The Real State of Nuclear Fusion

>by Roman Tsyupryk

>July 11, 2026

This text is just my thoughts out loud. I'm only a human being trying to analyze current information and imagine what might happen in the future. My thoughts could be completely wrong or might be just "noise" or they could be food for brainstorming about "what if..." scenarios.

Where the technology that powers the stars actually stands: the physics, the competing machines, the milestones, and the honest distance still to travel.

1. The Foundation: What Fusion Is

Nuclear fusion is the process that powers the Sun and every star: light atomic nuclei merge into heavier ones, and the tiny mass difference is released as energy according to E = mc².

It is the opposite of nuclear fission (today's nuclear plants), which splits heavy atoms like uranium. The comparison matters:

Property	Fission (today's nuclear)	Fusion
Fuel	Uranium, plutonium (mined, finite)	Hydrogen isotopes (effectively unlimited)
Long-lived radioactive waste	Yes, thousands of years	No (only short-lived activated materials)
Meltdown risk	Yes (needs active shutdown)	No (plasma simply dies if anything fails)
Weapons proliferation link	Direct	None from the reaction itself
Energy density	~1 million × chemical	~4 × fission per unit fuel mass

One kilogram of fusion fuel yields roughly the energy of 10 million kilograms of coal. Seawater contains enough deuterium to power civilization for millions of years.

The fuels. The easiest reaction, and the one nearly everyone pursues, is deuterium-tritium (D-T): it ignites at the "lowest" temperature (~100–150 million °C) and yields the most energy per collision. Harder but cleaner alternatives exist:

D-D (deuterium only): fuel from seawater, no tritium breeding needed, but requires higher temperatures
D-³He (deuterium + helium-3): mostly aneutronic (few neutrons), enabling direct electricity conversion; helion-3 is rare on Earth
p-¹¹B (proton + boron-11): fully aneutronic, abundant fuel, but needs ~1 billion °C or clever non-thermal tricks

2. Why It Is So Hard

Atomic nuclei are positively charged and repel each other ferociously (the Coulomb barrier). To force them to fuse, you must make them collide at enormous speeds, which means heating the fuel into a plasma at over 100 million °C, roughly ten times hotter than the Sun's core. (The Sun compensates with crushing gravitational pressure and sheer patience; we have neither.)

The whole game is captured in one figure of merit, the Lawson triple product:

density × temperature × confinement time  ≥  threshold

Hold enough plasma, hot enough, together long enough, and the fusion reactions release more energy than you spend heating it. Every fusion machine ever built is a different strategy for maximizing this product, and every strategy fights the same enemy: plasma at 100 million °C is the most unruly substance imaginable. It writhes, kinks, and leaks through instabilities faster than theory once predicted. No material can touch it, so it must be held by magnetic fields or crushed faster than it can escape.

The key metric: Q. Fusion gain Q = fusion energy out ÷ heating energy in.

Q = 1: "scientific breakeven"
Q ≈ 5: burning plasma (self-heating dominates)
Q ≥ 10: the traditional target for a power plant's plasma
Engineering breakeven: the whole facility (including inefficient lasers, magnets, cooling) produces net electricity. This is much harder than Q = 1, and nobody has done it.

3. The Implementations: Every Way We Try to Bottle a Star

There are two grand families: hold the plasma with magnets for a long time, or crush it so fast it fuses before it can fly apart, plus hybrids and outliers.

3.1 Magnetic Confinement Fusion (MCF)

Tokamak: the mainstream. A doughnut-shaped (toroidal) chamber where magnetic coils plus a current driven through the plasma itself create a twisted magnetic cage. The most studied, best understood design; nearly all record plasma results come from tokamaks (JET, JT-60, EAST). Weaknesses: the plasma current can violently collapse (disruptions), and operation is naturally pulsed. Who: ITER, Commonwealth Fusion Systems (SPARC/ARC), Tokamak Energy, China's EAST and CFETR.

Stellarator: the tokamak's twisted sibling. Instead of driving a current through the plasma, the magnetic twist is built into fantastically complex 3D-shaped coils. Result: intrinsically steady-state, no disruptions, at the cost of nightmarish engineering that only became practical with supercomputer optimization. Germany's Wendelstein 7-X is the flagship: in May 2025 it set a world record for the triple product sustained over long durations (43 seconds), surpassing all long-pulse tokamak results. Who: W7-X (Max Planck IPP), Proxima Fusion, Type One Energy, Thea Energy.

Spherical tokamak: squash the doughnut into a cored apple. Better plasma pressure for a given magnetic field (higher β), potentially much smaller and cheaper machines. Who: UK's MAST-U and the STEP program (grid pilot targeted ~2040), Tokamak Energy.

Field-Reversed Configuration (FRC): no doughnut at all. The plasma is a self-organized smoke-ring of current confined largely by its own magnetic field, in a simple linear machine. Naturally high β, geometrically suited to advanced aneutronic fuels. Who: TAE Technologies (targeting p-¹¹B), Helion Energy (a pulsed FRC variant).

Z-pinch: the oldest idea, reborn. Run a colossal current through a plasma column; its own magnetic field pinches it inward. Classically doomed by instabilities, but sheared axial flow stabilization revived it. No external magnets, no lasers, potentially the cheapest hardware in the field. Who: Zap Energy (backed by Chevron).

Magnetic mirror: a straight tube with magnetic "plugs" at each end. Abandoned in the 1980s, quietly returning with modern superconductors. Who: Realta Fusion, University of Wisconsin's WHAM.

3.2 Inertial Confinement Fusion (ICF)

Laser indirect-drive: the approach that made history. 192 laser beams heat a gold cavity (hohlraum) that bathes a peppercorn-sized fuel capsule in X-rays, imploding it to densities greater than the center of the Sun for a few billionths of a second. In December 2022, the National Ignition Facility (NIF) achieved the first-ever ignition: more fusion energy out than laser energy in. It has since repeated ignition roughly ten times, reaching a record gain of 4.13 (8.6 MJ from 2.08 MJ of laser light) on April 7, 2025 (still the standing record), with its tenth ignition shot in October 2025 coming in lower (gain 1.74, 3.5 MJ). Crucial honesty: NIF's lasers draw ~300 MJ from the wall to deliver those ~2 MJ of light: the facility is nowhere near net energy, and was never designed to be. It is a physics demonstration (and a weapons-science facility), not a power plant prototype. Who: NIF (LLNL); commercial laser-fusion startups Xcimer Energy and Focused Energy aim to redo it with modern, efficient lasers.

Magnetized Target Fusion (MTF), the hybrid: a magnetized plasma (easier to heat) is mechanically compressed (easier than lasers). General Fusion's design collapses a rotating wall of liquid metal around the plasma: the liquid metal doubles as the neutron absorber and heat-extraction system. Who: General Fusion (Canada).

Projectile / pulsed approaches: first-principles cost reduction: First Light Fusion (UK) fired hypervelocity projectiles at targets; Helion's Polaris rams two FRC plasma rings into each other and, uniquely, plans to recover electricity directly from the expanding magnetized plasma pushing back on the coils, skipping the steam turbine entirely.

3.3 The Outliers

Muon-catalyzed fusion: replace electrons with heavy muons and fusion happens at room temperature; sadly, making muons costs more energy than the fusion returns. Dormant but not dead.
Fusion-fission hybrids: use fusion neutrons to drive a subcritical fission blanket; studied seriously in China.
"Cold fusion" / LENR: not supported by reproducible evidence; excluded from serious roadmaps.

4. Where We Actually Are (mid-2026)

The scoreboard, stripped of hype:

Milestone	Status
Scientific breakeven (Q > 1, plasma)	✅ Achieved (NIF, Dec 2022, repeated ~10× since)
Target gain > 4	✅ NIF, April 2025 (8.6 MJ)
Burning plasma (self-heating dominant)	✅ NIF (2021–2025 shots)
Long-duration high-performance plasma	✅ W7-X record: triple product held 43 s (May 2025)
Private machine running D-T fuel	✅ Helion's Polaris (2026, first privately built)
Engineering breakeven (net electricity from the wall)	❌ Nobody. Not close.
Electricity delivered to a grid	❌ Nobody
Commercial power plant	❌ None under operating license anywhere

The big public projects:

ITER (France, 35 nations): the largest science project on Earth, a tokamak designed for Q ≥ 10. In November 2024 its schedule was formally rebaselined: the "Start of Research Operation" (first D-D plasmas) slipped from 2025 to 2034, with full D-T operation targeted for 2039. Assembly is accelerating (vacuum-vessel sector modules are being lowered into the pit and the central solenoid is stacking up), but ITER will never sell a watt; it is a physics machine.
NIF (USA): keeps extending the ignition record book, but its 1980s-era lasers make it a dead end for energy.

The private race: Private fusion investment reached roughly $9.8 billion cumulative (per the Fusion Industry Association's 2025 report, through mid-2025), and the company count more than doubled from 23 to 53. The front-runners:

Commonwealth Fusion Systems: the best-funded (>$2 B). Its breakthrough is high-temperature superconducting (HTS) magnets (~20 tesla), which shrink a Q>1 tokamak from ITER-size to gymnasium-size. SPARC's magnet ring is due to be complete in 2026; first plasma targeted 2027; the follow-on ARC plant in Virginia (with a 200 MW Google power-purchase agreement) targets the early 2030s.
Helion Energy: the most aggressive timeline. Helion's own newsroom reports Polaris hit 150 million °C and measurable D-T fusion in February 2026 (a company-reported result, not yet independently verified by a third party). Microsoft has a signed contract for power from Helion's first 50 MW plant (Orion, Washington state) starting 2028 (a date almost everyone outside Helion considers extremely optimistic, but it is a real contract with penalties).
TAE Technologies: FRC machines aimed at p-¹¹B aneutronic fusion; planning a 50 MWe plant, going public via merger in 2026.
Zap Energy, General Fusion, Tokamak Energy, Proxima Fusion, Xcimer: each betting a different geometry can beat the tokamak on cost.
Pacific Fusion: a $900M-funded pulsed-power (MagLIF-style inertial) approach; a company-reported figure (via press exclusive, not an independent or peer-reviewed publication) claims 440 gigawatts of peak power in an 80-nanosecond pulse in June 2026, with net facility gain targeted by 2030 (a serious newer entrant alongside the laser-fusion startups above, but treat the specific number as unverified).

5. The Unsolved Problems Nobody Should Skip

Ignition made headlines, but four quieter problems stand between Q > 1 and a power plant:

Tritium. D-T plants must breed their own tritium (from lithium blankets hit by fusion neutrons): the world's entire civilian tritium inventory is ~25 kg, and no one has ever demonstrated a self-sufficient breeding blanket. This is arguably the hardest unsolved engineering problem in fusion.

Materials. A D-T reactor wall absorbs a neutron bombardment fiercer than anything in fission, swelling and embrittling the best steels. Dedicated test facilities (IFMIF-DONES) are only now being built.

The duty-cycle gap. Record shots are one thing; a power plant must run ~90% of the year for decades. NIF fires about once a day; a laser-fusion plant needs ~10 shots per second.

Cost. Fusion doesn't just need to work: it needs to beat solar-plus-storage, which gets cheaper every year fusion stays in the lab. This is why so many startups obsess over machine simplicity (Zap's magnet-free pinch, Helion's turbine-free direct conversion) rather than raw plasma performance.

6. An Honest Timeline

Now–2030: SPARC attempts Q > 1 in a compact tokamak (2027–28); Helion attempts first fusion electricity (2028); W7-X pushes toward 30-minute plasmas. Expect at least one high-profile failure and one genuine surprise.
2030s: First pilot plants (ARC in Virginia, UK's STEP, China's CFETR) attempt net electricity to a grid. ITER finally reaches first plasma (2034) and D-T burn (2039). Market analyses cluster first commercial operation between 2030 and 2035 at the optimistic end, 2040+ conservatively.
2040s and beyond: If pilot plants work, the question becomes purely economic: can fusion be cheap, not just possible?

The most honest single sentence about mid-2026: the physics is essentially proven, the engineering is genuinely underway, and the economics is entirely unproven. Fusion has stopped being "thirty years away," but it has not yet become "ten years away" for grid electricity, no matter what any press release says.

See also: The First Sun, a short story imagining what humanity does with fusion once it finally, fully works.

Sources

Written as a companion piece to The First Sun, the facts first, then the dream.

Part of the Still Becoming series. These two articles are just a small part of a larger Still Becoming series exploring how different technologies could complement each other.

→ Next in the series: The First Sun