Segal Fusion
Hold a star still enough to tap it.
Magnetic-confinement fusion research. We simulate, confine, and control high-temperature plasma — driving toward repeatable net-positive energy gain, not a single demonstration shot.
Confine. Ignite. Repeat.
Fusion taps the steepest gradient there is — the binding energy inside the nucleus, the same reaction that lights a star. The world can ignite a plasma. The harder question is whether it can do so again the next day, and the day after — fueled, maintained, and connected to a grid — without spending more energy than it makes. Segal Fusion engineers the whole machine around the shot: confinement, fuel breeding, and materials are design constraints we carry from the first simulation, not downstream logistics. Net gain is the milestone. A repeatable power plant is the product.
Fusion needs a plasma held at over 100 million kelvin — hotter than the Sun's core — long enough and dense enough that the energy released beats the energy spent confining it. That product of density, temperature, and confinement time is the Lawson criterion, and it is the entire game. We pursue it with magnetic confinement: high-temperature-superconductor (REBCO) magnets generating the high fields that hold the plasma off the wall in a compact volume, where the economics of a power plant actually close.
High field is leverage — fusion power density scales roughly with the fourth power of the magnetic field, so a 20-tesla magnet does the work of a far larger, weaker machine. The remaining physics is unforgiving: disruptions that must be predicted and steered before they quench a shot, a divertor that must survive heat fluxes comparable to a rocket nozzle, a first wall that must tolerate 14 MeV neutrons, and a lithium blanket that must breed more tritium than the plasma burns. We treat each as an instrument to be measured, not a hope to be stated.
- Q>1
- Net-positive gain — energy out beyond energy in— engineering target
- 20 T
- On-axis field from HTS magnets, full operating point
- >1
- Tritium breeding ratio — fuel self-sufficiency— target
- 100 M K
- Plasma temperature the confinement must hold
SIMULATE
MHD + transport models, compiled to real-time control.
CONFINE
High-field HTS magnets hold the plasma off the wall.
IGNITE
Drive the plasma past break-even — Q>1, repeatably.
CONVERT
Fusion heat to a turbine to grid-firm electrons.
BREED
Lithium blanket breeds tritium; the fuel loop closes.
Plasma simulation & real-time control
High-fidelity MHD and transport models compile straight to a real-time control loop — predicting and steering the plasma shot by shot.
Disruption prediction on the control loop
Machine-learning inference runs inside the latency budget, flagging an incipient disruption with a stated confidence interval in time to act.
High-field HTS magnet engineering
REBCO-tape magnets with active quench detection and engineered joints hold the plasma at the fields that make a compact reactor possible.
Confinement geometry & first-wall layout
Detail on confinement geometry & first-wall layout is gated under U.S. export-control rules. Verified partners can request access — we confirm jurisdiction and eligibility before sharing.
Request accessDivertor & first-wall materials
Divertor heat-flux handling and neutron-tolerant first-wall materials — the surfaces that decide how long a machine runs between rebuilds.
Tritium inventory & breeding control
Lithium breeding blankets target a breeding ratio above one; inventory is tracked end to end with Segal Resources so the fuel loop can close.
Net gain is necessary but not sufficient. The unsolved variables are repeatability on a duty cycle and survival under neutron flux — a divertor and first wall that last, and a blanket that breeds more tritium than the plasma burns. A record shot is a press release; a machine that does it every day, fueling itself, is a power source.
Plasma Control Suite
A real-time plasma control and disruption-prediction terminal — coupling the simulation stack to magnet and heating actuators on a shot-by-shot basis.
- Real-time MHD inference inside the control latency budget
- Actuator coupling — magnets, heating, fueling on one loop
- Shot replay with full state reconstruction
- Disruption-prediction confidence intervals, not point alarms
Helix-0 Test Stand
HTS magnet coil sustained at 20 T, full operating field.
First in-house REBCO coil to reach full design field with active quench detection, validating the confinement geometry the compact reactor design depends on.
First Plasma
Initial confined plasma, control loop in the loop.
Commissioning the integrated machine — magnets, heating, and the Plasma Control Suite — toward a first sustained, instrumented discharge.
Divertor Heat-Flux Test
Divertor target under reactor-relevant steady-state heat flux.
Standalone test of divertor materials and cooling against the heat fluxes a power-plant scrape-off layer will impose — measured, not modeled.
Control-Loop Latency Target
Disruption inference + actuation within the disruption growth time.
Close the loop fast enough that a flagged disruption can actually be steered — the latency budget that makes prediction useful rather than forensic.
Sustained-Discharge Ladder
Stepwise increase in confined-discharge duration, shot over shot.
A duration ladder from seconds toward steady state — the honest path to a duty cycle, climbed one validated rung at a time.
Net Energy Gain
Repeatable Q > 1, energy out beyond energy in.
Not a single record shot, but net-positive gain reproduced on a duty cycle — the threshold of a real power source.
Disruption prediction within the latency budget
We measure how early a model can flag a disruption and how fast the actuators can respond. The open question is margin — predicting in time to steer, not just in time to log.
Divertor heat-flux handling
Steady-state heat fluxes rival a rocket nozzle. We are testing target materials and cooling against reactor-relevant loads to find the duty-cycle ceiling.
Neutron-tolerant first-wall materials
14 MeV neutrons embrittle and transmute the first wall. Lifetime under fluence is an open materials problem we treat as a measured curve, not an assumption.
Tritium breeding self-sufficiency
A power plant must breed more tritium than it burns. We model and bench-test lithium blanket designs toward a breeding ratio above one — the line the fuel cycle has to clear.
Q is fusion power out divided by heating power in. Q > 1 is scientific break-even at the plasma. A power plant needs engineering gain well above that — we state the plasma Q and the plant-level number separately, never conflated.
Fusion power density scales roughly with the fourth power of magnetic field. HTS (REBCO) tape reaches ~20 T, which shrinks the machine that delivers a given power — the difference between a compact reactor and a stadium.
No. It is a research program with one achieved magnet milestone and a first-plasma commissioning in progress. Net gain and a duty cycle are on the roadmap, marked as such.
Work with the Fusion team.
Direct to the program team. No newsletter, no funnel.