In early May 2026, a 48-ton steel half-shell was lowered by crane onto the SPARC tokamak assembly floor at Commonwealth Fusion Systems’ Devens, Massachusetts campus. With that single lift, the vacuum vessel — the doughnut-shaped chamber that will hold the fusing plasma — became geometrically whole. The machine that intends to prove commercial fusion is real is now structurally 75 percent complete.
That number deserves to be sat with. Five years ago, the credible case against commercial fusion was that the machines were not being built. They were being modeled, simulated, and rendered. SPARC is now a real object with a measurable percent-complete number against a CPM schedule. Half of the eighteen toroidal field magnets are in their stands. The cryostat panels are staged. The central solenoid stack is on the floor. First plasma is targeted for late 2026, with the Q greater than 1 demonstration — net energy out of the plasma — targeted for 2027.
The investors paying attention right now are not just energy-tech venture funds. They are the capital-allocation desks at every hyperscaler trying to back out a 2030 power bill.
The Hardware Is the Story
SPARC’s design choice — the one thing every physicist in the building will defend to the wall — is high-temperature superconductor (HTS) magnets built from yttrium barium copper oxide (YBCO). YBCO retains superconductivity up to 77 K, though SPARC operates closer to 10 K for margin. The field on axis is roughly 12 tesla. That field strength is what makes the machine compact. The Lawson criterion for ignition scales steeply with magnetic field, so doubling the field is not a 2x improvement to confinement quality — it is closer to a 16x improvement in the figure of merit that matters. Compactness is not a styling choice. It is the entire commercial thesis.
The first of the eighteen D-shaped toroidal field magnets was completed in January 2026. By April, more than half of the individual magnet “pancakes” — the layered superconducting tape stacks that make up each TF coil — were assembled. The May vessel install means the orange assembly stands now hold the vessel itself between the partially populated magnet sets. The order of assembly is precise: TF magnets onto stands, vessel into the magnet interior, poloidal field (PF) magnets looped around the structure, central solenoid down the throat, cryostat closed, vacuum drawn.
For a physicist this is the standard sequence. For a refinery executive used to reading construction percent-complete reports on a hydrocracker, the parallel is exact. SPARC has crossed from steelwork into instrumented assembly — the phase where schedule risk drops sharply and the dominant remaining variable is commissioning, not construction.
What the Modeling Says About 2027
SPARC’s mission objective is Q greater than 2. The official baseline says it should clear that with margin. The headline number CFS publishes is Q approximately 11 under empirical scaling laws, derived from the ITER89-P confinement multiplier applied to the SPARC operating point. That is the optimistic case.
The conservative case is more interesting. Physics-based integrated modeling — TRANSP coupled with the TGLF turbulence model and EPED predictions for pedestal stability — projects Q approximately 9 in standard H-mode operation. Recent CGYRO nonlinear gyrokinetic predictions for first-campaign plasmas land in the same neighborhood. Even the pessimistic version of the SPARC physics case clears the mission threshold by more than a factor of four.
Peak fusion power is designed at 140 MW for 10-second bursts. That is not commercial output — it is a physics demonstration. But 140 MW of thermal fusion power, even pulsed, is roughly thirty times the energy gain ratio NIF achieved in its 2022 ignition shot, in a machine that costs roughly the same as a single mid-size combined-cycle gas plant. The cost-per-demonstration curve is what the venture investors are tracking, and SPARC’s slope is steeper than anyone else’s right now.
The ARC Plant Is Already Sited
Before SPARC has even produced its first plasma, CFS has already sited its first commercial plant. The ARC Fall Line Fusion Power Station will be built at the James River Industrial Park in Chesterfield County, Virginia, on a site leased through a non-financial partnership with Dominion Energy Virginia. The plant is rated at approximately 400 MW of grid-delivered electric power — enough for roughly 150,000 homes — and is targeted to begin generating in the early 2030s.
In April 2026, CFS submitted the first fusion interconnection request in U.S. history to PJM Interconnection — the largest wholesale electricity market in North America. The application is a procedural milestone, but it is also a signal: CFS is treating fusion as a generation asset that needs to clear queue, study, and impact-analysis steps like any other generator. The framework of “we will figure out the grid integration later” is over. Fusion is now a transmission-planning problem.
Virginia’s selection over a dozen other states was not subtle. Chesterfield County sits inside the data-center belt that runs from Loudoun County south through Richmond. PJM is the grid that carries the AI workload for the East Coast hyperscaler footprint. The siting decision is a bet that the first ARC plant will sell its 400 MW to a buyer that already lives within a 100-mile radius.
Why the Hyperscalers Care Right Now
The same eighteen months that have moved SPARC from steel to assembly have rewritten the data-center power conversation. Microsoft is reactivating Three Mile Island Unit 1 under a 20-year PPA with Constellation Energy, anchoring the 837 MW Crane Clean Energy Center. Amazon Web Services has a 17-year PPA with Talen Energy for 1.92 GW from the Susquehanna nuclear plant. Google has signed the largest corporate SMR offtake in history with Kairos Power. Meta has a 1.1 GW PPA with Constellation for Clinton, Illinois.
Add them up and the hyperscalers have contracted more than 10 GW of new clean firm power in under two years. Microsoft alone is now sitting on 34.7 GW of clean energy contracted, surpassing Amazon as the largest corporate clean-energy buyer in the world. None of this existed at this scale in 2023.
What the procurement teams have figured out is that the AI compute curve is steeper than the grid build-out curve. Every GPU cluster they want to deploy in 2028 needs a megawatt that does not exist yet on the system. SMRs help but they will not arrive in volume until 2031–2033. Existing nuclear is a finite resource. Gas peakers run into emissions ceilings that the same hyperscalers have publicly committed to. Fusion is the only generation technology with both a credible long-duration carbon-free profile and a plausible commercial date inside the back-half of this decade.
This is why SPARC’s 75-percent milestone is not just a fusion-community story. It is a procurement-strategy story for every hyperscaler trying to firm up a power roadmap before the next AI training cycle.
The Competitive Landscape Is Honest
SPARC is not running unopposed. Helion Energy‘s Polaris reactor recently hit 150 million degrees Celsius in its plasma — a real number with real witnesses — and Helion has a binding offtake agreement with Microsoft for 50 MW of electricity starting in 2028. Helion’s magneto-inertial approach skips the steam cycle entirely and aims for direct electricity conversion. If it works at the Trenta-class scale they are targeting next, the economics will be brutal for everyone else.
Tokamak Energy in the UK has reached 100 million degrees ion temperature in a spherical tokamak architecture. TAE Technologies is on its Norman device with a p-B11 fuel path that, if proven, eliminates the neutron-flux maintenance burden entirely. Type One Energy is the stellarator entrant with the most credible commercial story. General Fusion is the magnetized target outsider. China’s state-owned consolidated fusion entity is now executing on a 2027-completion timeline for the BEST burning-plasma tokamak.
The honest read: nobody has won. But Commonwealth Fusion Systems is the entrant whose machine is closest to being a finished piece of hardware, whose grid interconnection process is the furthest along, and whose investor base — Google, Bill Gates, Nvidia, Eni, Temasek, Breakthrough Energy Ventures, Khosla Ventures — has the deepest combined willingness to fund the next two power-plant builds without waiting for the federal financing window to open. Roughly $3 billion of private capital has now flowed into CFS, most recently through an $863 million Series B2 in August 2025.
What This Means for the Industrial Sector
For refineries, petrochemical plants, and heavy-industrial operators reading this with skepticism — the relevance is not whether SPARC produces power before your next turnaround. The relevance is that the same DOE program staff who fund fusion are funding everything adjacent: superconducting magnet manufacturing, plasma diagnostics, tritium fuel-cycle research, AI-controlled magnetic-field optimization, and physics-informed digital twins. Every dollar that goes into fusion creates demand for the precision engineering, safety case writing, and compliance documentation that the rest of the energy sector also needs.
Porritt Inc. builds the software the energy-engineering economy uses to compile this kind of work — NEXUS CAD for the engineering geometry, NORMEX Standards AI for the regulatory citation layer. When SPARC ships, the procurement teams supporting the build-out — and the eight ARC plants behind it — will need the same compliance and engineering infrastructure every refinery turnaround already uses. That is a non-coincidence. Industrial decarbonization is the same job no matter what is generating the heat.
The next eighteen months are the most important stretch in commercial fusion history. By Q4 2026, SPARC should be producing first plasma. By Q4 2027, the world will know whether the Q greater than 1 milestone holds against the modeling. By 2029, the ARC site preparation in Chesterfield will be underway. By 2032, a 400 MW fusion power plant is delivering electrons to the PJM grid — or it is not.
If you are a hyperscaler procurement lead, a federal evaluator, a project finance desk pricing energy-infrastructure debt, or an industrial-operations executive trying to model your 2030 carbon liability — this is the milestone to mark on the calendar. The fusion conversation has moved from physics to commissioning. The next number to watch is not a temperature. It is a date stamp on a first-plasma press release.
Timothy Porritt is founder of Porritt Inc., building AI-powered tools for process safety, engineering compliance, and industrial operations. Based in Salt Lake City, Utah.