- Nov 16, 2026: Best Of inline snippet B0D7PPG25F at midpoint

Image: Ken Filar/MIT Plasma Science and Fusion Center, 2018, via Wikimedia Commons (CC BY-SA 4.0)
In a 60-acre campus off Hospital Road in Devens, Massachusetts, northwest of Boston, a private company is bolting together a machine that is meant to do something no magnetic fusion device has ever done: release more energy from fusion than it takes to heat the fuel. The machine is called SPARC, the company is Commonwealth Fusion Systems (CFS), and as of November 2026 its chief executive says it is about 80% built and due to switch on in 2027.
The short answer: SPARC is a compact tokamak, a doughnut-shaped magnetic bottle for a 100-million-degree plasma, built around a new kind of high-temperature superconducting magnet. CFS expects to start operating it in 2027 and then push as fast as it can to "Q>1," the point where the fusion reactions put out more power than the heating that sustains them. SPARC will not make electricity; it is a physics proof. If it works, CFS plans to use what it learns to build ARC, a roughly 400-megawatt fusion power plant in Chesterfield County, Virginia, that Google and the Italian energy company Eni have already agreed to buy power from in the early 2030s.
That is a big "if," and it is worth being precise about what SPARC is, what Q>1 does and does not mean, why the magnets matter so much, and what could still go wrong. Here is the full picture, with the numbers.
Where SPARC stands in November 2026
SPARC sits inside a building CFS calls Tokamak Hall on its Devens campus, next to the factory that makes its magnets. The machine needs 18 D-shaped toroidal field magnets, the big coils that wrap around the doughnut. CFS shipped the first one, a 24-ton steel-clad magnet, across the campus from its factory to Tokamak Hall in December 2025 and announced it on January 6, 2026. By May 14, 2026, two of those magnets were in place, the 96-ton vacuum vessel had arrived in two halves for fitting out, and the company described the machine as "nearly 75% complete."
The most recent public updates go further. In an August 2026 Utility Dive report on a new $1 billion funding round, CEO Bob Mumgaard put SPARC at "about 80% complete" and said the company expects to show net energy gain in 2027. On November 2, 2026, ahead of the American Physical Society’s plasma physics meeting, CFS’s vice president of plasma physics, Phil Snyder, said: "We’re in the home stretch now. We’re assembling SPARC, and we expect to turn it on next year." And on November 13, Mumgaard told Semafor the machine is about 80% complete and due to start up in 2027.

Photo: 4300streetcar, 2026, via Wikimedia Commons (CC BY 4.0)
The schedule has already slipped once. When SPARC was announced in 2018 it was expected to be finished in 2025, and early reporting on the project used that date. The current target, first plasma in 2027 and Q>1 as soon as possible afterward, is the one CFS’s co-founder and chief science officer Brandon Sorbom has repeated this year.
Here is the machine in numbers, from the peer-reviewed design overview CFS and MIT researchers published in the Journal of Plasma Physics in September 2020 and from CFS’s own updates:
Size: a major radius of 1.85 meters and a minor radius of 0.57 meters, about the size of today’s "medium-sized" research tokamaks such as DIII-D in San Diego.
Magnetic field: 12.2 tesla at the center of the plasma, which CFS says is about 400,000 times stronger than Earth’s magnetic field at the equator.
Plasma: heated to roughly 100 million degrees Celsius, with a plasma current of 8.7 million amperes in the full-performance scenario.
Heating: up to 25 megawatts of radio-frequency heating at 120 megahertz, the machine’s only source of outside heating.
Output: a projected 140 megawatts of fusion power in 10-second pulses, with a predicted gain of about 11 under nominal assumptions and above 2 even under pessimistic ones.
Fuel: deuterium and tritium, the two heavy forms of hydrogen, with no more than 10 grams of tritium on site over the facility’s lifetime.
What "Q>1" actually means, and what it doesn’t
Fusion headlines tend to blur three different kinds of "breakeven." SPARC is aiming at the first one. The other two matter more for your electric bill.
Plasma gain: the number SPARC is chasing
In a tokamak, Q is the fusion power the plasma produces divided by the heating power absorbed by the plasma. At Q=1, the plasma releases as much fusion power as is being pumped into it. CFS’s goal for its first campaign is Q>1; the 2020 design papers set the machine’s mission at Q>2 and projected about Q=11 with standard assumptions about how well the plasma holds its heat. For comparison, the record for magnetic confinement still belongs to the Joint European Torus (JET) in England, which in 1997 produced 16 megawatts of fusion power from 24 megawatts of heating, a Q of 0.67, for about a second. JET later set the record for total fusion energy in a single pulse, 69.26 megajoules over about six seconds in late 2023, announced in February 2024, before it shut down for good at the end of that year. It never crossed Q=1.
Target gain: what the National Ignition Facility did
The one machine that has produced more fusion energy than it put into the fuel is not a tokamak. The National Ignition Facility (NIF) at Lawrence Livermore National Laboratory in California fires 192 lasers at a peppercorn-sized fuel capsule. On December 5, 2022, it delivered 2.05 megajoules of laser light and got more than 3 megajoules of fusion energy back, the first lab "ignition." On April 7, 2025, it set its current record: 8.6 megajoules from 2.08 megajoules of laser energy, a target gain of about 4.1. CFS’s own blog acknowledges NIF as the only machine to have crossed the threshold so far, and frames SPARC as the first attempt with a commercial mission behind it.

Photo: Lawrence Livermore National Security, 2008, via Wikimedia Commons (CC BY-SA 3.0)
Engineering gain: the one that pays the bills
Neither number counts the electricity needed to run the whole machine. To deliver the 2.05-megajoule shot in 2022, NIF’s laser system needed about 300 megajoules of energy, Livermore director Kim Budil said at the announcement, roughly a hundred times the fusion energy that came out. A tokamak also has to power its magnets’ cooling, its heating systems, pumps and controls, and then turn fusion heat into electricity through a steam cycle that recovers well under half of that heat as electricity. That is why SPARC’s designers wanted a gain closer to 10 than 1: a power plant needs plasma gain well above 1 just to break even on electricity. CFS’s own site is blunt about it: "SPARC is not a power plant connected to the grid." SPARC will not put a single watt on the grid. Its job is to prove the physics and the magnets at once.
Why the magnets are the whole story
Tokamaks have been around since Soviet physicists Andrei Sakharov and Igor Tamm proposed the design in 1950, and roughly 150 of them have been built worldwide, a figure CFS itself cites. The physics of how well they hold heat is well studied. What has limited them is size and cost: to get to high gain with the superconducting magnets available until recently, you had to build something enormous. ITER, the international project under construction in southern France, is designed to produce 500 megawatts of fusion power from 50 megawatts of heating, a gain of 10, but its 2024 revised schedule puts the start of deuterium-tritium operation in 2039, with about 5 billion euros in added cost for the ITER Organization alone.
The alternative is a stronger field. Fusion power density rises steeply with magnetic field strength, roughly with the fourth power of the field at a fixed plasma pressure ratio, so a machine with twice the field can get similar performance at a fraction of the size. The 2020 SPARC overview makes the point directly: gain depends on both field and size, and "by increasing field, one can decrease size and obtain the same gain without changing any physics assumptions."
That is where the superconductor comes in. Older fusion magnets use materials that have to be chilled to about 4 kelvins, 4 degrees above absolute zero. SPARC’s magnets use REBCO, short for rare-earth barium copper oxide, a high-temperature superconductor made as thin, flat tape. It works at about 20 kelvins and can carry far more current in a strong field. On September 5, 2021, in a test stand at MIT’s Plasma Science and Fusion Center, a full-scale model coil built by MIT and CFS reached a steady field of just over 20 tesla, a world record for a large high-temperature superconducting magnet. The 20,000-pound coil contained 267 kilometers of tape, roughly the distance from Boston to Albany.
Dennis Whyte, then director of the MIT center, later called the test "the most important thing, in my opinion, in the last 30 years of fusion research," and said it "changed the cost per watt of a fusion reactor by a factor of almost 40 in one day." That is his estimate, not an audited figure, but it shows why the field took notice.
The test also included a deliberate failure. In late 2021 the team pushed the coil into a "quench," a sudden loss of superconductivity that dumps the magnet’s stored energy as heat. The magnet used a radical "no-insulation" design, with bare superconducting tape wound into 16 steel plates called pancakes, and one corner of one pancake melted. In a set of six peer-reviewed papers published in IEEE Transactions on Applied Superconductivity in March 2024, the team reported that its best models predicted the damage almost exactly and that the design was revised to avoid it in SPARC’s magnets. Whyte’s summary: "a few percent of the volume of the coil" was damaged.
CFS has since turned that prototype into a production line. Each SPARC toroidal field magnet is a stack of 16 pancakes in a steel case, and the company says its factory is now running at production pace. "We didn’t just make a single magnet. We made a whole manufacturing facility that can make many, many magnets," Sorbom said in January. The first magnet left that factory in December 2025; the company said then that it had 17 more to deliver.
From Alcator to SPARC: a 50-year lineage at MIT
SPARC did not come from nowhere. MIT has built high-field, compact tokamaks since the 1970s under the name Alcator, and the last of them, Alcator C-Mod, ran from 1991 to 2016. On its final day of operation, September 30, 2016, after federal funding ended, C-Mod set a world record for plasma pressure in a magnetic fusion device, 2.05 atmospheres. Much of SPARC’s physics basis, its heat-exhaust research and even the radio-frequency heating hardware it is scoped around trace back to that machine.

Photo: Mike Garrett, 2012, via Wikimedia Commons (CC BY 3.0)
Commonwealth Fusion Systems spun out of that program in 2018. On March 9 of that year, MIT announced the collaboration and a $50 million investment from Eni, with a stated goal of a working pilot plant within 15 years. In September 2020, CFS and MIT published seven peer-reviewed papers on SPARC’s physics in the Journal of Plasma Physics, the basis for the Q of about 11 projection. In March 2021, the company picked its Devens site. Since then it has raised about $4 billion in total, according to the company, including a $1.8 billion round in 2021, an $863 million round in 2025 and a further $1 billion announced on July 30, 2026. Google, an investor since 2021, has added to its stake. Semafor, citing Axios, reports that Abu Dhabi’s government-owned fusion fund, Plynth Energy, took a minority stake worth more than $100 million this year.
How CFS plans to reach Q>1
SPARC will not start with its most dangerous fuel. The 2020 design overview lays out early campaigns in helium, hydrogen and deuterium to learn how the machine behaves, followed by a move "as rapidly as possible" to deuterium-tritium operation, the only fuel mix that gives enough fusion power to cross Q>1.
At this month’s American Physical Society Division of Plasma Physics meeting, CFS researchers described two main paths to the first goal. The simpler one runs the magnets at their full 12.2 tesla with the plasma in "low confinement mode," or L-mode. The other uses "high confinement mode," or H-mode, a more demanding state with a hotter edge and higher plasma pressure that could reach Q>1 at a lower field of 8 to 9 tesla. "Because SPARC is designed to run well beyond Q>1, we have lots of room to maneuver as we operate the machine," said Clayton Myers, who leads the company’s plasma diagnostics team. After Q>1, the plan is to push both full field and H-mode together toward much higher gain, into what physicists call burning plasma, where heat from the fusion reactions themselves does much of the heating.
SPARC will not put a single watt on the grid. Its job is to prove the physics and the magnets at once.
Software is a bigger part of that plan than it was for earlier tokamaks. In October 2025, Google DeepMind announced a research partnership with CFS built around TORAX, its open-source plasma simulator, which CFS uses to run what DeepMind describes as "millions of virtual experiments before SPARC is even turned on." The two companies are also testing reinforcement learning to steer the plasma in real time, including sweeping the intense exhaust heat across the wall so it does not burn one spot. It is a cousin of the AI plasma control at DIII-D that Princeton researchers demonstrated earlier.
Tritium, radiation and the rules
SPARC’s fuel is half deuterium, a stable isotope extracted from water, and half tritium, a radioactive form of hydrogen with a half-life of about 12 years. Tritium gives off weak beta radiation, the same kind that lights self-luminous exit signs and some watch dials. CFS says there will be only about 10 grams of tritium on site in total, used over the life of the facility. The bigger radiation issue is neutrons: each deuterium-tritium fusion reaction throws off a fast neutron, which is why SPARC sits inside thick concrete walls and why parts of the machine will become radioactive over time.

Photo: Gazebo, 2014, via Wikimedia Commons (CC BY-SA 4.0)
Unlike a fission reactor, a tokamak cannot run away. There is no chain reaction; if power or containment is lost, the plasma cools and the reaction stops within moments. U.S. regulators have treated it accordingly. In April 2023 the Nuclear Regulatory Commission voted unanimously to regulate fusion machines under its byproduct-material framework, the rules used for particle accelerators and medical isotopes, rather than the rules for fission power plants. Congress wrote that approach into law in the ADVANCE Act in July 2024. Because Massachusetts is an NRC "agreement state," SPARC itself is licensed by the state: on October 24, 2024, the Massachusetts Radiation Control Program granted CFS a broad-scope radioactive materials license covering the machine.
The national rulebook is still being finished. The NRC published its proposed rule for fusion machines on February 26, 2026, and the public comment period closed on May 27. As of mid-November, the agency’s rulemaking page still lists the proposed rule as the latest milestone, with no final rule published yet.
From SPARC to ARC: Virginia, Google and Eni
SPARC’s real purpose is commercial. On December 17, 2024, CFS announced that its first power plant, ARC, would be built at the James River Industrial Center in Chesterfield County, Virginia, on land leased from Dominion Energy, which is also contributing development and technical help. CFS says it will finance, build, own and operate the plant, which it calls the Fall Line Fusion Power Station, and that ARC will produce about 400 megawatts, enough for about 150,000 Virginia homes, starting in the early 2030s. The company has applied to connect the plant to PJM, the grid operator for much of the eastern United States.
Two buyers have signed up. On June 30, 2025, Google agreed to buy 200 megawatts, half of the plant’s output, in its first commercial commitment to fusion, with an option on power from future ARC plants. CFS’s announcement says the deal is anchored in SPARC achieving Q>1. On September 22, 2025, Eni, one of its earliest investors, signed an offtake agreement worth more than $1 billion. Under those contracts, CFS told Utility Dive, ARC’s electricity will be sold into the wholesale market through Dominion’s system while the buyers pay for and receive the clean-energy credits.
The pattern will look familiar to anyone who has followed the tech industry’s scramble for firm, carbon-free power. Microsoft is paying to bring a Three Mile Island reactor back as the Crane Clean Energy Center, and it has a separate deal with Helion, whose pulsed machine takes a very different path, covered in our explainer on Helion’s Polaris. Google’s bet on CFS is the tokamak version of the same idea: lock in power from a technology that does not exist yet, at a price that only makes sense if the physics works.
What could still go wrong
The schedule. SPARC has already moved from 2025 to 2027, and first plasma is not Q>1. Early operation in hydrogen and deuterium could take months before the machine is ready for tritium, and the company has not given a date for the Q>1 shot beyond "as fast as humanly possible," as Sorbom put it this year.
The magnets at scale. One full-scale coil worked in 2021. SPARC needs 18 toroidal field magnets plus a central solenoid and poloidal field coils, all working together with steady current, and a quench in a fully assembled machine would be far more costly than one in a test stand.
The plasma itself. SPARC’s design margin is real, Q>2 even if the plasma holds its heat only 70% as well as standard scaling predicts, but no one has run a burning plasma in a field this strong. Instabilities and disruptions, sudden losses of the plasma that slam energy into the walls, are a known hazard in every tokamak.
Heat and materials. At full power SPARC will dump intense heat onto small areas of tungsten at the top and bottom of the vessel. CFS says each plasma-facing piece is placed with 200-micron precision, about the width of two human hairs. That is an engineering problem SPARC must solve over 10-second pulses; ARC has to solve it around the clock.
The power plant problem. SPARC does not need a blanket to breed tritium or capture neutron heat. ARC does. "Blanket technology is immature globally," Andrew Sowder, a senior technical executive at the Electric Power Research Institute, told Utility Dive, though he added that "it can be done at scale." CFS has joined the UK’s LIBRTI tritium-breeding program to test blanket technology in parallel with SPARC.
What to watch in 2027
The milestones to look for, in rough order, are the last of the 18 toroidal field magnets going in; the two vacuum vessel halves being joined; the cooldown of the magnets to operating temperature; first plasma; the first deuterium-tritium shots; and finally a claim of Q>1, which will matter most once it is backed by published data and outside review. CFS has said it plans to keep publishing its work and presenting it to peers, as it did this month.
The broader field will be watching too. According to the Fusion Industry Association’s 2026 report, as summarized by Utility Dive, only two other U.S. fusion companies, Helion and Type One Energy, have power offtake agreements, and no private company has yet shown net energy gain. A clean Q>1 on SPARC would not make fusion electricity real, but it would move the argument from "can it work" to "can it be built cheaply, again and again."
The bottom line
SPARC is the most closely watched fusion machine in the private sector because it combines conservative tokamak physics with an aggressive bet on new magnets. The magnet part has largely been proven at full scale; the physics part will be tested in 2027. If SPARC reaches Q>1, CFS will have done what JET and every other magnetic fusion machine could not, and Google and Eni’s early-2030s power deals will look a lot less speculative. If it struggles, the reasons will be specific and measurable, and the whole field will learn from them. Either way, the next year in Devens will say more about fusion’s real timeline than another decade of promises.
Further reading

The Star Builders: Nuclear Fusion and the Race to Power the Planet
Physicist Arthur Turrell’s tour of the labs and startups racing to make fusion work, from NIF’s lasers to tokamaks like SPARC.

Sun in a Bottle: The Strange History of Fusion and the Science of Wishful Thinking
Charles Seife’s skeptical history of fusion’s broken promises, a useful counterweight to every "just around the corner" headline.

SNAP CIRCUITS Green Energy STEM Electronics Kit
More than 125 snap-together projects on solar, wind and hand-crank power, a hands-on way into how electricity gets made.

4M Kidzlabs Anti Gravity Magnetic Levitation Science Kit
A small maglev kit that shows how magnetic fields can hold an object in place without touching it, the same idea a tokamak uses on plasma.

SNAP CIRCUITS Explorer SC-100 STEM Electronics Kit
The best-selling beginner electronics kit: 100-plus projects that teach how current, switches and electromagnets work.
