Helion has a customer waiting for fusion electricity. In May 2023, the company announced an agreement to sell power to Microsoft, targeting initial plant operation in 2028 and at least 50 megawatts of generation after a one-year ramp-up. Constellation would handle power marketing and transmission. Those are development targets, not delivered power; megawatts describe generating capacity, rather than a fixed quantity of energy.
On Moonshots with Peter Diamandis, energy investor Ramez Naam used that deal to explore what separates a signed contract and encouraging physics from electricity a customer can use. When a panelist suggested that a favorable regulatory environment could explain Helion's aggressive timetable, Naam pushed back: “I'd say the barriers are still physics and engineering.”
Three ways to make fusion work
Fission, the nuclear process used in today's commercial reactors, splits heavy atomic nuclei. Fusion joins light ones, releasing energy. The challenge is to make the fuel hot and dense enough, and keep it together long enough, to produce more useful energy than the machinery consumes.
Naam sorted the field into three rough families, apologizing to founders whose designs the simplification left out.
Magnetic rings. Tokamaks use powerful magnets to confine plasma—a hot gas of electrically charged particles—in a doughnut-shaped chamber. Naam sees this approach, pursued by Commonwealth Fusion Systems, as having the strongest foundation in earlier experiments. CFS uses high-temperature superconducting magnets to pursue a more compact machine, with the aim of reducing the size and cost of the plant.
The company's ARC plan describes an approximately 400-megawatt commercial station in Chesterfield County, Virginia, targeting the early 2030s. It would fuse deuterium and tritium, two forms of hydrogen. A surrounding liquid blanket would collect heat and produce more tritium fuel, while transferring energy to a steam-turbine system that generates electricity. That is a proposed plant, not an operating demonstration of commercial output.
Lasers. The National Ignition Facility, or NIF, at Lawrence Livermore National Laboratory compresses a tiny fuel capsule with lasers. Its experiments support nuclear-stockpile stewardship as well as research relevant to fusion energy. Naam was excited by its scientific results but distinguished the experimental facility from a machine designed to sell electricity.
Colliding pulses. Helion uses magnetic coils to accelerate two plasmas toward each other, then compresses the combined plasma. Its attraction is what happens afterward: the expanding plasma pushes against a magnetic field, inducing electrical current in coils. The company's technology FAQ describes returning that energy to capacitor banks—devices that store electricity—rather than first collecting heat and running a steam turbine.
Skipping that conversion step could avoid energy losses and equipment costs. Among the heavily funded companies he discussed, Naam thought Helion had a possible path to the cheapest electricity if it works, while CFS had the stronger experimental foundation.
Helion's preferred deuterium–helium-3 fuel helps explain both sides of that bargain. Much of its fusion energy emerges in charged particles, which suit direct electrical recovery, but the fuel requires more demanding fusion conditions than deuterium–tritium. And recovering electricity is not the same as producing a surplus across the entire plant.
What NIF's record actually measured
The distinction becomes clear in Livermore's account of NIF's December 5, 2022 experiment. The fuel released 3.15 megajoules of fusion energy after receiving 2.05 megajoules of laser energy. The laboratory reported three further target-gain successes in 2023, including a July shot that produced 3.88 megajoules from 2.05 megajoules.
That was a measured scientific achievement, not merely a theoretical gain. But its boundary was the target: the comparison counted laser energy delivered to the capsule, not all the electricity needed to operate the lasers and facility. Nor was the fusion energy converted into electricity for customers.
A power plant has a bigger energy bill to beat. It must recover useful energy, supply its own lasers or magnets and other equipment, and still have electricity left to export.
Naam's optimism also drew on a chart of the “triple product,” a measure combining particle density, temperature and energy-confinement time. In ordinary terms: how much fuel is packed together, how hot it is and how long it retains its energy. Pressure already incorporates density and temperature; it is not an additional factor to multiply by temperature again.
Samuel Wurzel and Scott Hsu's review of progress against the Lawson criterion compares published results across magnetic, laser-driven and related fusion experiments. It distinguishes energy-gain measures at different boundaries, from the fuel to the electrical and engineering systems. In its historical dataset, tokamaks and laser-driven systems led performance, but many approaches improved rapidly at first and then slowed as they approached breakeven requirements. Progress in fuel conditions does not supply a commercial delivery date.
For pulsed designs such as Helion's, the next challenge is doing the useful thing repeatedly: recovering enough energy from successive pulses, powering the next pulse and supporting dependable electricity delivery. Components also have to survive that workload. Neutrons can damage reactor materials, creating replacement costs and downtime; the maintenance interval depends on the design and operating conditions.
That is where cheap fuel can meet expensive power. Construction costs must be recovered through electricity sales, and a machine undergoing repairs is not generating revenue.
A reaction that stops still leaves material to manage
Fusion has a safety advantage Naam considers commercially important. It does not sustain the neutron-driven chain reaction used in fission. If the conditions needed for fusion are lost, the reaction stops. Keeping it going is the difficult part.
That does not mean a damaged fusion plant has no hazards. The Nuclear Regulatory Commission's fusion FAQ explains that neutrons can make components radioactive and that tritium, a radioactive form of hydrogen, can contaminate equipment. Those materials still require controls during operation, maintenance, recycling and disposal.
The United States is approaching fusion through its byproduct-material framework rather than treating it as conventional fission-reactor licensing. The 2024 ADVANCE Act incorporated fusion-produced radioactive material into that legal category. Oversight involves the NRC and Agreement States—states authorized to regulate specified radioactive materials—with fusion-specific rulemaking intended to accommodate different designs.
Naam sees that regulatory direction as an advantage for the industry. It is not an exemption from managing radioactive material: even Helion's preferred fuel system includes tritium handling as part of its proposed helium-3 production and recycling.
A reason to invest, not to stop building alternatives
Naam's enthusiasm rests on experimental progress and the range of approaches now being attempted. He describes a trade-off: designs with a stronger scientific foundation may offer more conventional power economics, while radical improvements in size or cost often come with greater technical uncertainty.
Compact fusion is one example. Avalanche Energy's Orbitron proposal targets machines producing 1–100 kilowatts of electricity. Its first proposed energy-extraction phase uses deuterium–tritium and a thermal conversion cycle. Naam sees enormous possibilities in compact machines, but less evidence that they will work than for the tokamak approach pursued by CFS.
He is similarly cautious about founders' calendars. “Every startup exaggerates how quickly they can get things done,” he told the panel. Optimism gets companies started; it does not remove the remaining experiments.
Asked whether solar, batteries and fusion could settle the clean-power question, he resisted treating fusion's success as assured. The companies might fail, or succeed technically and still produce electricity that costs too much. That leaves a case for investing in fusion alongside other clean-energy approaches, rather than waiting for one winner.
For Microsoft and the growing electricity needs of AI infrastructure, Helion's target therefore entails much more than a successful fusion pulse in 2028. It calls for a plant that can cover its own energy use, keep its machinery working and then ramp to at least 50 megawatts for sale. The purchase agreement gives that work a customer. The machine still has to supply the power.