This is a linkpost for "The future of fusion" book review.

IMO it's not really a book review, it's a well-written and data-driven argument that commercial fusion power is about 10-15 years away, no really for real this time. (!!!) Good discussion in the comments too. I'm interested to see what people here think. [ETA: the argument is for fusion power in less than 10-15 years; I mentally added a few years and slapped "commercial" on the front, though yeah I admit it could plausibly take longer idk.]

The author's credences/timelines:



I'm going to update my post on what 2040 looks like assuming no singularity accordingly.

New Comment
8 comments, sorted by Click to highlight new comments since:

IMO it's not really a book review, it's a well-written and data-driven argument that commercial fusion power is about 10-15 years away, no really for real this time.

It doesn't say that. It says that (he thinks) they'll get Q>5 in research reactors in 10 to 15 years.

Then they can start doing all the research to figure out the systems surrounding the core reactor.

Then they can do actual engineering on systems that produce commercial power. Probably a lot of that commercialization engineering will require further experimental tests, since there's so much new stuff involved.

Then they can get the permits and build them.

If the most optimistic predictions in that review came true right on schedule, you might have actual commercial fusion up and putting power into the grid in 30 years.

I said IMO. But yeah, I should clarify that this isn't a quote; thanks for pointing that out.

And yeah now that you mention it the regulatory hurdles will probably be a nightmare. Plus it's not like Elon Musk is in charge; they probably won't be able to go from prototype to commercially viable in 5 years... I'm updating my timelines upwards.

As someone who works in this field, let me add some thoughts:

  • Marvel is a joke. I put it at less than 10% that they would even beat the NIF record by 2035.
  • ITER is planned to turn on by 2025, but by now that is late 2025 ("turn on" means "create a plasma that is no more interesting than that in a fluorescent lamp"). From inside channels I know they plan to announce further delays later this year. Even if everything goes according to the present schedule, actual deuterium-tritium operation is planned to start in 2035. They will ramp thing up carefully, there will be no q=10 success on the first try.
  • Even now, the design of crucial parts of the ITER heating system is not finished.
  • The design of CFETR is not finished, it is not intended to be a commercial power plant (more of a post-ITER experiment), and the Chinese institute of plasma physics plans to build an intermediate tokamak, called BEST (I kid you not), before they start building CFETR.
  • Don't confuse Tokamak Energy with the Culham Centre for Fusion Energy, the location of the Joint European Torus. Tokamak Energy merely rents some office space there (possibly in a deliberate ploy to leech credibility?). The successes of JET are not Tokamak Energy's.

Update: Commonwealth Fusion Systems is the Same Scale as OpenAI (aiimpacts.org)

Seems like CFS (which is responsible for SPARC) is now aiming for 2025? So basically things are on track maybe?

I have a question about tritium breeding. How many free neutrons does D-T fusion produce? According to Wikipedia it is only one. Assuming less than 100% efficiency, this means we produce less than one tritium nucleus per reaction, meaning than we are losing tritium overall. How can we make up the shortfall?

Beryllium acts as a neutron multiplier: .

The consumable Be is supplied either as a solid metal layer, or as part of the molten salt (such as FLiBe) coolant, or both. The right amount of Be in the path of D-T neutrons theoretically allows the fusion reactor to have a Tritium Breeding Ratio (TBR) greater than one. That's the good news!

Speaking practically, this hasn't been shown to work yet, and my understanding is that it is a must-have for commercial D-T fusion. Recycling tritium within an operating D-T fusion plant (and beyond that, increasing global T inventory to bootstrap / enable the world to bring new fusion reactors online) is one of the key remaining engineering challenges to viable D-T fusion power, and is an active research area. The greater-than-unity breeding side of the problem is only the first challenge; separating, extracting, and capturing the freshly baked tritium is also an unknown. But we've known about this problem for a while, and there is room for optimism; ITER, for example is planning to test 6 different designs for tritium breeding blankets.

Does the Beryllium convert one fast neutron into two thermal neutrons? I assume we're not getting money for nothing.

Tritium has a half life of twelve years, so perhaps we could wait to disassemble the blanket, no rush to extract the tritium.

There is a plasma physicist on YouTube who made a video explaining why, in his opinion, fusion power is far away. Tritium was one of the issues he raised.

You also get neutrons from D-D reactions (which produce 1 neutron about half the time). Even if the plasma isn't hot enough to get a lot of D-D reactions, it will likely produce some to make up the shortfall.

Edit: and I completely forgot (lol) that when DD doesn't produce a neutron it produces a tritium. So, that will also help.