead of a neutron, so it pushes on the magnetic field and can be harvested as electricity. The prices are real: helium-3 is scarce on Earth (it leaks to space; it may have to come from the Moon or Jupiter), and the reaction wants 200–300 million degrees rather than 100. But if you recover energy at triple the efficiency and reclaim your input, the machine ends up about the same size for the same electricity out.

Build fast, build cheap, build a factory

Kirtley’s engineering philosophy is the human core of the episode. Helion built six prototypes before its seventh, Polaris, scaling FRC formation and compression each time; the 2020 machine Trenta reached 100 million degrees and did bulk deuterium–helium-3 fusion. The discipline is to choose the smallest machine that accomplishes the mission and iterate it fast on commonly available materials — aluminium, copper, G10 fibreglass (the substrate of circuit boards), turbo pumps bought used off eBay and tested in threes. Helion is roughly half technicians, vertically integrated down to its own power-supply production lines (and, Kirtley claims, the only fusion company with a conveyor belt). The counter-intuitive payoff: relentless manufacturing speed makes the science faster, because each cheap fast build teaches you the next.

The destination is not a single demonstration but an industry. The 2023 Microsoft deal commits Helion to a grid-connected plant delivering first electricity in 2028 — a deadline Kirtley calls daily and tough. Beyond it he wants a gigafactory turning out 50-megawatt generators by the truckload, to start replacing the world’s 4,000 gigawatts of installed fossil-fuel capacity.

Fusion, AI, and the shape of the future

The closing arc widens to civilisation. The cost of AI computation, Kirtley argues, asymptotes to the cost of electricity — concentrated, on-site, dense — which is exactly fusion’s profile and exactly a data centre’s need. A 50-megawatt plant fits on roughly an acre where equivalent solar would need 2,000; its output is already DC, which is what GPUs want, opening the possibility of feeding compute directly and skipping grid losses. Unlimited dense power, he suggests, unlocks desalination, vertical farming that returns land to nature, even electric propulsion to space. On the Fermi Paradox he is an optimist: rather than a great filter destroying everyone, he favours the Matrioshka brain — civilisations that stop expanding outward and pour their fusion-fed energy into cognition, growing inward until their star, sheathed in collectors, can no longer be seen. Asked for the most beautiful idea in his field, he answers that he is continuously amazed it works at all — that the balance between the electromagnetic force and the strong force is so fine it is hard to believe it is an accident.

Related

• Don Lincoln on Particle Physics, Antimatter, and the Biggest Mysteries in Physics — shares the E=mc² and strong-force foundations from the particle-physics side
• Alex Wellerstein on the Atomic Bomb Decision, Truman's Accountability, and the Nuclear Taboo — the fission-weapons and proliferation territory Kirtley distinguishes fusion from
• Sean Carroll on General Relativity, Quantum Mechanics, Black Holes and Aliens — complementary treatment of the Fermi Paradox and alien civilisations
• Sara Walker on the Physics of Life, Time, Complexity and Aliens — overlapping questions on physics, aliens, and where intelligence is headed
• Nuclear Taboo — the norm against nuclear weapons that frames the proliferation discussion
• Lex Fridman — host