Compendium

Walkthrough: Design an ITER-Class Fusion Tokamak (and the Post-ITER Fusion Landscape)

26 min read

Walkthrough: Design an ITER-Class Fusion Tokamak (and the Post-ITER Fusion Landscape)

ITER (originally International Thermonuclear Experimental Reactor; now the official name in its own right) is the largest scientific instrument ever attempted — a 30 m tall, 30 m diameter superconducting tokamak under construction at Cadarache in southern France since 2010, designed to demonstrate Q = 10 (500 MW of fusion thermal output from 50 MW of injected heating power) for 400 second pulses with a 50:50 deuterium-tritium plasma. Its scale is what differentiates it from every previous tokamak — by volume it is roughly 8 to 10 times JET (Joint European Torus, Culham UK, 1983 to 2023), and the engineering it requires has never been done at the required scale before: niobium-tin superconducting magnets at 11.8 T peak field weighing 360 tonnes per coil, a 5,000 tonne stainless-steel vacuum vessel welded from 9 sectors made in Korea + EU + India, a tungsten divertor that must dissipate 20 MW/m² steady-state, a tritium plant breeding and reprocessing 5 mol/h of the rarest hydrogen isotope.

The original ITER cost estimate was €5 billion (1998 design); the project has repeatedly inflated to a current total estimated cost of 65 billion depending on accounting convention (the 35-country partnership counts in-kind contributions at the partner’s local cost basis, which makes a single dollar figure hard to nail down). First plasma was originally scheduled for 2025; the project formally rebaselined in 2024 with first plasma slipped to 2034 and full DT (deuterium-tritium) operation now targeted for 2039.

ITER is not a power plant. It will not generate a watt of electricity. Its mission is to demonstrate net-energy-gain plasma physics and to validate the engineering for DEMO — the next-step demonstration plant that is intended to generate electricity (EU DEMO target: 1 to 2 GW electrical net, operational 2050s). Meanwhile, an aggressive private fusion sector has emerged with very different timelines and approaches — Commonwealth Fusion Systems targets first plasma in 2027 and net electricity in the 2030s; Helion Energy claims a 50 MW PPA with Microsoft for 2028.

This walkthrough designs an ITER-class tokamak, working through the magnetic confinement physics, the engineering subsystems, the partner organization, and then surveys the broader post-ITER fusion landscape including SPARC, JT-60SA, EAST, KSTAR, K-DEMO, ARC, STEP, CFETR, Wendelstein 7-X, NIF inertial confinement, and the private startups.

1. The fusion problem — why is this so hard

Deuterium-tritium (DT) fusion has the lowest ignition threshold of any fusion reaction usable for power. The reaction is:

D + T → ⁴He (3.5 MeV) + n (14.1 MeV)

Total energy release per reaction: 17.6 MeV (4 to 5 times higher per nucleon than fission of U-235). The 3.5 MeV alpha (helium nucleus) stays in the plasma and heats it (the route to ignition); the 14.1 MeV neutron escapes and deposits its energy in the surrounding blanket structure, where it is converted to heat → steam → electricity (and where it is also used to breed more tritium from lithium).

To get the DT reaction rate high enough for net energy gain, the Lawson criterion must be satisfied. In its simplest fusion-triple-product form:

n × T × τE ≥ 3 × 10²¹ keV·s/m³ (for DT ignition)

where n is plasma density (m⁻³), T is plasma temperature (keV; 1 keV ~ 11.6 million K), and τE is energy confinement time (s — how long the plasma holds onto its heat).

ITER design point:

  • n ~ 10²⁰ m⁻³ (10⁻⁵ of atmospheric density — a very good vacuum of charged particles)
  • T ~ 15 keV (~ 150 million K — about 10 times hotter than the core of the Sun)
  • τE ~ 3.7 s
  • Triple product: ~ 5.5 × 10²¹ keV·s/m³

The Sun, by comparison, has central T ~ 1.5 keV but enormous density and τE (gravity holds it together for 10 billion years). Earthbound DT fusion has to compensate for far lower density × confinement-time with much higher temperature.

Confining 150 million K plasma without it touching any wall (which would instantly cool it and vaporize the wall) requires magnetic confinement. The plasma is a fully ionized gas of charged particles; they spiral along magnetic field lines and are confined transverse to them. The tokamak geometry uses a toroidal (doughnut-shaped) magnetic field generated by external coils, plus a poloidal (around-the-short-way) field generated by a current flowing in the plasma itself, to make helical field lines that confine particles.

2. ITER core specifications

  • Major radius R: 6.2 m (distance from torus axis to plasma center)
  • Minor radius a: 2.0 m (plasma column radius)
  • Aspect ratio R/a: 3.1 (conventional tokamak; spherical tokamaks like ST40 push to ~1.5)
  • Plasma volume: 840 m³
  • Plasma current Ip: 15 MA (compare JET 5 MA, SPARC 8.7 MA target)
  • Toroidal magnetic field on axis: 5.3 T (peak field at coil 11.8 T)
  • Fusion power: 500 MW thermal
  • Heating power injected: 50 MW
  • Gain Q: 10 (target; ITER must demonstrate Q ≥ 10 for 400 s pulses, with Q ≥ 5 for 1000 s pulse extension)
  • Pulse length: 400 s flat-top (inductive scenarios); up to 3000 s with non-inductive current drive
  • Plasma fuel consumption per pulse: ~0.5 g DT mixture (50:50 by atom)
  • Reactor envelope: 30 m tall × 30 m diameter (tokamak hall)
  • Total mass: ~23,000 tonnes of magnets alone; complete tokamak ~28,000 tonnes; total site facility much larger

ITER will not achieve “engineering Q” (Q including the energy to run the cryoplant, magnets, RF systems, vacuum pumps, etc.) of 1. The 50 MW heating power is the fusion-physics input; the wall-plug electrical input to ITER is several hundred MW. Net-electric-Q > 1 is the DEMO target, not ITER’s.

3. ITER partner organization — the multi-decade negotiated structure

ITER is governed by an international agreement signed November 2006 by seven parties representing roughly half of humanity. The agreement created the ITER Organization (IO) headquartered at Cadarache (Saint-Paul-lès-Durance, Bouches-du-Rhône, southern France), with 35 nations participating through the seven Members.

  • European Union: Domestic agency is Fusion for Energy (F4E), Barcelona. EU is the host party and bears ~45% of construction cost. Major components supplied: vacuum vessel sectors (5 of 9), poloidal field coils 2-5 (manufactured at Cadarache itself in purpose-built winding building), in-vessel components, divertor cassettes, neutral beam injectors, building construction.
  • United States: USIPO (US ITER Project Office, Oak Ridge TN, hosted by ORNL). ~9% in-kind plus cash. Major components: central solenoid (jointly with Japan; 6 modules manufactured at General Atomics Poway CA), 12 of 18 toroidal-field coil cases (manufactured Italy), Tokamak Cooling Water System, pellet injection, electron-cyclotron sources.
  • China: CNDA (China Domestic Agency, Beijing). Major components: AC/DC power converters, magnet supports, glow discharge cleaning, water-cooling, correction coils, internal magnetics diagnostics.
  • Russia: Russian Domestic Agency (RF DA, ROSATOM). Despite Russia’s invasion of Ukraine in 2022 the ITER project has continued Russian deliveries under the original treaty (a notable exception to broader sanctions; ITER components are technically civilian scientific equipment). Major components: poloidal field coil 1, port plugs, equatorial port stubs, gyrotrons, switching networks, blanket modules.
  • India: DAE (Department of Atomic Energy) Indian Domestic Agency. Major components: cryostat (largest stainless steel vacuum vessel ever fabricated, 30 m × 29 m, manufactured by Larsen & Toubro at Hazira Gujarat, shipped sectionally to France 2015 to 2020 — final lid lift August 2024), cooling water, cryogenic distribution, ion-cyclotron sources.
  • Japan: QST (National Institutes for Quantum Science and Technology). Major components: 9 of 19 toroidal-field coils (manufactured Mitsubishi Heavy Industries Kobe + Toshiba Yokohama), 4 of 6 central-solenoid modules (Mitsubishi for winding, General Atomics for module assembly — split work), high-voltage gyrotrons (170 GHz, ~1 MW each, 24 units total).
  • Korea: NFRI (National Fusion Research Institute, now Korea Institute of Fusion Energy KFE). Major components: 4 of 9 vacuum vessel sectors (manufactured Hyundai Heavy Industries Ulsan and Hyosung Changwon), thermal shields, tritium storage and delivery.

The in-kind contribution structure means that 90% of the project’s value is delivered by member-nation industries as components rather than as cash; the ITER Organization assembles and integrates these on site. This has been the source of considerable scheduling complexity: a single component delay anywhere in the seven-member supply chain delays the integrated assembly.

4. Magnets — the largest superconducting system ever built

The ITER magnet system has three families:

Toroidal field (TF) coils

  • 18 coils arranged like the petals of a flower around the vertical axis of the torus
  • D-shape, 18 m tall × 14 m wide, 360 tonnes each (after potting + casing)
  • Conductor: cable-in-conduit (CICC); ~900 strands of Nb₃Sn (niobium-tin) superconducting filaments in a copper matrix, encased in a stainless-steel jacket, with central helium-cooling channel
  • Operating current: 68 kA per coil
  • Operating temperature: 4.5 K (supercritical helium coolant flowing through the conduit and through cooling channels in the winding pack)
  • Peak field at conductor: 11.8 T (on the inside leg of each TF coil)
  • Field on plasma axis: 5.3 T
  • Manufacturer split: 9 coils Japan (Mitsubishi Heavy Industries + Toshiba); 10 coils EU (ASG Superconductors Italy + Iberdrola Spain + Alstom France + SIMIC Italy — note 10 EU + 9 JA = 19, one spare); 9 sets of D-shape stainless-steel coil cases (the structural housing around each winding pack) Korea (Hyundai HEC); plus winding-pack inserts and joints
  • Nb₃Sn strand supply was a major early bottleneck: total strand requirement ~600 tonnes; world production capacity in 2005 was ~15 tonnes/year, principally for MRI manufacturers. ITER drove a 10× expansion in global Nb₃Sn production capacity from suppliers including Bruker EAS (Germany), Furukawa (Japan), Hitachi (Japan), Western Superconducting Technologies (China), Oxford Instruments (UK), Luvata (Finland), KAT (Korea), Bochvar (Russia), Hitachi Cable (Japan); the first ITER-grade strand was qualified 2009.

Central solenoid (CS)

  • 6 stacked modules in the bore of the tokamak (the vertical column running through the middle of the torus)
  • Total height ~18 m, diameter 4 m, mass ~1000 tonnes
  • Conductor: Nb₃Sn cable-in-conduit
  • Operating current: 45 kA
  • Peak field: 13 T
  • Operating temperature: 4.5 K
  • Function: provides the changing flux that inductively drives 15 MA of plasma current via the transformer principle (plasma is the secondary winding)
  • Manufacturer: Modules wound and assembled at General Atomics (Poway California) under contract to USIPO; Nb₃Sn conductor by Mitsubishi (Japan) via QST; testing at GA before shipment to Cadarache
  • Each module weighs ~110 tonnes; first module delivered to ITER site September 2020 by Mighty Servant 1 heavy-lift ship through Berre-l’Étang port

Poloidal field (PF) coils

  • 6 large horizontal ring coils (PF1 through PF6) arranged at varying heights above and below the plasma
  • Function: shape and stabilize the plasma — controlling its position, elongation (vertically stretched plasma — the ITER “D” cross-section is ~1.8 elongation), triangularity, and divertor geometry
  • Conductor: NbTi (niobium-titanium, the “ordinary” superconductor as compared to Nb₃Sn for high-field) cable-in-conduit
  • Operating current: 45 to 55 kA depending on coil
  • Field: 4 to 6 T
  • Operating temperature: 4.5 K
  • PF1 (top, smallest): supplied by Russia, manufactured Sredne-Nevsky Shipyard St Petersburg
  • PF2 through PF5 (middle, largest — up to 25 m diameter, 400 tonnes): manufactured on the ITER site itself in a purpose-built winding building, because they are too large to ship; F4E (EU)
  • PF6 (bottom): manufactured ASIPP Hefei China; arrived 2020

Correction coils

  • 18 smaller coils (CC) to correct field-error harmonics from manufacturing tolerances in the larger coils
  • Nb-Ti CICC; manufactured ASIPP China (the same Institute of Plasma Physics that operates EAST)

Cryogenics

  • Largest helium cryoplant ever built; 75 kW of refrigeration at 4.5 K (compared to LHC at CERN which uses 4 × 18 kW for the LHC ring, so ITER is roughly equivalent to the entire LHC in helium-cryogenic capacity)
  • Liquid helium inventory: ~25 tonnes
  • Helium supplied by Linde + Air Liquide; cryoplant integrated by Air Liquide France with helium turbines, cold-boxes, screw compressors, and 80 K thermal shield circuit (gaseous helium at 80 K for the thermal-shield buffer between the 300 K cryostat exterior and the 4.5 K coil cold mass)

5. Vacuum vessel — the plasma chamber

  • Double-wall stainless steel 316LN (low-carbon, nitrogen-enhanced) torus
  • Inner diameter ~13 m; outer diameter ~19.4 m at equator; vertical extent ~11 m
  • Total mass: ~5,200 tonnes (one of the heaviest welded stainless-steel structures ever fabricated)
  • 9 sectors of 40° each, welded together on site
  • Sector supply: 5 EU (Mangiarotti / Walter Tosto / AMW Consortium Italy), 4 Korea (Hyundai Heavy Industries Ulsan + Hyosung Changwon)
  • Double wall serves as: (a) primary tritium boundary (one of three nested tritium barriers — first wall + vacuum-vessel + cryostat), (b) cooling water flow path between walls for nuclear-heat removal, (c) shielding (water + steel mass between plasma and superconducting coils to keep neutron fluence on superconductors below damage threshold)
  • 44 ports through the vacuum vessel walls (18 upper, 17 equatorial, 9 lower) for diagnostics, heating systems, test blanket modules, remote handling access, vacuum pumping
  • All welds undergo full radiographic + helium leak-rate testing (leak rate must be < 10⁻⁸ Pa·m³/s)

6. Plasma-facing components — divertor and blanket

The plasma touches solid structure only at controlled locations. The geometry of the ITER plasma is “lower single null” — a specific magnetic configuration where field lines from the outer plasma layer (the “scrape-off layer” SOL) all funnel down to a small ring at the bottom of the vacuum vessel, the divertor.

Divertor

  • 54 modular cassettes lining the bottom of the vacuum vessel, forming a 360° ring
  • Each cassette: ~10 tonnes, 3.4 m long, removable as a unit using the remote-handling system
  • Plasma-facing material: tungsten monoblock armor on copper-alloy (CuCrZr) heat-sink, water-cooled
  • Heat flux during steady-state operation: 10 MW/m² (the divertor strike point) — comparable to the surface of the Sun
  • Transient heat flux during ELMs (edge-localized modes — short-duration plasma instabilities): up to 20 MW/m² for milliseconds
  • Cooling: pressurized water at ~3 MPa, inlet ~70°C, outlet ~120°C (subcooled — kept liquid, no nucleate-boiling crisis allowed)
  • Supplier consortium: cassette bodies — Iberdrola + Mangiarotti EU; inner-vertical-target plasma-facing units — Plansee Austria; outer-vertical-target PFUs — ANSALDO Italy + RHP Austria; dome — JADA Japan; cassette assembly — Ferrari Tosto / RINA Italy
  • The mission to make tungsten cope with 10 MW/m² steady plasma flux is one of the central engineering challenges of fusion power — the JET legacy program included tungsten divertor testing 2011 to 2023 (the JET-ITER-Like-Wall campaign) which provided the validation data ITER’s divertor is based on

Blanket

  • 440 modules covering the inside surface of the vacuum vessel above and around the divertor
  • Each module: ~4.5 tonnes, includes shield block + first-wall panel (beryllium-coated stainless steel)
  • Function: (a) absorb the 14.1 MeV neutrons from the plasma and slow them down, depositing their kinetic energy as heat in the structure; (b) shield the superconducting coils from neutron damage; (c) host the tritium breeding test modules (TBMs) in 4 of the equatorial ports
  • Cooling: water at 4 MPa, inlet 70°C, outlet 240°C (close to PWR temperatures)
  • ITER’s reference blanket does not breed tritium for the plasma — instead it tests competing breeding concepts in the TBM ports:
    • HCPB (Helium-Cooled Pebble-Bed) — EU; lithium-ceramic pebble bed (Li₂TiO₃ + Li₄SiO₄) with beryllium neutron multiplier, helium-cooled
    • HCCR (Helium-Cooled Ceramic Reflector) — Korea; lithium-ceramic
    • WCCB (Water-Cooled Ceramic Breeder) — Japan; lithium-ceramic, water-cooled
    • LLCB (Lead-Lithium Cooled Ceramic Breeder) — India; PbLi liquid metal
    • DCLL (Dual-Coolant Lithium-Lead) — US; PbLi + helium
    • HCCB (Helium-Cooled Ceramic Breeder) — China; lithium-ceramic
  • The TBM program is the central technology test ITER provides for DEMO — it shows which breeder blanket concept(s) work at the relevant 14 MeV neutron spectrum.

7. Heating systems — 73 MW external heating to reach Q = 10

The plasma must be heated externally to ~15 keV before fusion alpha-heating becomes self-sustaining. ITER provides 50 MW for routine operations, with 73 MW total installed (upgrade path to 110 MW reserved). Three concurrent heating methods:

Neutral beam injection (NBI)

  • 2 NBI lines, each 16.7 MW into the plasma, total 33 MW (upgrade port reserved for third NBI line, +16.7 MW)
  • Beam: D⁰ (neutral deuterium) atoms at 1 MeV kinetic energy (relativistic; classical particles at this energy would penetrate ~2 m of plasma before slowing)
  • Generation: D⁻ negative ions (because at 1 MeV, positive ions cannot be efficiently neutralized — neutralization cross-section falls off with energy; negative ions can be photodetached or stripped to neutral D⁰ with ~60% efficiency)
  • Source: caesium-conditioned hydrogen plasma; D⁻ extracted at -1 MV through 5-stage electrostatic accelerator
  • Test bed: PRIMA (Padova Research on ITER Megavolt Accelerator), at Consorzio RFX, Padova Italy — a full-scale prototype NBI built to validate the ITER design before ITER itself starts. SPIDER (source-only test bed) operational since 2018; MITICA (full-energy test bed) ramping up.
  • Supplier: F4E (EU) + JADA (Japan) shared procurement; major industrial partners Elytt Energy (Spain), Galileo Vacuum (Italy), Tecnimont (Italy)

Ion cyclotron resonance heating (ICRH)

  • 20 MW from 2 antennas
  • Frequency 40 to 55 MHz (tunable to match ion cyclotron resonance for D, T, ³He, H minority species)
  • Generation: tetrode tubes (similar technology to broadcast transmitters, but at higher power)
  • Suppliers: F4E (EU) + INDA (India); antennas manufactured ENEA Frascati Italy and IPR Gandhinagar India

Electron cyclotron resonance heating (ECRH)

  • 24 MW from 24 gyrotrons at 170 GHz
  • Frequency tuned to electron cyclotron resonance at 5.3 T toroidal field → 170 GHz for second-harmonic X-mode at plasma core
  • Suppliers: 8 gyrotrons Russia (GYCOM/INP Nizhny Novgorod), 8 Japan (QST + TOSHIBA + Mitsubishi), 6 EU (Thales France + KIT Karlsruhe), 2 India (CEERI Pilani)
  • Beam steering: quasi-optical transmission via mitre-bend mirror system + final-stage steerable launchers; allows real-time MHD-mode suppression by depositing power on resonant magnetic surfaces

8. Tritium plant

Tritium is the rarest hydrogen isotope on Earth — total world inventory ~30 kg (mostly from CANDU heavy-water reactor moderator overheads — Ontario Power Generation’s Darlington Tritium Removal Facility is the principal supplier). Half-life 12.3 years; decay-loss ~5%/year. ITER’s site tritium inventory limit is set at 4 kg total (700 g in-vessel + balance in tritium plant storage).

The ITER tritium plant:

  • Throughput: 5 mol/h of DT mixture flowing through the tokamak fueling-recycle loop
  • Cryogenic distillation: H/D/T isotope separation column at 20 to 25 K (boiling-point separation: H₂ 20.4 K, HD 22.1 K, D₂ 23.6 K, HT 22.9 K, DT 24.4 K, T₂ 25.0 K). Tritium-rich product stream is the fuel; protium tail is exhausted.
  • Storage: getter beds (depleted uranium U-238 or ZrCo intermetallic) at room temperature absorb DT chemically; release on heating to 350 to 500°C
  • Fueling:
    • Gas puffing for edge fueling (low penetration)
    • Pellet injection for core fueling (5 mm cryogenic DT pellets at 500 m/s by gas-gun, 5-10 Hz)
    • Supplied F4E (gas) + USIPO (pellet)
  • Detritiation: residual room air, glovebox atmospheres, water effluent all pass through catalyst beds + getter beds to capture stray tritium before release. Stack release limit < 1.5 g T₂ per year.

The tritium plant is the technology subsystem most directly transferable to DEMO + commercial fusion — every fusion power plant will need essentially this system.

9. Diagnostics

ITER has ~50 diagnostic systems measuring plasma conditions; each system has its own port allocation, vacuum interface, and remote-handling-compatible installation. Examples:

  • Thomson scattering (multiple): laser scatters off plasma electrons; spectrum reveals Te (electron temperature) and ne (electron density) at multiple radial locations. Nd:YAG and ruby laser systems.
  • Neutron camera + neutron spectrometer: measures the 14.1 MeV fusion neutron spectrum, yielding total fusion power + emissivity-profile information (where in the plasma fusion is happening)
  • Bolometers: measure radiated power (synchrotron, bremsstrahlung, line radiation) for power balance
  • Reflectometry + interferometry: microwave-based density-profile measurements
  • Charge-exchange recombination spectroscopy (CXRS): ion temperature, plasma rotation, impurity concentration
  • Erosion + deposition monitors: track tungsten erosion + deuterium retention in PFCs
  • Bolometer arrays, magnetic diagnostic coils, Mirnov coils: equilibrium reconstruction + MHD-mode identification

10. Remote handling

ITER plasma activates the in-vessel structure with 14 MeV neutrons; in DT operation the in-vessel components will be too radioactive for hands-on maintenance (gamma dose rates of order Sv/h immediately after shutdown). All in-vessel maintenance must be remote-handled.

  • Cask + transporter system: airlock + transfer-cask docks at vacuum-vessel ports; modules are extracted to cask, transported to Hot Cell Facility on site for refurbishment/disposal
  • In-vessel Viewing System (IVVS): laser-scanning + camera-based inspection of in-vessel surfaces between pulses
  • Multi-Purpose Deployer + Divertor Cassette Mover: large in-vessel robots that exchange divertor cassettes, blanket modules, and diagnostic plugs
  • Suppliers: CETMA Italy (cask + transporter), Oxford Technologies UK (manipulators), Hyundai Heavy Industries (heavy-lift), Tractebel ENGIE (controls + integration)

11. The post-ITER fusion landscape

While ITER builds, the rest of the fusion world has moved.

Operating large tokamaks (the comparison and supplementary machines)

  • JET (Joint European Torus), Culham UK: 1983 to 2023 final shutdown. The previous record-holder for fusion energy in a single pulse — 69 MJ over 5 seconds in its 2021 DT-2 campaign with ITER-like-wall (beryllium + tungsten). Provided most of the validation data for ITER. Now in decommissioning.
  • JT-60SA, Naka Japan: largest operating tokamak in the world post-JET-shutdown. Joint EU-Japan project (Broader Approach Agreement). NbTi superconducting magnets (lower field than ITER), 5.5 m major radius, 1.18 MA plasma current first plasma October 2023. Designed as the bridge machine between current generation and ITER, providing ITER-relevant scenario development at smaller scale.
  • EAST (Experimental Advanced Superconducting Tokamak), Hefei China: operational since 2006, fully superconducting. Held record for longest plasma at 1056 seconds (~17 min) in April 2023 at ~70 million K.
  • KSTAR (Korea Superconducting Tokamak Advanced Research): at NFRI Daejeon Korea. NbTi/Nb₃Sn fully superconducting. Held 100-million-K record for 30 seconds in 2023 (subsequently exceeded by EAST).
  • DIII-D, San Diego CA: US, General Atomics-operated, copper-coil (not superconducting), runs short pulses; primary US tokamak for advanced scenario development.
  • TCV (Tokamak à Configuration Variable), Lausanne: at EPFL Switzerland, copper-coil; very flexible plasma shaping for control-physics studies.
  • Alcator C-Mod, MIT (1992-2016): pioneering high-field LWR design heritage that became the SPARC + ARC concept

Stellarators

Different geometry from tokamaks — the helical field comes entirely from external coils (no plasma current), so the plasma is steady-state by construction (no inductive flux-swing limitation, no current-driven instabilities). Trade-off: extremely complex 3D coil geometry.

  • Wendelstein 7-X (W7-X), Greifswald Germany: operational since 2015. 5.5 m major radius. NbTi superconducting modular + planar coil set. Set record February 2023 for combined plasma duration × temperature × density (the stellarator triple product proxy) by sustaining 8 minutes of high-performance plasma. The flagship stellarator demonstration.
  • Helically Symmetric eXperiment (HSX), Wisconsin: smaller stellarator with quasi-helical symmetry
  • NCSX (cancelled), PPPL: cancelled in 2008 due to cost overrun in coil manufacturing
  • Type One Energy: US startup commercializing quasi-symmetric stellarator (founded 2023 by ex-CFS engineers)
  • Realta Fusion (Madison WI): mirror + stellarator hybrid concepts; Wisconsin spinout

China — CFETR + CRAFT

  • CFETR (China Fusion Engineering Test Reactor): design phase; intended as a DEMO-equivalent successor to ITER and EAST, ~200 MWe net target, post-2035 construction
  • CRAFT (Comprehensive Research Facility for Fusion Technologies): testing facility under construction Hefei

Korea — K-DEMO

  • ITER-class DT machine; design phase; targeted for first plasma ~2037-2040; intended as a fusion electricity demonstrator parallel to EU DEMO

Private fusion (the disruption)

Roughly $7 billion in private fusion funding has been deployed 2020 to 2026, more than the cumulative private investment in fusion in the entire previous 50 years.

  • Commonwealth Fusion Systems (CFS), Devens MA: MIT-spinout, founded 2018. Building SPARC — a compact high-field DT tokamak using REBCO (rare-earth barium copper oxide) high-temperature-superconductor (HTS) magnets at 20 T peak field (vs ITER’s 11.8 T Nb₃Sn). Compact = R = 1.85 m (vs ITER R = 6.2 m) with same physics × scaling. First plasma target 2027. Series B 2024 raised 3 billion. Follow-on ARC (Affordable, Robust, Compact) commercial pilot plant ~400 MWe targeted for 2030s, site Chesterfield County Virginia (Dominion Energy host partner) announced 2023.
  • Helion Energy, Everett WA: field-reversed-configuration (FRC) pulsed device; D + ³He fuel (aneutronic, sort of — but Helion plans direct-electric-conversion via pulsed magnetic compression). Microsoft 50 MW Power Purchase Agreement 2023 (delivery target 2028 — almost certainly to slip). Sam Altman ($375M personally invested), Peter Thiel, others. Prototype Polaris first plasma 2024.
  • TAE Technologies, Foothill Ranch CA: FRC + neutral-beam plasma stabilization. ~$1.2 billion raised cumulative. “Norman” device operational, “Copernicus” device under construction targeting D-³He plasma.
  • Tokamak Energy, Milton Park UK: spherical tokamak (low aspect ratio, R/a ~ 1.5) with HTS magnets. ST40 device achieved 100 million K in March 2022 — first private machine to do so. Next-generation ST80-HTS planned.
  • First Light Fusion, Yarnton UK: projectile-driven inertial confinement (hyper-velocity projectile compresses target). $90M+ raised. Pyxis-class machines.
  • Marvel Fusion, Munich: ultrafast laser-driven inertial fusion using p + B¹¹ (boron) aneutronic target
  • Focused Energy, Darmstadt Germany / Austin TX: laser-driven inertial; founded by ex-NIF physicists; partnership with Forschungszentrum Jülich
  • Zap Energy, Everett WA: sheared-flow-stabilized Z-pinch; pulsed; FuZE devices
  • General Fusion, Vancouver BC: magnetized target fusion; piston-driven lead-lithium liquid-metal compression of magnetized plasma. Site at Culham UK for demonstration plant (announced 2021; delayed).
  • Xcimer Energy, Denver: KrF-laser-driven inertial fusion; founded 2022 by NIF veterans
  • Pacific Fusion, Bay Area: stealth-mode pulsed-power magnetic-target startup; raised $900M Series A Oct 2024 — one of largest fusion fundings ever
  • Avalanche Energy, Seattle: orbitron-based small-fusion concept; not yet at scientific breakeven scale
  • Stellar Fusion + Realta Fusion + Thea Energy + Princeton Stellarators: stellarator-pathway startups

Inertial confinement — the NIF achievement

The National Ignition Facility (NIF) at Lawrence Livermore National Laboratory (Livermore CA) is the world’s largest laser — 192 beam lines totaling 1.8 MJ of UV light energy delivered in ~10 ns pulse to a 2 mm hohlraum containing a DT-filled cryogenic capsule. NIF is mission-funded by NNSA for nuclear weapons stockpile stewardship; fusion-energy research is a secondary application.

  • December 5, 2022 ignition shot: 2.05 MJ laser energy delivered to target → 3.15 MJ fusion energy released. Fusion fuel gain Q_fuel = 1.5 (>1 for first time in any laboratory) — the first “scientific ignition” in human history.
  • Subsequent shots in 2023: Q_fuel reached 1.9 (5.2 MJ out from 2.2 MJ in) in July 2023; further increases through 2024.
  • Engineering-Q context: the 192 NIF lasers consume ~422 MJ of wall-plug electrical energy to deliver 2 MJ to target — engineering gain is far below 1. NIF was never designed to be an energy-positive system; it was designed to demonstrate ignition physics. But ignition was demonstrated, and that is the gate that ICF-as-energy must pass through before commercial design can begin.

12. Materials challenges — the long-pole problem

The deuterium-tritium reaction produces 14.1 MeV neutrons. These do enormous damage to surrounding structural material:

  • Displacement damage: each 14 MeV neutron creates ~10 to 100 atomic displacements in the lattice. DEMO first-wall materials must withstand 30 to 150 dpa (displacements per atom) over plant lifetime. ITER reaches only ~3 dpa peak in 20 years of operation, so ITER does not fully qualify DEMO materials — a parallel program is needed.
  • Helium embrittlement: 14 MeV neutron + nucleus → α (helium) + lighter nucleus; helium accumulates in grain boundaries causing brittle failure
  • Transmutation: neutron capture transmutes elements (Cr → V → Ti chain in ferritic steel; chemistry of the wall shifts over lifetime)
  • Tritium permeation: tritium diffuses through metals at fusion-temperature; must be contained with tritium-permeation barriers (aluminide coatings, oxide layers)

Structural material candidates:

  • EUROFER-97: reduced-activation ferritic-martensitic (RAFM) steel, Fe-Cr-W-V-Ta; the EU baseline. Activation products decay to free-release levels in ~100 years (vs >10,000 years for conventional stainless steels) due to removal of Nb + Mo + Ni + Cu.
  • F82H: Japanese RAFM steel, similar Fe-Cr-W base
  • ODS steels (oxide-dispersion-strengthened, e.g., 14YWT, MA957): Y₂O₃ nano-precipitates pin dislocations, raise creep strength, suppress He bubble growth
  • V-4Cr-4Ti: vanadium alloy; low activation, high temperature, but very oxygen-sensitive (lithium coolant compatible, not water)
  • SiC/SiC composites: silicon-carbide fiber in silicon-carbide matrix; ultra-high temperature (>1000°C), very low activation, but ductility + permeability + manufacturing are all open issues

IFMIF-DONES (International Fusion Materials Irradiation Facility — DEMO-Oriented Neutron Source) is under construction at Granada Spain (first stones 2024, operational ~2030). It will produce a 14 MeV neutron flux to qualify materials at DEMO-relevant dpa rates within a 5-10 year campaign. Without IFMIF-DONES (or equivalent), DEMO materials qualification is the long-pole limit on fusion deployment timeline.

13. Economics + outlook

The 25 to 65 billion USD ITER cost is not representative of future fusion plant cost — it includes one-time R&D, six site civil constructions in seven member countries, manufacturing first-of-a-kind learning, and integration complexity from in-kind contributions. The relevant figure for commercial fusion is the LCOE (levelized cost of energy) of a serial-build plant.

LCOE projections from independent analysis (Lazard 2024, IEA 2024, MIT Energy Initiative 2023):

  • Optimistic (assuming HTS magnets + compact designs deliver promised cost reduction): 100 / MWh (competitive with Gen III+ nuclear; cheaper than firm renewables + storage)
  • Central: 200 / MWh (premium to Gen III+ nuclear but valuable for firm + dispatchable power)
  • Pessimistic (FOAK + commercial-cost-overruns): $300+/MWh (uncompetitive)

Timeline reality-check (as of 2026):

  • ITER first plasma 2034, DT operation 2039 → confirmed
  • SPARC first plasma 2027 → on schedule per CFS; net-energy 2028
  • ARC commercial pilot 2030s → speculative; depends entirely on SPARC results
  • Helion 2028 50 MW commercial → almost certain to slip
  • EU DEMO 2050s → conceptual; EUROfusion design phase
  • Commercial fleet deployment 2040s + → most realistic scenario in IEA NZE

The wave of corporate offtakes (Microsoft + Helion, Google + Kairos for fission, Microsoft + Constellation for Three Mile Island restart, etc.) reflects the data-center industry’s willingness to pay a premium for low-carbon firm baseload, which has substantially shifted the commercial calculus for advanced nuclear including fusion.

Whether fusion enters the energy mix at scale by 2050, 2060, or 2080 — and whether the dominant technology is tokamak, stellarator, FRC, ICF, or something not yet identified — is genuinely uncertain. But ITER first plasma in 2034 + SPARC in 2027 will, between them, dramatically change the empirical basis for those projections.

14. Adjacent

Graph View

Backlinks