At the risk of coming off as a nay-sayer, let's say engineering hurtles related to fusion power generation is overcome. How is the presumably high upfront capital costs going to compare with the ROI?
That is, it would seem likely that fusion power would be costly to build. It would also seem apparent that if it were to fulfil its promise then the power it generates is sold at or less than the current amount. That would then seem to imply a lengthily time to make a return on the initial investment. Or am I missing something else with this equation?
It's not only initial investment. Half of the fusion fuel is tritium, which is one of the most expensive substances on Earth (a google search finds that the price of tritium is about $30k per gram [1]). For comparison, fission reactors need enriched uranium, and that costs only about $4000 per kilogram [2]. People have the idea that fusion produces many times more energy than fission, probably because fusion bombs have a higher yield than fission bombs. This is not true. The most typical fusion reaction involves one deuterium and one tritium and yields 17.5 MeV from a total or 5 nucleons. A fission reaction involves one neutron and one atom of U-235 and yields 190 MeV from 236 nucleons. So fusion yields about 4.3 times more energy per nucleon. That's respectable, but in the popular imagination fusion yields 100 or 1000 times more energy than fission, so the fuel cost can be neglected. Nothing could be further from the truth.
The myth of unbounded / free energy from fusion comes from being able to use any old hydrogen atoms, rather than the much rarer deuterium and tritium.
Perhaps one day we'll get there, but I worry that the current advancements using the rarer isotopes will end up proving to be a dead end on that road, much like so many attempts at GAI. In the short term I suspect we'd have better odds with getting thorium reactors to be economical.
Tritium is rare but lithium isn't, and we can make tritium from lithium using the neutrons from fusion. (We also get tritium from fission plants, which is how we'd build the first fusion reactors.)
> we can make tritium from lithium using the neutrons from fusion
Each fusion reaction consumes one tritium atom and produces one neutron. If that neutron hits a lithium atom, it can split that and produce a tritium atom. If everything goes perfectly and there are no losses, then you get a 100% replacement of all the tritium that you consume. If you have a 90% replacement ratio (highly optimistic), you essentially lower the cost of your tritium fuel by a factor of 10, so from $30000 per gram to $3000 per gram, or $3 MM per kilogram.
> We also get tritium from fission plants
Yes we do. Mainly from Candu reactors. There are 49 Candu and Candu-like reactors in the world, and each produces less than 1kg of tritium per year. According to [1] a 1 GW fusion power plant would consume about 55 kg of tritium per year. So you'd need to run more than 50 fission power plants to operate one fusion power plant. Most people who dream of fusion think that fission will become irrelevant, not that you'll need 50 fission power plants for each fusion power plant.
That's why fusion blankets for D-T reactors use lead or beryllium as neutron multipliers. CFS for example uses FLiBe molten salt. Doing it this way a tokamak can not only sustain its own tritium supply, but periodically provide startup fuel for additional reactors.
Initial tritium load for a small, high-field reactor like CFS is much smaller than for ITER. And I'll note that the paper you linked has this conclusion:
> The preliminary results suggest that initial operation in D–D with continual feedback into the plasma of the tritium produced enables a fusion reactor designed solely for D–T operation to start-up in an acceptably short time-scale without the need for any external tritium source.
Ok, let's talk about that. For those who are not familiar, CFS stands for Commonwealth Fusion Systems, as startup with links to MIT. CFS aims to build a fusion reactor similar to ITER, but many times smaller, the secret sauce being that they use superconductors to achieve high magnetic fields. Back in 2022 some of the MIT guys got an ARPA-E grant to investigate the use of FLiBe to achieve atritium breeding ratio higher than 1 [1]. The results are in [2], they were published in January 2025. Here are some quotes:
> The long-term goal of LIBRA is to demonstrate a TBR ⩾ 1 in a large volume (1000 kg ∼ 500 l) of FLiBe molten salt using D–T neutron generators. Note that a full-scale LIB in an ARC-class FPP will require ∼250 000 l of FLiBe, hence the importance of understanding tritium behavior in large salt volumes.
ARC is the fusion reactor designed by CFS. This paper states that it will need 250000 liters of FLiBe. This is an insane amount. To understand how large this amount is, consider this: this ARPA-E project that took 3 years, used a quantity of 100 ml, so 0.1 liters.
Anyway, what breeding ratio was achieved? 3.57 x 10^(-4), or 0.0357%. It's a long way to go from here to 1.
I'm not saying it's impossible, but too many things related to fusion are just "engineering details".
Wow. The ARC reactor is supposed to deliver 270 MWe (when/if it will be built) [1]. So 20% of the world annual production of beryllium for one power plant that would deliver about one quarter of the power of an AP-1000 fission reactor.
Beryllium is very efficient as a neutron multiplier, but it is also extremely rare. It would not be acceptable as a consumable for energy production, as it is much more useful for other purposes.
In the Solar System, the abundance of beryllium is similar to that of gold and of the platinum-group metals. On Earth, the scarcity of beryllium is less obvious only because it is concentrated in the continental crust, where it is relatively easily accessible, even if its amount in the entire Earth is much smaller.
Lead neutron multipliers would be preferable, because they only inter-convert isotopes of lead, so it is not destroyed, like beryllium.
However lead used for this purpose becomes radioactive, with a very long lifetime, unless expensive isotope separation would be used for it.
I mean, it has to destroy the lead eventually, since lead is being used as a source of the extra neutrons. An individual lead nucleus will be converted to lighter lead isotopes by (n,2n) reactions, but eventually it will reach Pb-203 which decays to Tl-203. Presumably the thallium (and then mercury) will also be subject to (n,2n) reactions.
No, it comes from foolishly thinking that the cost of fuel will dominate cost of energy. That doesn't require fusion of protons; deuterium and lithium are cheap.
Agreed. I think fusion power would be great, but the sales pitch of 'limitless free power' just isn't true. The thought experiment I use is this: Let's imagine coal is magically free in every way. How does my power bill change? The answer is "barely at all" because the cost of utility electric power is mostly in distribution. We pay around 30c/kWh while the wholesale energy price is more like 2c/kWh.
It'll still make a difference in large scale energy intensive stuff, like desalination, aluminium refining, etc. but the average punter is going to save a lot more by installing solar panels.
We'll never know until (or if it ever comes) but there's reason to believe Fusion could be >50% cheaper than Fission.
That would still be more expensive than Solar and Wind (by 100% or more) - but I am skeptical in the same time frame those sources will be able to take over baseload generation.
It's really comparing apples to oranges.
Plus, it's a very hypothetical future. Anything could happen between now and then.
Because IMO the only approach that is even capable of delivering here is the Helion one (=> direct conversion). And that design is incredibly far from ready, the whole approach is completely unproven and their roadmap is mainly wishful self-delusion (from what we can tell by evaluating past milestones, like "first 50MW reactor finished by 2021"-- there is no 50MW reactor even now).
From my PoV, ITER-style tokamaks are the most conservative/certain design, and also the furthest along by far. That would imply:
=> Cryogenics for the magnets
=> big hightemperature vacuumchamber for plasma
=> all the thermal/turbogenerator infrastructure needed in conventional plants
=> super high neutron radiation flux (this is a problem)
I just don't see where you save anything. This is basically just a fission reactor, only a magnitude more complicated and demanding. I absolutely don't see how it could ever get significantly cheaper than conventional nuclear powerplants.
Fission reactor has to be big and has to deal with storage of a lot of nuclear waste and must implement a lot of expensive measures to stop runaway reaction in case of unexpected events.
Fusion has none of this. Assuming Q >> 1 will be demonstrated in a design that can be commercialized the next biggest problem is dealing with high-energy neurons on a scale never experienced before with potential much faster degradation of materials than anticipated leading to prohibiting operational costs.
That's a problem, but it's not necessarily even the biggest problem. Other huge problems include the shear size of the machines per MW of output (and hence cost per MW), coupled with their dreadful complexity and the difficulty of keeping them operating when they become too radioactive for hands-on maintenance. Designs typically just assume the reactors will be reliable enough, when there's no empirical evidence to support that (and the one study that tried to estimate uptime based on analogies with other technologies found the reactor would have an uptime percentage of just 4%!)
Helion's promised dates were conditioned on funding, which they didn't actually get for several years. Adjusting for when they did get funding, they're pretty much on track.
Without any of the meltdown concerns a fusion powerplant is a lot simpler to actually build than a fission plant. It has a small fraction of the security, reliability, regulatory, etc concerns (not none, just way way less). Unless it's so marginal that it's barely producing electricity I'd be pretty surprised to find out we had Q>1 fusion and yet it couldn't out compete fission anywhere fission is practical.
Modern fission designs mitigate meltdown concerns well enough that I'm not sure the safety & security around a fusion plant would actually be any better/cheaper, although public sentiment may be enough of an advantage. Tritium & neutron activated metals are dangerous enough to require keeping the traditional nuclear plant safeguards IMO. As far as proliferation concerns go, I don't see any reason you couldn't breed plutonium in the neutron flux of a fusion reactor, & the tritium is clearly viable for boosted warheads.
Modern fission designs plausibly mitigate meltdown concerns well enough...
To move that "plausibly" into "actually" you have to have very careful design review by regulators. Very careful review of construction to make sure what is constructed is what was designed. And so on and so forth. It's a lot of friction that skyrockets costs. Legitimately. People inevitably attempt to cut corners, and there's no way to make sure they aren't on the safety parts without checking. Actual currently regulatory costs seem to bear out the difference between these, with SMR people spending large amounts of money to convince regulators they didn't screw up, vs Helion fusion being "regulated like a hospital".
I'm not saying fusion has no proliferation concerns. But it's the difference between "low grade nuclear waste, or a very high tech very advanced program to weaponize a working reactor" and "even a broken reactor can be strapped to some explosives to make a dirty bomb". I can't say I'm very aware of how much proliferation concerns drive costs.
I was thinking more of large scale D-T fusion, e.g. the tokamak design, which requires breeding tritium & is expected to create a lot of neutron activated waste. The tritium is especially concerning, as it's roughly as deadly as polonium-210 & highly bioavailable in the form of super heavy water.
You're probably right for smaller aneutronic designs like Helion's. If they can actually be made to work, they'll be much safer.
That's astounding, I've never heard anybody claim that the reactors would be simpler before! Do you have any estimates of anybody working on the problem that thinks that?
Every schemer I have ever seen is quite a bit more complex than a fission reactor. Often, designs will depend on materials that do not yet exist.
That said there is a tremendous variety of techniques that fit under the umbrella term of "fusion," so I'm hoping to learn something more.
Not simpler in terms of technology, but simpler in terms of deployment, regulation, and security. Those are the majority of costs in fission power plants.
The majority of the cost in fission is in the massive construction build, change orders, logistics, massive concrete pours, welding, etc.
I've looked a lot into this in terms of how to get a project like Georgia's Vogtle to have cost less, or Olkioluoto in Finland, or Flamanville 3 in France. Big complex construction projects are expensive, and it's not clear at all to me that fusion would be simpler or smaller, or escape the rest of Baumol's cost disease that has been plaguing fission in highly developed economies.
The more plausible looking modern fusion companies tend to be designing very small reactors compared to those projects. Vogtle is 5000 MWs. Olkioluoto is 1600. Helion is promising reactors that are 50 and can be shipped via trains by 2028 (or 2030 depending on how you read some statements/interpret what I just said). They still need some neutron shielding to actually operate them safely (boronated concrete, probably not shipped by train), but nothing on the scale of what you need for a fission plant.
Helion also promised 50MW prototypes by 2021. It's 2025 and they have no 50MW prototype still. Fusion power roadmaps are generally exercises in wishful thinking, while fusion power startup roadmaps are basically gaslighting-as-a-serivce.
I still think its worth researching and we'll get there at some point, but I'm not holding my breath-- the whole industry has overpromised in the past, continues to overpromise now and will be probably be irrelevant for de-carbonizing the grid by the time the technology is actually ready at an industrial scale.
Mass media reporting on the whole sector is admittedly even worse; especially for uninformed readers without an engineering background.
I believe the promise you are referring to was contingent upon Helion getting funding that they did not get. It's not gaslighting to publish an amibitious timeline for a startup and be wrong, but that's not even what happened here. They said "if X (funding), then Y". X didn't happen.
I believe the current timetable is no longer contingent upon funding (since they've got the funding they think they need). It's no doubt still an optimistic startup timeline, a target, that they might well fail to achieve (even without the startup failing, just being late).
Meh. They promised 50MW by 2021 in 2018. In 2021, they got half a billion $, but the 50MW plant is still not running. But the bigger problem is that those are all just pretty prototypes; according to a 2018 ARPA report, magnetic field compression in the 40T range is needed for commercial viability. The various prototyes have pushed this from 4T to 10T (and presumably soonish 15T) since 2014. Extrapolating that trend is much less promising than Helions roadmaps, and doing linear extrapolation there is probably doing them a favor...
It's a cool concept, but probably not gonna be viable anytime soon (if ever!).
With regards to scaling, I think there are two components:
Does the physics change as they scale up the field strength? No one is really going to know until they try (unless we get a lot better at simulating plasma real fast). If not, they lost a bet, but they lost it honestly and as far as I can tell (not a physicist) it was a reasonably good bet to make.
Can they physically build the bigger magnets they need fast enough to meet their timelines (and everything else. I understand they are currently bottlenecked on capacitors)? Apart from normal "startups are overly optimistic" issues I don't see any reason to think that they shouldn't be able to reliably predict how fast they can scale magnet size, or be limited to a linear rate. While they are big magnets, it's not exactly new physics.
I'm not sure I'd say they are "probably going to be viable" anytime soon either. I think they have a good chance, but "probably" as in ">50%" is probably pushing it. (Also depends on where you put the goalposts of course)
FWIW I believe that 2018 report was for a high gain low pulse rate plan that Helion rejected, and they are aiming for substantially lower strength magnets as a result. I can't find anything more than rumors to confirm that though.
Just to be clear: I'm not accusing Helion of being dishonest or even fraudulent.
It's just that from everything I know about the project, they still have a long way to go, and there are a lot of milestones to hit that are just pipe dreams for now (actually fusing He3, breeding it, net-gain energy extraction, ...).
I would expect progress to slow down significantly as the scale of prototypes and their complexity increases (like what happens for basically every engineering project ever)-- but progress is already slow/behind schedule to begin with...
That’d be interesting to learn more about. What I’ve seen always leans toward regulation driving costs.
Though I guess some of that infrastructure could be overbuilt due to excessive regulation.
Also much of the concrete and steel is needed for the containment domes. Fusion power likely wouldn’t require nearly as much protection. Perhaps just a fairly standard industrial building.
People won't be afraid of fusion, fusion plants can't be used to make bombs, fusion plants could maybe explode, but they won't poison the nearby land (or the whole planet) for decades-eons.
Helion's reactor, if it works, could become a source of the cheapest neutrons on the planet. It would greatly enable nuclear proliferation by providing neutrons for breeding of fissionable material for bombs.
A 50 MW DD reactor would produce enough neutrons to make half a ton of plutonium per year. Remember, none of these neutrons have to be turned around to make tritium, as they would have to be in a DT reactor.
IMHO, dislike of masks is built into us as a social species that place significant value on facial expressions. Makes sense from an evolutionary game theory perspective for societies to discourage them.
There is a certain amount of "who cares about the cost" when it comes to fusion power. Nations will want to build them to lower or eliminate reliance on foreign energy, to address climate change concerns, and as a backup for renewables, and for other non-economic reasons. Many things that governments will want to fund that have nothing to do with directly "how much does the electricity cost?" or "when can we expect a return on investment?"
And the first generation will be expensive. That's how all new technology is.
There's definitely an existential question around if fusion will ever be able to beat renewables plus batteries, but who knows with our energy demands ever increasing at some point renewables may hit a breaking point in land cost.
I'm generally pro-publicly funded research. There is not any direct ROI on say the LHC, but it does fund advanced manufacturing and engineering work that might enable other more practical industrial applications. The ROI might be a century away.
> At the risk of coming off as a nay-sayer, let's say engineering hurtles related to fusion power generation is overcome. How is the presumably high upfront capital costs going to compare with the ROI?
I'm not sure if you're being serious, but I'm going to assume you are. Let's say energy costs 1/10th it does today. That's far cheaper than I see anybody predicting fusion will be, but I think renewables will get there. How much does cheap energy change in the economy? What is bottlenecked by expensive energy at the moment? It turns out that matter, people, people's wants, still have a huge impact.
Make all energy free. What does that change? It lowers operating costs for many things, but up front capital costs are still there. Land still matters. Food still matters.
Money will still matter. Allocation of time, of resources, all that still matters a lot. Energy is big for the economy, but if its free we shift our focus to other matters of logistics.
That is, it would seem likely that fusion power would be costly to build. It would also seem apparent that if it were to fulfil its promise then the power it generates is sold at or less than the current amount. That would then seem to imply a lengthily time to make a return on the initial investment. Or am I missing something else with this equation?