Most fusion timelines are slides. Commonwealth Fusion Systems has a building. In Devens, Massachusetts, its SPARC tokamak was almost 80% complete as of August 5, 2026, its cryogenic plant has run the full system at 8 kelvin, and the company says operations will begin in 2027.
That makes Commonwealth Fusion Systems (CFS) the private fusion company with the clearest near-term physics test on the calendar. It's also the one with the most money: CFS says it has raised $4 billion, about 30% of all capital raised by the fusion industry.
This full guide covers what SPARC is, why its high-temperature superconducting (HTS) magnets matter, the real timeline, ARC in Virginia, the Google and Eni deals, funding, how CFS compares with Helion, TAE, Tokamak Energy and ITER, and the honest limits. For the wider field, start with Inside Deep Tech's roundup of the fusion startups closest to commercial power.
Last updated: October 6, 2026.
Key Takeaways
- SPARC was about 80% complete in August 2026; CFS targets first operations in 2027.
- The design rests on a 20 tesla HTS model coil tested September 5, 2021 with MIT.
- ARC, a roughly 400 MW plant in Chesterfield County, Virginia, targets the grid in the early 2030s.
- Google signed for 200 MW and Eni for more than $1 billion; CFS has raised $4 billion.
- Honest limits: SPARC slipped from a 2025 target, and net tritium breeding is still undemonstrated.
What Commonwealth Fusion Systems is building in October 2026
CFS spun out of MIT in 2018 and is building two machines in sequence. SPARC is the demonstration tokamak at its Devens headquarters. ARC is the first power plant, planned for the company's Fall Line Fusion Power Station in Chesterfield County, Virginia, per CFS's July 2026 funding release.
The two are tightly linked by design. CFS says SPARC "proves most of the technology" ARC will need, and that lessons from more than 1,000 employees building SPARC are already feeding the ARC design (CFS, August 2026).
The core idea is simple to state. Keep conventional tokamak physics, swap in much stronger magnets, and shrink the machine. In 2021, MIT's Martin Greenwald described the niche as "conventional plasma physics, and conventional tokamak designs and engineering" combined with "this new magnet technology" (MIT News).
SPARC and ARC at a glance
| Parameter | SPARC (demonstration) | ARC V3A (power plant) |
|---|---|---|
| Site | Devens, Massachusetts | Chesterfield County, Virginia |
| Major radius / minor radius | 1.85 m / 0.57 m | 4.62 m / 1.18 m |
| Magnetic field on axis | 12.2 T | 11.4 T |
| Fusion power (design) | about 140 MW (nominal case) | about 1.13 GW |
| Headline goal | Q>2 mission, Q of about 11 nominal | at least 400 MW net electric |
| Status, Oct 2026 | almost 80% complete, operations in 2027 | Pre-construction, PJM application filed |
Source: Overview of the SPARC tokamak, Journal of Plasma Physics (2020), OSTI record, ARC physics basis overview, Journal of Plasma Physics (2026), and CFS.
Why the HTS magnets are the whole bet
Fusion power in a tokamak scales steeply with magnetic field, so a stronger field lets a smaller machine reach the same plasma conditions. CFS's magnets use REBCO (rare earth barium copper oxide) superconducting tape, which MIT says can match the performance of a conventional low-temperature superconducting device about 40 times larger in volume (MIT News).
The proof point came on September 5, 2021. CFS and MIT's Plasma Science and Fusion Center ramped a large-bore HTS toroidal field model coil to 20 tesla, the strongest field of its kind at the time. The coil used 267 km of tape across 16 stacked plates (MIT News).
SPARC needs 18 D-shaped toroidal field (TF) magnets, each about 24 tons, per CFS's January 2026 magnet delivery post. Each TF magnet is a stack of 16 "pancakes" built with CFS's non-insulated, non-twisted (NINT) design, and the machine also uses poloidal field and central solenoid HTS magnets.
For the materials science behind REBCO and why it changed the size of fusion machines, see Inside Deep Tech's guide to high-temperature superconductors in fusion.
The SPARC timeline: from magnet test to first plasma
| Date | Milestone | Source |
|---|---|---|
| Sep 5, 2021 | 20 T HTS model coil test with MIT | MIT News |
| Dec 1, 2021 | Series B of more than $1.8 billion, led by Tiger Global | CFS |
| May 31, 2023 | DOE picks CFS among 8 Milestone program awardees | DOE |
| Dec 17, 2024 | ARC site announced in Chesterfield County, Virginia | CFS |
| Jun 30, 2025 | Google signs 200 MW offtake from the first ARC plant | CFS |
| Aug 28, 2025 | $863 million Series B2 | CFS |
| Sep 22, 2025 | Eni signs offtake worth more than $1 billion | CFS |
| Jan 6, 2026 | First of 18 TF magnets delivered to SPARC | CFS |
| Apr 9, 2026 | About 75% complete; 2 TF magnets and both vacuum vessel halves in Tokamak Hall | CFS |
| Jul 1, 2026 | First international partner in UKAEA's LIBRTI tritium program | CFS |
| Jul 30, 2026 | $1 billion raise; $4 billion total | CFS |
| Aug 5, 2026 | DOE approves ARC milestones; SPARC almost 80% complete | CFS |
| Sep 16, 2026 | Full cryogenic system runs at 8 K | CFS |
| Sep 30, 2026 | Fujikura order for more than 10,000 km of HTS tape | CFS |
| 2027 (target) | SPARC operations begin, then the push to Q>1 | CFS |
Source: CFS newsroom, The Tokamak Times, MIT News, and DOE; each row links its primary source.
Two details matter more than the headline percentages. First, the magnet count. CFS's April tour reported 2 of 18 TF magnets in Tokamak Hall, while CEO Bob Mumgaard told TechCrunch in January he expected all 18 installed by the end of summer. CFS's August update says installation is still underway and gives no new count.
Second, the support systems. The cryoplant, magnet power and radio-frequency heating systems are being commissioned in parallel with assembly, ahead of a "dry dress rehearsal" that runs them as if SPARC were pulsing (CFS). CFS is also building a digital twin of the machine with NVIDIA and Siemens (TechCrunch).
What Q>1 on SPARC means, and what it doesn't
Q is fusion power divided by the external heating power absorbed by the plasma. The peer-reviewed SPARC overview sets a mission goal of Q>2 under conservative assumptions and projects Q of about 11 and about 140 MW of fusion power in its nominal case (Journal of Plasma Physics).
CFS's commercial deals key off a lower bar. Google's offtake agreement is "anchored in CFS' SPARC achieving net fusion energy, known as Q>1" (CFS).
Here's the catch. Q>1 is plasma gain, sometimes called Q-scientific. It ignores the electricity needed to run magnets, cryogenics, heating systems and pumps, and the losses in converting heat to power. Engineering gain, measured at the plant's electrical meter, is the number a utility pays for.
ARC's design shows the gap. Its physics basis targets about 1.13 GW of fusion power to deliver at least 400 MW of net electricity (Journal of Plasma Physics). By Inside Deep Tech's arithmetic, that's roughly 35% of fusion power ending up as net grid power, after every internal load.
ARC: the Virginia power plant
On December 17, 2024, CFS said it would build ARC at the James River Industrial Center in Chesterfield County, Virginia, and would independently finance, build, own and operate it. Dominion Energy Virginia owns the site and provides non-financial collaboration and leasing rights (CFS).
CFS describes ARC as a plant of about 400 MW. Its 2024 release equated that to about 150,000 homes (CFS), while its August 2026 post says 280,000 average American households (CFS). The household figure depends on the assumptions used, so plan on the megawatts.
The commercial groundwork is moving ahead of the physics. CFS says it was the first fusion company to submit an application to PJM Interconnection, the largest U.S. wholesale electricity market (CFS), and it's working on a zoning permit and site development (CFS).
The design is modular on purpose. The ARC physics basis notes that early SPARC results can be folded into the first ARC and into "subsequent replacements of the ARC vacuum vessel" (Journal of Plasma Physics). That's a design choice that admits in-vessel components will wear out.
Who's buying ARC power: Google, Eni, and Dominion
Google moved first. On June 30, 2025, it signed a power purchase agreement for 200 MW from the first ARC plant, with an option on power from future plants, and increased its equity stake (CFS). Google called it the largest direct corporate offtake agreement for fusion energy.
Eni followed on September 22, 2025 with an offtake agreement worth more than $1 billion. Eni has been a CFS shareholder since 2018, and the companies didn't disclose financial terms (CFS).
Together, Google and Eni have agreed to buy more than half of the plant's output, per CFS's July 2026 release. Dominion's role is site and development support, not offtake or investment, per the 2024 announcement.
The buyer list tells you who fusion is really for. Hyperscalers want firm, carbon-free power near data center clusters, and Virginia hosts the biggest one. Inside Deep Tech covers the nearer-term options in its guides to small modular reactors for AI data centers whether fusion can power the AI data center boom, and behind-the-meter generation.
Funding: $4 billion and who's paying
| Round | Date | Amount | Notable investors | Source |
|---|---|---|---|---|
| Series B | Dec 1, 2021 | More than $1.8 billion | Tiger Global (lead), Bill Gates, Google, Eni, Temasek | CFS |
| Series B2 | Aug 28, 2025 | $863 million | NVentures (NVIDIA), Google, Eni, Breakthrough Energy Ventures | CFS |
| Equity raise | Jul 30, 2026 | $1 billion | Pension, sovereign wealth, infrastructure and industrial investors | CFS |
| Strategic add-on | Sep 10, 2026 | Part of the $1 billion | Hyundai Motor Group | CFS |
| Total raised | As of Jul 30, 2026 | $4 billion | About 30% of fusion industry capital | CFS |
Source: CFS Series B release, Series B2 release, July 2026 release, and Hyundai release.
CFS's Asian backers now include a consortium of Japanese industrial companies led by Mitsui and Mitsubishi, Temasek of Singapore, Woori Venture Partners in Korea, and Hyundai, per the September 2026 release. CFS didn't name the July 2026 investors individually or disclose a valuation.
Public money is a small slice. DOE announced $46 million in May 2023 for eight companies in its Milestone-Based Fusion Development Program (DOE). CFS says it funds more than two-thirds of its milestone costs and that DOE's 14-expert panel approved its ARC preconceptual design and technology roadmap in 2026 (CFS).
Supply chain: tape, cryogenics, and scale
HTS tape is the long pole for ARC. On September 30, 2026, CFS agreed to buy more than 10,000 km of HTS tape from Fujikura, which is expanding capacity to deliver it, starting with the Virginia plant (CFS).
For scale, the 2021 model coil used 267 km of tape (MIT News). By Inside Deep Tech's arithmetic, the Fujikura order is about 37x that test coil, which shows why tape supply sits in CFS's critical path (CFS).
Cryogenics is the other piece. CFS says SPARC's system cools helium to 8 kelvin and is essentially the same equipment ARC will need, though ARC's version will drop the blowdown tanks and is meant to run for decades (CFS).
How SPARC compares with Helion, TAE, Tokamak Energy, and ITER
| Program | Approach | Status, Oct 2026 | Next commercial step | Source |
|---|---|---|---|---|
| CFS SPARC / ARC | HTS tokamak, D-T fuel | SPARC almost 80% complete; operations 2027 | ARC, about 400 MW, early 2030s | CFS, CFS |
| Helion Polaris / Orion | Pulsed FRC, D-He3 for commercial use | First private D-T fusion; 150 million °C (Feb 2026) | Orion for Microsoft; full 50 MW not before 2029 or 2030 | Helion, Axios |
| TAE Norm / Da Vinci | Beam-driven FRC, hydrogen-boron goal | NBI-only FRC formation; Copernicus skipped (Nov 2025) | Da Vinci prototype plant; merger with TMTG pending | TAE, TAE |
| Tokamak Energy | Spherical tokamak, HTS magnets | Demo4 magnets hit 13.7 T in a 14-month campaign | STEP magnet partner, £70 million contract to 2029 | Tokamak Energy |
| ITER | LTS tokamak, intergovernmental | Assembly; full magnetic energy in 2036 | D-T operations in 2039, Q≥10 goal | ITER |
Source: CFS, Helion, Axios, TAE, TAE and TMTG S-4 release, Tokamak Energy, ITER.
The comparison isn't apples to apples. Helion has already run D-T fuel in Polaris, which SPARC hasn't, but its field-reversed configuration has a far thinner experimental record than tokamaks. Axios reports that two 2026 academic papers questioned assumptions behind Helion's approach, and that Helion says full 50 MW operation at Orion won't come before 2029 or 2030 (Axios).
TAE skipped its planned Copernicus device after Norm's results and is moving to its Da Vinci plant (TAE). Its all-stock merger with Trump Media & Technology Group, agreed December 18, 2025, had an S-4 filed on September 30, 2026 that isn't yet effective (TAE).
ITER is the useful baseline. Its 2024 plan puts deuterium-tritium operations in 2039 and adds about EUR 5 billion in cost for the ITER Organization, with a goal of 500 MW of fusion power from 50 MW of heating (ITER). SPARC is trying to reach burning-plasma physics a decade earlier in a much smaller machine.
Honest limits
The real limits as of October 2026:
- Schedule. In 2021, SPARC was targeted for completion in 2025, and the Series B roadmap said 2025: SPARC achieves net energy. Operations now begin in 2027.
- Physics still to prove. The ARC physics basis itself lists model uncertainties and physics risks to be retired through SPARC operation.
- Plasma gain is not plant gain. ARC needs about 1.13 GW of fusion power for at least 400 MW net; a Q>1 shot on SPARC produces no electricity.
- Tritium breeding. ARC must make its own fuel. CFS joined UKAEA's £220 million LIBRTI program to demonstrate net tritium production; that test hasn't happened yet.
- Neutron damage. MIT found the instant "beam on" effect on REBCO isn't a concern, but long-term degradation over years or decades is still being studied, and ARC plans vacuum vessel replacements.
- Cost. Neither Google nor Eni disclosed PPA pricing, and CFS hasn't published an ARC capital cost.
- Supply. ARC depends on more than 10,000 km of HTS tape from a supplier that's still expanding capacity.
Frequently asked questions
What is Commonwealth Fusion Systems SPARC?
SPARC is a compact, high-field tokamak that Commonwealth Fusion Systems is building in Devens, Massachusetts, to show net fusion energy. Its peer-reviewed design uses a 12.2 T field and a 1.85 m major radius.
When will SPARC turn on?
CFS says SPARC operations will begin in 2027, per its May 2026 vacuum vessel update. The machine was almost 80% complete in August 2026.
Has Commonwealth Fusion Systems achieved net fusion energy yet?
No. SPARC hasn't made a plasma yet, so Q>1 is still a target. Google's 200 MW offtake is anchored on SPARC reaching Q>1.
Where will the ARC fusion power plant be built?
ARC is planned for the James River Industrial Center in Chesterfield County, Virginia, on land leased from Dominion Energy Virginia, per CFS. CFS targets grid power in the early 2030s.
How much money has Commonwealth Fusion Systems raised?
CFS says it has raised $4 billion, including a $1 billion round in July 2026, the $863 million Series B2 and the $1.8 billion Series B.
Is Commonwealth Fusion Systems a public company?
No. CFS is privately held, backed by investors including Google, Eni, NVIDIA's NVentures and Hyundai Motor Group, per its September 2026 release.
How is SPARC different from ITER?
What to watch next
Three signals will tell you whether the timeline holds: all 18 TF magnets installed, a completed dry dress rehearsal, and SPARC's first plasma in 2027 (CFS). After that, the number to watch is how fast CFS climbs from first plasma to Q>1.
If you're planning power for 2030s data centers, treat ARC as an option with a physics gate, not as firm capacity. Keep it in the plan next to gas, SMRs and on-site generation.
