Fusion power has been thirty years away for roughly sixty years. This is the single most cited fact about the field, deployed by skeptics with the confidence of someone who has discovered a devastating argument that the entire fusion research community somehow failed to consider. The joke is real. The timeline failure is real. And the underlying implication—that fusion is a perpetual mirage, always receding as you approach it—is no longer accurate, though you’d be forgiven for not believing that given the track record.
Here’s what changed: in December 2022, the National Ignition Facility at Lawrence Livermore National Laboratory achieved ignition—a fusion reaction that produced more energy than was delivered to the fuel. In 2025, Commonwealth Fusion Systems completed its first high-temperature superconducting magnet and began assembling SPARC, a compact tokamak that the company says will achieve net energy in 2027. Private fusion investment has exceeded $10 billion since 2021. Helion Energy is building a plant to deliver 50 megawatts to Microsoft data centers by 2028. Google, Nvidia, Sam Altman, Bill Gates, and—in a plot twist that nobody anticipated—the Trump family have all put money into fusion companies. The Fusion Industry Association now has 45 member companies.
Something is different this time. Whether that something is sufficient to close the gap between “it works in a lab” and “it powers your house” is the question that matters, and answering it honestly requires understanding why the gap has been so persistent in the first place.
What fusion is and why it’s hard (the 90-second version)
Fusion is the process that powers the sun. You take light atomic nuclei—typically isotopes of hydrogen—and force them together at temperatures exceeding 100 million degrees Celsius until they fuse into heavier nuclei, releasing enormous amounts of energy in the process. The fuel is abundant (deuterium comes from seawater, tritium can be bred from lithium), the energy density is extraordinary (a few grams of fuel produces as much energy as tons of coal), and the waste products are helium and neutrons rather than CO2 or long-lived radioactive waste. On paper, it’s the energy source that solves everything.
The problem is that “forcing nuclei together” requires overcoming the electromagnetic repulsion between positively charged protons, which means you need to create and sustain a plasma—a gas so hot that electrons are stripped from atoms—at temperatures roughly ten times hotter than the center of the sun, and you need to confine that plasma long enough and at sufficient density for enough fusion reactions to occur to produce net energy. The sun accomplishes this through gravity. It has 1.3 million times the volume of Earth pressing inward on its core. We don’t have that luxury, so we have to use either magnetic fields or inertial compression to do the job, and both approaches are engineering nightmares of the first order.
The measure of success is Q—the ratio of fusion energy produced to the energy required to heat and confine the plasma. Q > 1 means you’re getting more out than you’re putting in. Q > 10 means you’re getting enough out to run a practical power plant after accounting for conversion losses and plant operations. NIF achieved Q ≈ 1.5 in its best shot. ITER, the international tokamak under construction in France, is designed for Q ≥ 10. SPARC is targeting Q > 2 as a demonstration, with the expectation that its commercial successor, ARC, will operate at Q > 10.
Why it’s taken so long
The conventional answer is “it’s really hard physics.” That’s true but incomplete. The more complete answer involves three factors that have nothing to do with plasma physics.
Funding volatility. Fusion research has been subject to boom-and-bust funding cycles since the 1970s. In 1976, the U.S. Energy Research and Development Administration published a report projecting timelines for fusion based on different funding levels. The “fusion never” line corresponded to flat or declining budgets. The actual funding trajectory for the next four decades tracked almost exactly along the “fusion never” line. The thirty-years-away joke is less a reflection of the physics being intractable and more a reflection of the fact that governments kept funding the field at the level that their own projections said would produce no result. Fusion researchers have been trying to build a reactor on a budget calibrated to produce a perpetual research program instead.
ITER’s governance structure. ITER is a collaboration among 35 nations, which means it’s simultaneously one of the most ambitious scientific projects ever undertaken and one of the most bureaucratically encumbered. Components are manufactured across multiple countries, shipped to southern France, and assembled according to specifications that were locked in over a decade ago. The project is billions over budget and years behind schedule. First plasma is now expected sometime in the next decade—decades after the project was conceived. ITER may ultimately demonstrate Q ≥ 10, and if it does, it will validate the tokamak concept at reactor scale. But it will not have done so quickly, cheaply, or in a way that suggests the model is replicable for commercial deployment.
The tritium problem. Most fusion reactor designs use deuterium-tritium fuel because the D-T reaction has the lowest ignition temperature and highest cross-section—it’s the easiest fusion reaction to achieve. But tritium doesn’t exist naturally in useful quantities. It’s radioactive with a half-life of 12.3 years and is currently produced almost exclusively as a byproduct of heavy-water fission reactors, primarily in Canada. The global supply is roughly 25 kilograms. A single fusion power plant would consume 50 to 100 kilograms per year. The plan is for fusion reactors to breed their own tritium by surrounding the plasma chamber with a lithium blanket that captures the neutrons produced by fusion and transmutes lithium into tritium. This breeding cycle has never been demonstrated at scale, and achieving a tritium breeding ratio greater than 1.0—producing more tritium than you consume—is an unsolved engineering challenge that every D-T fusion design depends on.
What’s actually close
Commonwealth Fusion Systems is the company most likely to demonstrate net energy fusion in a tokamak in the near term. SPARC is roughly 60 percent complete as of early 2026. The first of 18 toroidal field magnets has been installed. The cryostat base—the structural foundation that supports the 1,000-tonne tokamak—is in place. The company expects all 18 magnets installed by mid-2026, first plasma in 2027, and net energy demonstration shortly after. SPARC is designed to achieve Q > 2 in pulses lasting roughly 10 seconds, possibly extending to 30 seconds—though at 30 seconds, heat loading on the inner wall becomes a problem because the SPARC chamber is compact and the neutron flux is intense.
The key innovation is the magnet. SPARC uses high-temperature superconducting (HTS) tape—specifically REBCO (rare-earth barium copper oxide) superconductor—to produce magnetic fields roughly twice as strong as ITER’s conventional superconducting magnets. Stronger fields mean you can confine the same plasma in a much smaller volume, which means a much smaller, cheaper, faster-to-build machine. SPARC is about 24 feet across. ITER is the size of a building. CFS demonstrated a full-scale HTS magnet producing a field of 20 tesla in September 2021, which was the engineering milestone that convinced most of the fusion physics community that the compact tokamak approach was viable. The company has raised nearly $3 billion, including investments from Google, Nvidia, and Breakthrough Energy Ventures.
If SPARC succeeds, the commercial plant is ARC—a 400-megawatt machine planned for a site in Chesterfield County, Virginia, in collaboration with Dominion Energy, with grid power targeted for the early 2030s. Google has signed a 200-megawatt power purchase agreement. That’s not a research grant. That’s a customer.
Helion Energy is taking a fundamentally different approach—pulsed field-reversed configuration fusion, using deuterium and helium-3 fuel instead of deuterium-tritium, which avoids the tritium supply problem entirely but requires a much harder fusion reaction to achieve. Helion has signed a power purchase agreement with Microsoft to deliver 50 megawatts from its Polaris plant in Washington state by 2028, backed by $425 million in funding led by Sam Altman. Whether Helion can achieve its stated timeline is an open question—the field-reversed configuration approach has less experimental history than the tokamak, and the company’s claims have drawn skepticism from mainstream fusion physicists.
TAE Technologies, in one of the stranger corporate developments of 2025, merged with Trump Media & Technology Group in a $6 billion deal that will make it the first publicly traded fusion company—and partially owned by the Trump family. TAE uses a beam-driven field-reversed configuration targeting proton-boron-11 fusion, which is aneutronic (no neutrons, no radioactive waste, no tritium needed) but requires even higher temperatures than D-T and has even less experimental validation. The merger injects capital into a company that has been operating for over two decades; whether it injects physics remains to be seen.
Pacific Fusion, one of the newest entrants, raised $900 million in Series A funding—the largest initial round in fusion history—from Eric Schmidt, Patrick Collison, Reid Hoffman, and other Silicon Valley investors. Pacific is pursuing inertial confinement fusion, the approach used by NIF, but with a pulsed power driver instead of lasers. The company is deliberately quiet about its technical details, which is either prudent IP protection or a sign that the details don’t yet support the valuation.
The “done” question
Here’s the part that most fusion coverage skips: achieving Q > 1 is not the finish line. It’s the starting line. A fusion reaction that produces net energy for 10 seconds in a laboratory is a physics milestone. A fusion power plant that produces net electricity, continuously, reliably, for decades, at a cost competitive with fission, natural gas, and renewables—that’s a different problem entirely, and it involves challenges that have nothing to do with plasma physics.
Materials. The inner wall of a fusion reactor is exposed to a neutron flux that degrades materials at a rate for which we have almost no operational data, because no fusion machine has run long enough at high enough power to test it. ITER’s plasma-facing components use beryllium and tungsten, and even those are expected to require periodic replacement.
Maintenance. A fusion reactor is radioactive—not from the fuel, but from the neutron activation of structural materials. Maintenance must be performed robotically behind heavy shielding, on a machine operating at temperatures from minus 269°C (the magnets) to 100 million degrees (the plasma) with everything in between.
Electricity conversion. Fusion produces heat. That heat must be captured and converted to electricity through steam turbines or another thermal cycle, which means the plant has all the complexity of a conventional thermal power plant in addition to the fusion-specific systems. The overall electrical efficiency—from fusion energy produced to electricity delivered to the grid—must be high enough that the plant is economically viable after paying for fuel, operations, maintenance, and capital costs.
Regulatory framework. There is currently no licensing pathway for commercial fusion plants in most countries. The U.S. Nuclear Regulatory Commission has been working on a framework, but it’s not finalized. You can’t build a power plant if there’s no regulatory process for approving it.
None of these problems is unsolvable. All of them are the kind of grinding, unglamorous engineering challenges that don’t generate press releases or attract venture capital but determine whether a technology actually ships. The gap between “SPARC achieves Q > 2 in 2027” and “you can buy electricity from a fusion plant” is measured in these problems, and the honest estimate for closing that gap—from the most optimistic credible voices in the field—is the early to mid 2030s, with widespread deployment following over the subsequent decade.
Kim Budil, the director of Lawrence Livermore, put it well at Davos in early 2026: “Historically, we’ve always said fusion energy is 30 years away from whatever day you ask and will always be that. I think that’s not true anymore. But fusion is hard. There’s a lot of work to be done.”
That’s the right framing. Not “it’s here.” Not “it’ll never happen.” It’s close in a way it has never been close before, the money is real, the engineering is advancing, and the physics has been demonstrated. But the path from physics demonstration to power plant passes through a decade of materials science, manufacturing, regulation, and the kind of reliability engineering that turns a prototype into something that runs on a random Tuesday without a team of PhDs babysitting it.
We cover fusion power in depth—the physics, the engineering, the economics, every company and approach, and the constraints that determine whether any of it actually reaches the grid—as the opening lecture of our Moonshot 2169 course. If the thirty-years-away joke bothers you as much as it should, that’s the place to get the full picture.