Don't conFUSEme

Nuclear fusion has been thirty years away for about seventy years now. This is not an accident. It is not even particularly a failure. It is a reflection of the fact that what we are attempting to do is essentially build a small star in a room, then extract useful electricity from it without melting the room, and it turns out this is quite difficult.

I've been following fusion research for a long time, partly out of genuine fascination and partly because the stakes are high enough that it deserves serious attention. If we crack it, energy scarcity becomes a historical curiosity. If we don't, the alternatives are various degrees of uncomfortable. So: what's actually going on, and when, if ever, should we expect to be able to plug our kettles into a star?

 

How Fusion Actually Works, and Why It's So Hard

The basic physics is straightforward enough. You take two light atomic nuclei, typically isotopes of hydrogen called deuterium and tritium, and you force them together with enough energy that they fuse into a heavier nucleus, releasing a substantial amount of energy in the process. This is what the sun does. The sun, however, has the advantage of being roughly 1.4 million kilometres across and having the gravitational force of an enormous amount of mass to do the forcing. We are attempting to replicate this process in a machine that fits on a reasonably large industrial site.

The challenge is confinement. To initiate fusion, you need a plasma, which is matter heated to the point where electrons have separated from nuclei, at temperatures exceeding 100 million degrees Celsius. This is hotter than the core of the sun. You cannot put this in a box made of ordinary materials, because ordinary materials would instantly vaporise. You have to hold it using something that isn't a material, and there are essentially two serious candidates: magnetic fields and inertial momentum. The engineering required to do either of these reliably, at scale, and without consuming more energy than the fusion reaction produces, is the challenge that has consumed several generations of physicists.

 

The Tokamak, and Its Competitors

The dominant approach for the past several decades has been the Tokamak, a doughnut-shaped chamber in which powerful magnetic fields contain the plasma. It is, by any reasonable standard, an extraordinary piece of engineering. It is also extremely expensive, extremely temperamental, and has a persistent tendency to experience plasma instabilities at inconvenient moments, which is something of an understatement when you consider that "plasma instability" in a Tokamak means the containment briefly fails and several hundred million degrees of superheated gas touches the reactor wall.

The Tokamak's main institutional expression is ITER, the International Thermonuclear Experimental Reactor being built in southern France by a collaboration of 35 nations. ITER is breathtaking in ambition and somewhat sobering in timeline. First plasma is currently expected sometime in the 2030s. A demonstration power plant connected to the grid is a 2040s proposition. Commercial fusion on any meaningful scale is a 2050s story at the optimistic end. I am not criticising ITER for this, because the timescales reflect the actual difficulty of the problem rather than any particular institutional failure. But it is worth being clear-eyed about what "fusion is coming" actually means in calendar terms.

The stellarator is an alternative magnetic confinement approach, using a more complex twisted field geometry that offers potentially better plasma stability than the Tokamak. It is considerably harder to engineer and has historically been less developed, but Germany's Wendelstein 7-X has been producing encouraging results and the approach is attracting renewed interest. Magnetic confinement research more broadly is exploring various field configurations beyond the classical Tokamak, and this diversity of approach is probably healthy.

The other main strand is inertial confinement fusion, where instead of containing plasma in a magnetic field you compress it so rapidly and intensely that it fuses before it has time to escape. The National Ignition Facility in California fires an enormous array of lasers simultaneously at a tiny fuel pellet, crushing it to densities and temperatures sufficient for fusion. In late 2022, NIF achieved ignition for the first time, meaning the fusion reactions produced more energy than the lasers delivered to the target. This was a genuine scientific milestone. The laser system itself consumed roughly a hundred times more energy than the target received, so "net energy gain from the facility" is still some way off, but ignition is the fundamental physics milestone and it matters.

 

The Private Sector Enters the Room

What's changed significantly in recent years is the arrival of well-funded private companies pursuing fusion on compressed timescales, with a variety of novel approaches. This is not the same as the speculative fusion startups of previous decades. Several of these companies are backed by serious investors, run by credible physicists, and are genuinely exploring approaches that differ meaningfully from the academic mainstream.

Helion Energy is probably the most interesting of these. Rather than the standard deuterium-tritium fuel cycle, Helion is pursuing deuterium and helium-3 fusion using a pulsed approach where two magnetised plasma rings are fired at each other and compressed. Unusually, they aim to extract energy directly from the plasma rather than using the fusion reaction to boil water and drive a turbine, which is thermodynamically more efficient if you can make it work. Microsoft has signed a power purchase agreement with them for delivery by 2028, which is either a bold bet on a genuine technological breakthrough or an extremely ambitious piece of corporate positioning, and we will find out which in due course.

Commonwealth Fusion Systems is pursuing a more conventional Tokamak approach but using new high-temperature superconducting magnets that are significantly more powerful than previous designs, allowing a much smaller and cheaper device. Their SPARC reactor, if it performs as modelled, would achieve net energy gain in a machine roughly a tenth the size of ITER. First experiments are planned for the late 2020s.

The entry of the private sector doesn't mean fusion is suddenly easy. It means that for the first time, there are organisations with genuine incentives to move quickly, the capital to do so, and the freedom to take approaches that large international collaborations can't easily pivot to. This changes the probability distribution of outcomes over the next decade rather than guaranteeing any particular result.

 

What Fusion Actually Offers, and What It Doesn't

When fusion does arrive, the advantages are substantial. The fuel, primarily deuterium extracted from seawater and tritium bred from lithium, is for all practical purposes inexhaustible. The reaction produces no carbon dioxide. The radioactive waste is orders of magnitude less problematic than fission waste, and decays to safe levels in decades rather than millennia. There is no chain reaction to run away with you, no Chernobyl scenario, no risk of a fusion plant becoming a weapon. A fusion reactor that loses containment simply stops fusing, because the conditions for fusion are so precise and so difficult to maintain that the reaction extinguishes itself.

What fusion doesn't offer is a free pass on physics. A fusion power plant still generates heat, and heat still has to be managed. If you're imagining that fusion will make data centres trivially cheap to run by eliminating their energy costs, you've solved half the problem. The other half is that data centres generate heat as a fundamental consequence of computation, and that heat still has to go somewhere regardless of where the electricity came from. Thermodynamic limits are not negotiable, even with unlimited clean electricity. Efficiency improvements in computing remain important independently of what happens with energy generation.

 

Where We Actually Are

The honest position in September 2022 is this: fusion is more likely to arrive as a meaningful energy source this century than at any point in the past thirty years, the range of serious approaches has broadened considerably, and the entry of private capital has introduced a genuine urgency that public research programmes struggle to sustain. NIF's ignition result and the rapid progress on high-temperature superconducting magnets are both genuinely significant developments that have shifted the probability landscape.

It is also still true that commercial fusion power before the 2040s would require something close to a breakthrough rather than incremental progress, and that "fusion is coming" has been said so many times, for so long, that treating any given timeline with scepticism is entirely rational.

My position is cautious optimism held at arm's length. The physics is sound. The engineering challenges are enormous but not obviously insurmountable. The incentives are better aligned than they've ever been. I would not bet my retirement on a specific delivery date, but I no longer think it's reasonable to dismiss fusion as permanently just over the horizon.

Watch the 2020s. They should be informative.

All Posts