How close is fusion, really

In December 2022 a laser facility in California got more energy out of a fusion reaction than the lasers delivered into the target. That was a genuine scientific milestone, fifty years in the making, and it was widely misreported.

The gain measured was against the laser energy hitting the target. The lasers themselves consumed roughly a hundred times that from the electricity grid, because they are inefficient. Net energy in the sense that matters for a power plant, more electricity out than the whole facility drew in, has not happened anywhere.

This is not a debunking. Reaching scientific ignition was the precondition for everything that follows. It is a caution about which number a headline is quoting.

The two families and where each stands

Magnetic confinement holds a hot plasma in a magnetic bottle, usually a doughnut-shaped tokamak, continuously. The large international project pursuing this has been under construction for years with a long schedule and a large budget. What changed the field was high-temperature superconducting tape, which allows much stronger magnets in a much smaller machine. Since confinement improves steeply with magnetic field strength, a stronger magnet means a far smaller and cheaper reactor, and that single materials advance is why private fusion became fundable.

Inertial confinement compresses a fuel pellet with lasers, a brief event repeated many times a second. This is where ignition was achieved. Turning a shot every few hours into ten shots a second, with a fresh precisely manufactured target each time, is the engineering problem, and nobody has demonstrated it.

Several other approaches exist, including magnetised target and field-reversed configurations, and some are well funded. The honest position is that no approach has a decisive lead.

The problems that are not physics

Neutron damage. The reaction throws off high-energy neutrons that embrittle and activate the structure around them. Materials that survive years of this flux are still being developed, and the facilities to test them properly barely exist. This is arguably the single hardest unsolved problem.

Tritium. The most accessible reaction needs tritium, which is radioactive, has a twelve-year half life and exists in tiny quantities worldwide. A plant must breed its own from lithium in a blanket around the reaction, recover it and reuse it, with a breeding ratio above one. This has never been demonstrated in a closed loop.

Heat to electricity. A fusion plant still boils water to drive a turbine. The nuclear island is exotic, the power island is conventional, and conventional plant costs money.

Duty cycle. A plant that runs for a few minutes and then needs maintenance is not a plant. Continuous operation for months is a requirement nobody has approached.

Why the money arrived anyway

Private fusion has attracted billions, and several firms have signed power purchase agreements with technology companies for delivery in the 2030s. Those agreements are best read as funding commitments rather than as capacity anybody is counting on, and the buyers know it.

The case for the investment is straightforward: if it works, the payoff is enormous, and the cost of finding out has fallen by an order of magnitude thanks to better magnets, better simulation and cheaper prototyping. That is a reasonable bet even at a low probability, which is what it is.

The honest timeline

A machine that demonstrates net electricity, briefly, some time in the 2030s is plausible. A commercial plant selling power competitively is a 2040s question at the earliest, and that assumes materials, tritium and duty cycle all resolve. Anyone giving you a date in the 2020s is quoting a milestone, not a plant.

Meanwhile data centers need power now, and the technologies that will actually supply it this decade are wind, solar, storage, gas and existing nuclear.

What to watch

Watch for a demonstrated closed tritium breeding cycle, which would be a bigger deal than any plasma record. Watch sustained plasma duration in the new high-field machines. And watch whether any private firm publishes a wall-plug energy balance rather than a scientific gain figure.