Brennan Superconductors
Wrexham
REBCO tape and non-insulated pancake winding: 738 km of conductor for the TF set, on a 14-week drumbeat.
TARN-1 · Shot 41,208 · 19 Aug 2026 · 6.4 s burn
Confinement, not containment.
We build tokamaks in Oxfordshire. TARN-1 has been running deuterium plasmas since September 2024. TARN-2 is the machine that has to sell electricity — 315 MW of fusion power, 40 MW net onto the grid, first plasma 2033.
0 M°C
Peak ion temperature, TARN-1
0
Plasma shots since first light
£0 m
Series C, closed March 2026
0 T
Peak field on the REBCO conductor
Sheet 02 Record to date Rev C · 19.08.26
Eleven years,
two machines,
no press releases
about breakthroughs.
Tarn was spun out of a magnet group in 2015 with one bet: that rare-earth barium copper oxide tape would let a tokamak get small enough to build twice. TARN-0 proved the coil. TARN-1 proved the plasma. TARN-2 has to prove the balance sheet, which is a harder problem than either.
0
Shots on TARN-1
Every one archived, calibrated and open to our science partners within 48 hours.
0s
Longest flat-top
Held at 8.1 MA on 14 June 2026. The limit is the divertor, not the magnets.
0
People, 74% engineers
Abingdon, Didcot and a coil-winding hall in Wrexham. Nobody works remotely on a magnet.
0%
UK content by value
Forty-one British suppliers. The tungsten, the vacuum vessel and the tape are all made here.
Backed by
Fenchurch Deep Capital
Office for Fusion Deployment
Loughrigg Ventures
Bramhope Industrial
Ceres Climate Fund II
Sheet 03 TARN-2 shot console · illustrative model Rev C · 19.08.26
Set the plasma.
The machine will
tell you if it holds.
The same five-parameter envelope our physics team argues about on Tuesdays, reduced to something you can drag. Move a slider and the boundary, the flux surfaces and the colour of the plasma redraw, then the panel re-solves for temperature, confinement time, triple product, Q and the megawatts that would reach a substation. Break a real limit and the console refuses the shot, and says which one.
01 Shot request
TARN-2 · SIM
11.0MA
3.0Ohmic + non-inductive drive14.0
9.8T
4.0Peak on conductor 19.1 T13.0
2.00
1.30Vertical stability control2.40
0.94n/nGW
0.35ne = 2.49×1020 m−31.20
38MW
15NBI + electron cyclotron120
Fuel mixture
50/50 deuterium–tritium. 17.59 MeV per reaction, 20% of it carried by the alpha particle that stays behind and heats the plasma.
Load a set-point
02 Section A–A · poloidal cross-section
Solution converged
-
Fusion gain Q
8.28
Alphas carry 20% of it back into the plasma
-
Fusion power
315MW
80% in neutrons
-
Energy confinement τE
2.14s
IPB98(y,2) × H = 1.15
-
Net electrical
+40MW
After all recirculating loads
03 Edge turbulence · mid-plane view
χeff nominal
04 Achieved Q · shot history
17 shots
05 Operational limits
Safety factor q954.95 / min 2.00
Greenwald fraction0.94 / max 1.00
Normalised beta βN1.50 / max 3.50
Peak conductor field19.1 / max 23.0 T
Divertor loading Psep/R20 / max 30 MW/m
H-mode threshold margin1.27× / min 1.00×
06 Interlock log
Every entry names the physics that produced it. An interlock in red means the pulse would not be permitted on a real machine, and the reason is not a matter of taste.
07 What this model does and does not do
A deliberately simplified zero-dimensional power balance, published so you can see the shape of the problem rather than a marketing number. Honest about its own arithmetic and useless as a design tool.
- Reactivity ⟨σv⟩ from the Bosch–Hale parameterisation for D–T, both D–D branches and D–³He, valid 0.2–100 keV.
- Confinement from the IPB98(y,2) ELMy H-mode scaling, with a house H-factor of 1.15 claimed for the high-field, lithium-conditioned wall.
- Temperature is solved as a fixed point: alpha heating feeds the power balance, which sets τE, which sets stored energy, which sets temperature. Damped iteration, 200 passes.
- Flat density and temperature profiles, one ion species pair, Zeff fixed per fuel, bremsstrahlung only — no synchrotron, line radiation, impurity transport, sawteeth, ELMs or pedestal physics.
- Balance of plant: 42% thermal-to-electric, 1.14 blanket energy multiplication, 45% injector wall-plug efficiency, 38 MW fixed load for cryoplant, coils, pumps and the tritium plant.
- Limits applied: q95 ≥ 2.0, n/nGW ≤ 1.0, βN ≤ 3.5, conductor field ≤ 23 T, Psep/R ≤ 30 MW/m, and the Martin 2008 L–H threshold reduced by 45% for the lithium-conditioned wall.
1 Toroidal field coil
Eighteen D-shaped REBCO pancakes, 24.9 T peak on conductor, 20 K inlet.
2 Vacuum vessel
316LN double wall, 62 mm, borated water between the skins.
3 First wall
Tungsten monoblock on CuCrZr, lithium conditioned between campaigns.
4 Poloidal field coils
Six shaping coils plus a divertor coil; they set κ and hold the X-point.
5 Plasma boundary
The last closed flux surface. Everything outside it is being thrown away.
6 Divertor cassette
Fifty-four removable cassettes, exchanged by remote handling in 26 days.
7 Central solenoid
Six modules, 42 Wb of flux swing, the thing that starts the current.
Sheet 04 TARN-2 subsystems Rev C · 19.08.26
Four hard problems
and how far along
each one is.
A tokamak is not one invention. It is a magnet problem, a materials problem, an exhaust problem and a fuel-cycle problem sharing a building and a budget. Where each stands in August 2026, with what is built kept separate from what is still a drawing.
Eighteen coils, one tape, no liquid helium
The whole company exists because REBCO tape carries current at 20 K that niobium-tin cannot carry at 4 K. That single fact lets the machine shrink from a 6.2 m major radius to 3.6 m, and a tokamak that is half the size is roughly an eighth of the cost to build and a great deal quicker to rebuild when you learn something.
The coils are wound as non-insulated pancakes in Wrexham, 41 km of tape each. TF-04 finished its 12,000-cycle fatigue campaign in April 2026 with no measurable loss of critical current. Cooling is helium gas at 20 K, so there is no cryogenic liquid inventory in the building.
| Conductor | REBCO, 12 mm | Brennan Superconductors, Wrexham |
|---|---|---|
| TF coils | 18 | Non-insulated D-section pancakes |
| Peak field on conductor | 24.9 T | Demonstrated on TF-04; qualified to 23.0 T |
| Field on axis | 9.8 T | At R₀ = 3.60 m, design point |
| Stored magnetic energy | 4.1 GJ | Dumped in 11 s on a quench |
| Operating temperature | 20 K | Gaseous helium, no liquid inventory |
| Status | TRL 6 | Full-scale coil tested to spec, April 2026 |
Tungsten, and the neutrons that ruin it
A D–T plasma throws 14.1 MeV neutrons at the wall at roughly 1 MW per square metre. They do not care about magnetic fields. They knock atoms out of the lattice, leave helium behind, and over a few full-power years turn good structural steel into something that fails a bend test.
TARN-2 plans a first-wall replacement every 2.4 full-power years from the outset rather than discovering it later. All 1,120 tungsten-on-CuCrZr panels are exchanged by the same remote handling arm that services the divertor.
| Armour | W, 6 mm | Monoblock on CuCrZr heat sink |
|---|---|---|
| Panels | 1,120 | Halstead Ceramics, Stoke-on-Trent |
| Neutron wall load | 0.97 MW/m² | At the 315 MW design point |
| Displacement damage | 9.6 dpa | Per replacement interval |
| Replacement interval | 2.4 fpy | Full-power years, scheduled |
| Exchange outage | 71 days | Whole-wall, remote handling |
| Status | TRL 5 | Panels irradiated at Culham, 2025–26 |
The exhaust problem, which is the real one
Everything the plasma loses arrives somewhere. In a conventional divertor it arrives on a strike point a few millimetres wide, at a heat flux that would cut through a steel plate. Nothing survives 40 MW per square metre, so the answer is never a better material; it is to stop the power arriving concentrated.
TARN-2 runs a nitrogen-seeded, partially detached double null: the exhaust radiates most of its energy into the divertor volume before it reaches a surface. It works on TARN-1 for the full 6.4 s flat-top. Holding it for eight hours with tritium in the machine is a different question, and we do not pretend otherwise.
| Configuration | Double null | Nitrogen-seeded, partially detached |
|---|---|---|
| Cassettes | 54 | Ilkley Tungsten, Leeds |
| Psep / R | 20 MW/m | Design point; 30 MW/m is the ceiling |
| Peak target flux | 8.1 MW/m² | Attached; under 2 when detached |
| Coolant | H₂O, 155 °C | 4.2 MPa, 1,900 kg/s |
| Cassette exchange | 26 days | Full set, remote handling |
| Status | TRL 4 | Detachment demonstrated on TARN-1 |
Twenty kilogrammes of tritium exist
Global civil tritium stocks are somewhere around 20 kg and falling as CANDU reactors retire. TARN-2 burns 48 g a day at full power. There is no version of this industry that buys its fuel; a power plant must breed more tritium than it burns or it is a very expensive science experiment.
Our blanket is lithium-lead eutectic with a beryllium multiplier: breeding ratio 1.11 against a break-even of 1.05 once decay, holdup and wall retention are counted. Everything downstream is sized to hold under 400 g of in-process inventory, which is what keeps the site out of the top regulatory tier.
| Breeder | PbLi + Be | Eutectic, 90% ⁶Li enriched |
|---|---|---|
| Breeding ratio | 1.11 | Break-even is 1.05 with holdup |
| Tritium burn | 48 g/day | At the 315 MW design point |
| Burn fraction | 2.9% | The rest is recycled each pass |
| In-process inventory | < 400 g | Site licensing threshold |
| Startup inventory | 3.1 kg | Contracted, Skerne Power Systems |
| Status | TRL 3 | Blanket mock-up under test at Harwell |
Sheet 05 Fusion is not fission Rev B · 02.07.26
The two are
opposites, and
the difference
is the safety case.
Fission splits heavy nuclei that want to split, and the hard engineering problem is stopping the reaction. Fusion forces light nuclei together against their will, and the hard engineering problem is keeping it going for one more second. That inversion is why a tokamak cannot melt down: take away the heating and the reaction stops in under a second, because it was never self-sustaining without you. Switch the diagram to compare the two.
Reaction to display
Deuterium and tritium fuse to helium-4 and a 14.1 MeV neutron. Nothing left behind is fissile and nothing decays for centuries.
Property
Fusion — D–T tokamak
Fission — PWR, U-235
Fuel in the vessel
About 2 grammes of hydrogen isotopes at any moment
Roughly 100 tonnes of uranium oxide, loaded for 18 months
If control is lost
Plasma cools and the reaction stops within a second
Decay heat continues for years and must be actively removed
Chain reaction
None. Every reaction needs the conditions maintained
Self-sustaining by design; moderated by control rods
Long-lived waste
Activated steel; almost all below-ground disposal at 100 years
Spent fuel requiring isolation on a 10,000-year timescale
Proliferation route
No fissile material produced or present on site
Plutonium in spent fuel; safeguarded internationally
Energy per kilogramme
339 TJ from D–T, about four times fission per unit mass
80 TJ from fully burnt U-235
Waste figures are volumes at 100 years after shutdown for a notional 500 MWe plant of each type. Fusion’s activation inventory depends heavily on the steel you choose; ours is a reduced-activation ferritic-martensitic grade specified for exactly this reason.
Sheet 06 Programme to grid connection Rev C · 19.08.26
Eleven gates.
Six are behind us.
Drag the scrubber, or use the arrow keys, to walk the programme from the first coil test in 2019 to grid connection in 2035. Dates after August 2026 are planned; we have moved two of them once and published the reason both times.
2026
Complete
Series C closed at £412m
Led by Fenchurch Deep Capital with the Office for Fusion Deployment taking 18%. The round funds TARN-2 long-lead procurement to the end of 2029.
Gate 6 of 11 · closed 27 March 2026
Sheet 07 British supply chain Rev C · 19.08.26
Forty-one suppliers.
Thirty-eight of them
within a day’s drive.
Tape wound in Wrexham, vessel forged in Dundee, tungsten sintered in Leeds, cryoplant built in Sheffield. This is not sentiment: a supply chain you can visit on a Tuesday is one you can fix on a Wednesday. Filter by discipline to see who makes what.
Filter suppliers by discipline
Schematic. Not to survey accuracy. 12 sites shown.
Ashfield Cryogenics
Sheffield
The 20 K helium refrigeration plant, 41 kW at temperature, with the cold box and quench dump resistors.
Tay Forge & Fabrication
Dundee
316LN double-wall vessel sectors, electron-beam welded, and the 3,100 t cryostat base plate.
Delamere Vacuum
Runcorn
Cryopumps, torus exhaust and 62 port assemblies, each leak-tested to 1×10⁻¹⁰ mbar·l/s.
Ilkley Tungsten
Leeds
Divertor monoblocks and cassette bodies. Sinters its own powder, the only UK firm that still does.
Halstead Ceramics
Stoke-on-Trent
1,120 first-wall tungsten-on-CuCrZr panels, brazed and ultrasonically inspected in-house.
Marchwood Beryllium
Southampton
Beryllium multiplier pebbles for the blanket, and the handling cells they must be packed in.
Skerne Power Systems
Darlington
Tritium storage beds, cryogenic distillation columns and the startup inventory contract to 2036.
Lledr Diagnostics
Bangor
Neutron cameras, gamma spectrometry and the tritium accountancy instruments the regulator reads.
Kessock Magnet Structures
Inverness
316LN coil cases and inter-coil structures that take 4.1 GJ of magnetic load without moving.
Pentridge Robotics
Bristol
The 11 m articulated boom that exchanges divertor cassettes with no human in the hall.
Tarn Fusion
Culham
Assembly hall, plasma control, integrated modelling, and the only place the machine exists at once.
Sheet 08 Open roles Rev C · 19.08.26
We are 188 people
and need about
forty more.
Salaries are published because arguing about them wastes everyone’s week. Everything is on site at Culham or Wrexham — you cannot commission a magnet over a video call. Four days a week in the building, the fifth wherever you like.
Filter open roles by team
Senior Plasma Control Physicist
£82,000–£98,000+ 12% bonus
You will own vertical stability on TARN-1 and write the control laws TARN-2 inherits. We want someone who has personally lost a plasma at 3 a.m. and knows why.
Principal Engineer, Cryogenics
£95,000–£112,000+ 12% bonus
Own the 20 K helium loop end to end, from the Ashfield cold box to the last current lead. Process background welcome; you do not need to have worked on fusion before.
Coil Winding Technician (×4)
£41,000–£52,000+ shift premium
Hands on 12 mm tape, ten hours a day, to a tension tolerance of ±2 N. Composites, aerospace bonding or high-voltage winding backgrounds all transfer well.
Tritium Accountancy Lead
£74,000–£89,000+ 12% bonus
Every gramme, tracked, to the satisfaction of the Environment Agency. Nuclear, pharmaceutical or precious-metals accountancy experience all count.
Remote Handling Operator
£54,000–£66,000+ shift premium
Drive the Pentridge boom in a mock-up now and inside an activated vessel later. Subsea ROV pilots have made this jump twice already and both were excellent.
Head of Grid & Offtake
£110,000–£135,000+ equity
Negotiate the connection agreement and the first offtake for a plant that does not exist yet. You will need to be comfortable being told no for about three years.
6 roles shown. Applications to careers@tarnfusion.co.uk — a person reads every one.
Sheet 09 What others say Rev B · 02.07.26
People who have
every reason
to be sceptical.
We asked for the quotes that were true rather than the ones that were flattering, which is why two of these contain a caveat.
The TF-04 result is the first time I have seen a non-insulated REBCO coil hold spec through a full fatigue campaign at that field. It does not solve the exhaust problem, and Tarn are the first to say so, but it changes what size of machine is worth arguing about.
Institute of Applied Magnetics
We have supplied four fusion programmes. Tarn is the only one that sent an engineer to stand in our winding hall for six weeks before placing the order. The tolerances they asked for were uncomfortable and they were also correct.
Brennan Superconductors, Wrexham
Their tritium accountancy submission was the most complete first draft we have received from a private operator. That is a low bar and they cleared it by a distance. The blanket work is still early and the 2033 date assumes nothing goes wrong twice.
(writing in a personal capacity)
Sheet 10 The awkward questions Rev C · 19.08.26
Yes, we have
heard the joke
about thirty years.
If your question is not here, write to ask@tarnfusion.co.uk. A physicist or an engineer answers, usually within two working days, and they are allowed to say they do not know.
Is fusion always thirty years away?
It was, and the reason was funding rather than physics: the 1976 US programme plan set out what each budget level would deliver, and the level actually funded was the one labelled “fusion never”. What changed is the magnets. HTS tape reaches the same triple product in a machine a third of the linear size, and small machines can be built, broken and rebuilt on a commercial timescale. We promise a machine in the ground in 2031 and first plasma in 2033, and we publish the slip when there is one.
Can a tokamak melt down?
No, and the reason is structural rather than procedural. There are about two grammes of fuel in the vessel at any moment. The plasma is not self-sustaining: it needs continuous heating, fuelling and magnetic confinement, and losing any one of them stops the reaction in well under a second. The genuine hazards are a magnet quench, which we dump in eleven seconds, and the tritium inventory, which is why the in-process figure is held below 400 grammes.
What about the waste?
There is waste, and anyone who says otherwise is selling something. Fourteen-MeV neutrons activate the steel they pass through, so at end of life you have a few thousand tonnes of radioactive structure. The difference is the half-life: with a reduced-activation ferritic-martensitic steel, chosen for this purpose and no other, essentially all of it suits near-surface disposal or recycling within a hundred years. No spent fuel, no plutonium, no geological repository.
Where does the tritium come from?
Initially from CANDU heavy-water reactors, of which there are a finite and shrinking number; we have contracted 3.1 kg for startup. After that the machine must breed its own from lithium in the blanket, which is the single largest technical risk in the programme and the reason our fuel-cycle TRL is a candid 3. Our target breeding ratio is 1.11 against a break-even of about 1.05. If that number comes in below 1.0 in the mock-up, the design changes.
Why a tokamak and not a stellarator or laser fusion?
Because the tokamak has sixty years of confinement data behind it, and the scaling laws, for all their flaws, are calibrated against dozens of real machines. A stellarator has better steady-state behaviour and much worse buildability. Inertial confinement has produced target gain, a genuine and important result, and is a long way from a driver that fires ten times a second for a year. We picked the boring option deliberately.
How much will the electricity cost?
TARN-2 is a pilot, not a power station, and its electricity will be absurdly expensive — the plant costs about £2.4bn and produces roughly 40 MW net. That is the point of a pilot. The first commercial machine, TARN-3, is sized at 480 MW net and our internal model puts it between £86 and £131 per MWh at an 8% discount rate. We are not going to publish a single number with a straight face until we have operated TARN-2 for a year.
Who regulates you?
Under the Energy Act 2023 fusion in the United Kingdom sits with the Environment Agency and the Health and Safety Executive rather than the Office for Nuclear Regulation. That is a proportionate framework rather than a light-touch one: it recognises that there is no fissile material and no criticality accident, while treating the tritium inventory and the activated structure exactly as seriously as they deserve. Our environmental permit application for TARN-2 goes in during Q2 2027.
Can I visit?
Yes. Open afternoon on the second Thursday of each month, capacity thirty: the TARN-1 hall from the gallery, and a real coil section you can put your hand on. School groups from Oxfordshire, Berkshire and Buckinghamshire are free and we keep four slots a month for them. Book at visits@tarnfusion.co.uk. Wear flat shoes; 118 steps and no lift on the gallery route.
Sheet 11 Get in touch
Come and stand
next to a magnet.
Investors, suppliers, engineers looking for a job and physicists who think our H-factor is optimistic — all equally welcome. We answer everything.
Direct lines
Mon–Fri 08:30–17:30
| General | ask@tarnfusion.co.uk |
|---|---|
| Careers | careers@tarnfusion.co.uk |
| Supply chain | procurement@tarnfusion.co.uk |
| Press | press@tarnfusion.co.uk |
| Visits | visits@tarnfusion.co.uk |
| Switchboard | +44 1235 610 4400 |