Do they even have the design of an actually working fusion reactor? This seems like a crucial detail that is suspiciously ommitted.
"but said it remains on track to deliver power by 2028" - so casually written! I HIGHLY doubt this.
AIPedant•2h ago
No, of course they don't. This is Sam Altman's fusion company, backed by Microsoft in 2023 with a signed power purchase agreement: either Altman has some serious dirt on Satya Nadella, or (more likely!) Satya Nadella is a gullible idiot who thought "Sam Altman is the Boy Genius Who Invented AI, so he can solve fusion too!"
(Remember this is same same Satya Nadella who offered Altman an unspecified CEO-level position after he got fired... while publicly admitting he didn't know why Altman was fired! If I was a MSFT investor I would be pretty upset about this.)
Presumably 2025 MSFT is more sober-minded about Altman. I wonder if they're gonna try to wiggle their way out of the PPA. Otherwise I am truly baffled.
trhway•1h ago
>Presumably 2025 MSFT is more sober-minded about Altman.
If nothing else, that, after all the good they had done to him, should do it:
"OpenAi v. Microsoft: Altman ready to sue for unfair competition
...
OpenAI has put a potentially devastating weapon on the table: accusing Microsoft of anti-competitive practices and raising the attention of the Federal Trade Commission. It would be a low blow...
"
Ugh. Didn’t realize this was Altman. That’s disappointing.
Seemed like there was a certain amount of magic thinking about neutron damage but a bit less than fission typically does. Guess we’ll see.
paddw•1h ago
Presumably, if it's just a per-agreement to purchase power, there's no downside in the likely case the project implodes.
AIPedant•53m ago
The problem is that the agreement does not seem to specify that the electricity has to come from fusion![1] This is actually common in renewable PPAs - if the specific project doesn't come online then the provider has to find an alternative. But those are usually done with established providers which have > 0MW overall capacity. Helion does not. If the PPA is fixed amount vs fixed price, Microsoft might end up on the hook for inflated wholesale prices instead of cheap fusion.
FWIW I agree with the author of that Data Center Dynamics post, it's quite likely that MSFT and Helion are essentially in cahoots by stoking investors with vaporware. But it also seems like Altman might have sold Nadella 50MW of magic beans.
Investors gets ruined on PPAs all the time. With renewables, there's usually some sort of wind turbine or solar park attached, so after declaring bankrupcy the new owners can settle out of court and continue production.
Here there's realistically no way to continue electricity production so the assets will likely be chopped up and sold, if possible. Nobody owns anyone anything and Microsoft doesn't have to pay a dime. They won't get their power of course, but there's no downside for them.
These types articles where a PPA contract is confused with an investment is really common, mostly for nuclear and renewables, but that doesn't make them any more true. Microsoft hasn't invested anything, likely because they know this is (pardon the pun) hot air.
dehrmann•13m ago
> either Altman...
Another option is this was to sweeten the pot during OpenAI negotiations.
cornholio•2h ago
I'm not aware of Helion publishing any peer reviewed data claiming a physics breakeven (the point where the total energy generated by the reactor exceeds the external energy fed in to maintain the reaction going); let alone an engineering breakeven (the point where the fission generates about an order of magnitude more energy, to allow for the energy conversion losses, cooling and fuel breeding etc. so as to actually output any useful amount of electricity); let alone an economic breakeven, where the reactor generates sufficient useful energy that its market price can allow the capital and operating costs of the reactor and associated infrastructure to be recovered in a certain number of decades.
If fusion had all three today, it would still e a though sell; fission has them and is still failing economically.
staplung•1h ago
I don't disagree with any of what you say but if Helion's approach works (and that's a huge if) it would generate electricity directly, without need for a steam turbine or any of the associated plumbing. My understanding is that a big part of the cost for fission is the turbine etc.
hinkley•1h ago
And danger. Turbines are more powerful at high temps, and now you have hot liquids near your reactor. Or you use molten salt as a middleman so the potential steam explosions are a little farther from the reactor.
captain_coffee•1h ago
And how would you "generate electricity directly", specifically? Let's talk physics and engineering, not vague statements.
How would that energy generated from nuclear fusion be transformed into electricity "directly"? By which process / series of processes?
cyberax•1h ago
> And how would you "generate electricity directly", specifically?
Using a particle accelerator (decelerator?) in reverse. I'm an investor in Tri-Alpha Energy, and they have tested a direct converter with the claimed 90+% efficiency.
Animats•56m ago
See "Induction systems"[1] The concept was proposed in 1963, but nobody ever made it work.
That's the plan. Magnetohydrodynamic generators [2] do work, but they have electrodes in the gas. That works for jet engine type MHD generators, but fusion plasma is too hot for any solid material.
What they're trying to do is known physics but very hard engineering.
They're also trying for aneutronic fusion, using helium-3. If the plant generates large volumes of neutrons, it chews up the first wall between the reaction and the outside. It also
makes what it hits radioactive, so there's a waste problem. Aneutronic fusion uses reactions that (mostly) generate alphas and betas. This is, again, known physics but very hard engineering.
If they can get a demo machine going which solves either problem, that would be a huge advance.
So far, that does not seem to have happened.
There are other startups in this space. It's probably the way commercial fusion power will eventually be done. Not via the tokamaks, like ITER.
> There are other startups in this space. It's probably the way commercial fusion power will eventually be done. Not via the tokamaks, like ITER.
There is literally no evidence to suggest this: Helion are making big claims but as noted have shown little incremental progress on their machines.
The balance of history says if it happens it'll come out of a large government funded project: that's how fission happened, and there's plain old fission startups too who also are yet to deliver anything and we know fission works.
thinkcontext•56m ago
From their Wikipedia page
> Energy is captured by direct energy conversion that uses the expansion of the plasma to induce a current in the magnetic compression- and acceleration – coils. Separately it translates high-energy fusion products, such as alpha particles directly into a voltage. 3He produced by D–D fusion carries 0.82 MeV of energy. Tritium byproducts carry 1.01 MeV, while the proton produces 3.02 MeV.
> This approach eliminates the need for steam turbines, cooling towers, and their associated energy losses. According to the company, this process also allows the recovery of a significant part of the input energy at a round-trip efficiency of over 95%.
https://en.wikipedia.org/wiki/Magnetohydrodynamic_generator is the gist of it. If you have enough conductive plasma then just moving it through a magnetic field generates a current. Applied to fusion power, you heat the plasma through the fusion reaction then divert part of the plasma through the MHD generator.
Tbh, I very much doubt that this is a realistic path in the coming decade (but would love to be proven wrong). AFAIK no experimental reactor has yet generated any net electrical power at all, let alone with any big (ie dozens to hundreds of) MW MHD generators. Getting even one of these aspects working would be a major advance, let alone doing both at once.
Sure, they aim to extract energy directly from the field, but the three breakeven points are still important. A significant part of the energy will be lost as x rays and neutrons, since their D-H3 fuel cycle is not aneutronic; they will also have significant D-D reactions that are required to breed Tritium which they capture and then let ti decay to Helium-3.
Overall, when you look at the total complexity and energy balance of the full reactor + fueling cycle, maintaining vacuum, keeping superconducting magnets at cryogenic temperatures, tritium extraction etc. then generating an order of magnitude more energy than inserted still seems necessary to achieve engineering breakeven.
fcpguru•1h ago
i think the idea is chatgpt-5 (or maybe 6) will very quickly solve the 3 main blockers:
1. Plasma Stability & Control
2. Neutron-Resistant Materials
3. Tritium Breeding
4. Heat Exhaust & Divertor Design
Because:
Trained on terabytes of tokamak operation data. Will Ssggest new coil configurations, feedback loops, and magnetic geometries in minutes instead of months. LLMs can read the entire materials science literature, cross-reference neutron scattering data, and propose new alloys or composites. ML models can simulate atom-level neutron impacts in hours, not months, narrowing the search space. Use reinforcement learning to optimize lithium arrangement, coolant flows, and neutron multipliers. LLMs can generate and evaluate hundreds of engineering CAD designs in parallel. Predict tritium production efficiency before we build the prototype. LLMs + physics-informed ML can propose thousands of divertor designs that maximize heat spread, self-cooling, or vapor shielding. Suggest novel coolant chemistries based on prior patents and obscure literature.
TLDR: Altman knows how well LLMs are getting at physics enough to be sure fusion will be solved very soon.
salynchnew•1h ago
Wow. A lot to unpack here. "Propose new alloys or composites" is certainly quite a phrase, given the context.
Has an LLM ever generated any complex CAD design like this that has been built & worked?
captain_coffee•1h ago
Not sure if that reply is intentional trolling or schizophrenia to be honest.
My experience using them to make OpenSCAD code is marginally better than you may expect from @simonw's pellican-on-a-bike challenge, but only marginally.
Animats•1h ago
They've been through a few generations of test machines. They have something called "Polaris". It was supposed to be finished around the end of 2024.[1]
Their own site still talks of it as being under construction.
Discussion on Reddit.[2]
They previously built something called "Trenta".[3] That generates two balls of plasma and fires them at each other. There's no fusion or fusion fuel. It's a test rig for plasma generation and manipulation. That was running two years ago.
"Polaris" is a scale-up of Trenta, with something to fuse, and with energy recovery. It's very unclear how far that project has progressed. If they were getting energy out, that would be big news.
Helion is vague about how that's progressing.
> If they were getting energy out, that would be big news.
That's rather underselling it. If they have a proven, working, commercially viable design for a fusion power plant, they could probably just write a paper about how it works and collect their Nobel prize for physics next year.
creato•41m ago
I'm ambivalent about whether their design can work but if they were confident in their design and have the necessary funding, a paper and a nobel prize are going to be very far down their priority list.
trhway•1h ago
with power hungry AI datacenters popping up like mushrooms after a rain the timing couldn't be better for fusion. I guess VCs see that too. Well, some VCs also paid for that back then https://en.wikipedia.org/wiki/Rotary_Rocket - single stage half-helicopter half-rocket to orbit (and $30M in VC funding 30 years ago were really huge money) - despite 6th grade math.
captain_coffee•2h ago
AIPedant•2h ago
(Remember this is same same Satya Nadella who offered Altman an unspecified CEO-level position after he got fired... while publicly admitting he didn't know why Altman was fired! If I was a MSFT investor I would be pretty upset about this.)
Presumably 2025 MSFT is more sober-minded about Altman. I wonder if they're gonna try to wiggle their way out of the PPA. Otherwise I am truly baffled.
trhway•1h ago
If nothing else, that, after all the good they had done to him, should do it:
"OpenAi v. Microsoft: Altman ready to sue for unfair competition
...
OpenAI has put a potentially devastating weapon on the table: accusing Microsoft of anti-competitive practices and raising the attention of the Federal Trade Commission. It would be a low blow... "
https://en.ilsole24ore.com/art/openai-v-microsoft-altman-rea...
hinkley•1h ago
Seemed like there was a certain amount of magic thinking about neutron damage but a bit less than fission typically does. Guess we’ll see.
paddw•1h ago
AIPedant•53m ago
FWIW I agree with the author of that Data Center Dynamics post, it's quite likely that MSFT and Helion are essentially in cahoots by stoking investors with vaporware. But it also seems like Altman might have sold Nadella 50MW of magic beans.
[1] https://www.datacenterdynamics.com/en/opinions/microsoft-and... This is second-hand, the agreement is not public.
xorcist•2m ago
Here there's realistically no way to continue electricity production so the assets will likely be chopped up and sold, if possible. Nobody owns anyone anything and Microsoft doesn't have to pay a dime. They won't get their power of course, but there's no downside for them.
These types articles where a PPA contract is confused with an investment is really common, mostly for nuclear and renewables, but that doesn't make them any more true. Microsoft hasn't invested anything, likely because they know this is (pardon the pun) hot air.
dehrmann•13m ago
Another option is this was to sweeten the pot during OpenAI negotiations.
cornholio•2h ago
If fusion had all three today, it would still e a though sell; fission has them and is still failing economically.
staplung•1h ago
hinkley•1h ago
captain_coffee•1h ago
How would that energy generated from nuclear fusion be transformed into electricity "directly"? By which process / series of processes?
cyberax•1h ago
Using a particle accelerator (decelerator?) in reverse. I'm an investor in Tri-Alpha Energy, and they have tested a direct converter with the claimed 90+% efficiency.
Animats•56m ago
What they're trying to do is known physics but very hard engineering.
They're also trying for aneutronic fusion, using helium-3. If the plant generates large volumes of neutrons, it chews up the first wall between the reaction and the outside. It also makes what it hits radioactive, so there's a waste problem. Aneutronic fusion uses reactions that (mostly) generate alphas and betas. This is, again, known physics but very hard engineering.
If they can get a demo machine going which solves either problem, that would be a huge advance. So far, that does not seem to have happened.
There are other startups in this space. It's probably the way commercial fusion power will eventually be done. Not via the tokamaks, like ITER.
[1] https://en.wikipedia.org/wiki/Direct_energy_conversion
[2] https://en.wikipedia.org/wiki/Magnetohydrodynamic_generator
[3] https://spectrum.ieee.org/aneutronic-fusion
XorNot•15m ago
There is literally no evidence to suggest this: Helion are making big claims but as noted have shown little incremental progress on their machines.
The balance of history says if it happens it'll come out of a large government funded project: that's how fission happened, and there's plain old fission startups too who also are yet to deliver anything and we know fission works.
thinkcontext•56m ago
> Energy is captured by direct energy conversion that uses the expansion of the plasma to induce a current in the magnetic compression- and acceleration – coils. Separately it translates high-energy fusion products, such as alpha particles directly into a voltage. 3He produced by D–D fusion carries 0.82 MeV of energy. Tritium byproducts carry 1.01 MeV, while the proton produces 3.02 MeV.
> This approach eliminates the need for steam turbines, cooling towers, and their associated energy losses. According to the company, this process also allows the recovery of a significant part of the input energy at a round-trip efficiency of over 95%.
https://en.m.wikipedia.org/wiki/Helion_Energy
WJW•51m ago
Tbh, I very much doubt that this is a realistic path in the coming decade (but would love to be proven wrong). AFAIK no experimental reactor has yet generated any net electrical power at all, let alone with any big (ie dozens to hundreds of) MW MHD generators. Getting even one of these aspects working would be a major advance, let alone doing both at once.
jcims•23m ago
cornholio•13m ago
Overall, when you look at the total complexity and energy balance of the full reactor + fueling cycle, maintaining vacuum, keeping superconducting magnets at cryogenic temperatures, tritium extraction etc. then generating an order of magnitude more energy than inserted still seems necessary to achieve engineering breakeven.
fcpguru•1h ago
1. Plasma Stability & Control
2. Neutron-Resistant Materials
3. Tritium Breeding
4. Heat Exhaust & Divertor Design
Because:
Trained on terabytes of tokamak operation data. Will Ssggest new coil configurations, feedback loops, and magnetic geometries in minutes instead of months. LLMs can read the entire materials science literature, cross-reference neutron scattering data, and propose new alloys or composites. ML models can simulate atom-level neutron impacts in hours, not months, narrowing the search space. Use reinforcement learning to optimize lithium arrangement, coolant flows, and neutron multipliers. LLMs can generate and evaluate hundreds of engineering CAD designs in parallel. Predict tritium production efficiency before we build the prototype. LLMs + physics-informed ML can propose thousands of divertor designs that maximize heat spread, self-cooling, or vapor shielding. Suggest novel coolant chemistries based on prior patents and obscure literature.
TLDR: Altman knows how well LLMs are getting at physics enough to be sure fusion will be solved very soon.
salynchnew•1h ago
Has an LLM ever generated any complex CAD design like this that has been built & worked?
captain_coffee•1h ago
fcpguru•1h ago
ben_w•1h ago
Animats•1h ago
Discussion on Reddit.[2]
They previously built something called "Trenta".[3] That generates two balls of plasma and fires them at each other. There's no fusion or fusion fuel. It's a test rig for plasma generation and manipulation. That was running two years ago.
"Polaris" is a scale-up of Trenta, with something to fuse, and with energy recovery. It's very unclear how far that project has progressed. If they were getting energy out, that would be big news. Helion is vague about how that's progressing.
[1] https://www.helionenergy.com/polaris/
[2] https://www.reddit.com/r/fusion/comments/1hlojqu/any_news_on...
[3] https://www.youtube.com/watch?v=_bDXXWQxK38
WJW•1h ago
That's rather underselling it. If they have a proven, working, commercially viable design for a fusion power plant, they could probably just write a paper about how it works and collect their Nobel prize for physics next year.
creato•41m ago