Marketing is the interface to the consumer and includes everything from deciding what you think customers will want and how to design the product to that to how you service the product in the field. It’s not a synonym for “bullshit” although marketing can be bullshit as much as any kind of investor communications or PR, either of which is a better description of this article.
Furthermore, if fusion ever works electricity will be too cheap to meter. So why sign a non-binding forward purchase agreement for something that if it works will cost $0?
which is like laser fusion but 10-100x more efficient in terms of wall plug to energy delivered to the pellet and also capable of multiple shots per second compare to laser systems that are doing good if they fire a few times a day.
Nobody is interested in developing it because the high energy requirement for the ion beams to transfer energy effectively into the pellet means you need multiple big accelerators so you can’t demonstrate it at less than a full scale power plant. Such a thing would cost about 1/10 of what a certain billionaire spent to buy a washed up social network.
Advanced particle accelerators are an area that could be proliferation sensitive and I think you might get some trouble if you tried to advance the technology. For instance a particle accelerator that drives a spallation neutron source could be used to make Plutonium with a smaller thermal footprint than a nuclear reactor. For that matter, a D+T fusion reactor produces copious very high energy neutrons which would be great for making Pu239 or U233.
For a data center operator that wants to maximize utilization and needs a lot of power, they might be willing to pay a premium for consistent on site power rather than dealing with intermittent sources or grid connection costs and battery backups.
That's where future markets come in. The producer won't get negative prices on shiny days. The consumer won't pay through the nose on rainy days. The marginal producer doesn't matter with a prenegotiated price.
This derisks both sides letting both of them build without worrying about the market moving against them and rendering their capital investments worthless.
> Furthermore, if fusion ever works electricity will be too cheap to meter.
Not to take this discussion in a different tangent, but I seriously doubt it. In CA, PG&E’s share of renewable keeps increasing, as do our rate. There’re a myriad of operating expenses as well as monopolistic rent seeking, and in this case IP development cost recovery that cost reduction may remain a pipe dream..
Oklo isn’t a fusion startup. They are planning to build small modular molten salt fission reactors aka next gen SMR fast reactors if you want the buzzwords.
Oklo's fission reactor is not molten salt based but is a liquid metal fast breeder reactor inspired by Argonne National Laboratory's old Experimental Breeder Reactor II, per their technology page:
Why do you think fusion will be cheaper than fission power? Fusion will, at least for a significant time, require investments in the fusion plants as well as maintenance and operations.
The only way I can see something like fusion become cheap quickly, is if all labor becomes cheap because of AI.
I think you're right, and I also wonder: Considering how hard it is for us to engineer a fusion plant that merely "breaks even" in terms of energy made usable to us, will fusion plants ever be competitive with fission plants? Like, I understand that hydrogen is a lot cheaper than plutonium or enriched uranium, but iirc we never had an issue getting fission plants to break even. As far as I'm aware, the technology for exploiting fission was obvious as soon as they realized that radioactive decay could be hot enough to boil water, and at risk of sounding glib, everything else about it is just safety measures. Whereas there is no fusion equivalent of "The Pile"[1] from Fermi et al.
I do expect fusion to some day be the dominant source of energy, at least if we intend to continue to increase our energy consumption, instead of going all Amish.
But that MAY be after we spent all the easily extractable uranium and (unfortunately) fossil fuels.
When looking at costs for almost anything, there tends to be 3 main components:
1) Raw materials for construction. (Edit: + energy consumed)
2) Raw materials for fuel/operations. (Edit: + energy consumed)
3) Labor
Fusion may require very significant amount of raw materials for construction, and also (with today's technology) a lot of labor. But it is nearly free in terms of fuel costs.
Now, let's assume (as a thought experiment) that AI will make labor virtually free within 100 years, we're down to fuel costs vs the materials needed for construction.
As long as we live in a world where we expect capital to yield some kind of significant/exponential "return on investment", we can calculate a "present day value" based on what we expect that return on investment to be.
But it's not a given that the assumption of exponential growth will hold true in the future Or rather, it's almost certain that SOME DAY it will cease to hold true. In particular, when we run out of other fuels (and we've covered Earth in solar panels), stagnation is inevitable. That means the natural interest is basically 0.
In such a world, and with the assumption that labor is free, fusion would give "free" energy, since capital costs no longer matter compared to the present day value of a permanent revenue stream.
So, given enough time, it seems clear that fusion will almost certainly take over as the dominant energy source. The question is when.
If our energy needs remain moderate for quite a long time or if we get a 300 year AI winter, it could take 100s of years for fusion to become dominant.
On the other hand, if AI makes labor virtually free 50 years from now (or makes labor only slightly more expensive than the energy used), and we continue to need more and more power to drive our data centers, manufacturing and every other need we may have, we may need it sooner.
Also, if we're not able to progress beyond barely breaking even in terms of energy produced vs energy consumed in the process, that will make progress slow.
But if we continue to make progress in terms of energy efficiency, and especially if we're able to scale fusion power to generate very large outputs (10's, 100s or 100s of GW), it may take over relatively quickly.
If I were to bet, I would guess it will take somewhere between 50-300 years for it to generate more than 50% of our power requirements.
"Too cheap to meter" means too cheap to meter, not free. Like residential internet — I pay for infrastructure, but my usage isn't metered. In some places water is that cheap, for example.
Fusion is an incredibly energy dense power source, but that does not necessarily make it cheap. e.g Uranium 235 is 3 million times more energy dense than coal, but nuclear power is not cheaper. And fusion is much closer in energy density to fission than fission is to coal.
It's more than a "yeah I am interested" email, or hoping that their costs will be lower than the spot price of electricity. To your point its less than a money down contract, but its not nothing.
The price of electricity has a high volatility, and having a guaranteed buyer at a given price can stabilize their business. Even if not guaranteed, it can make their business look more attractive to lenders/investors.
The point is it's not guaranteed at all, since it's a non binding letter of intent.
I don't see any point other than to give something to calm nervous investors or to get more investors.
> Without committed financial skin in the game, isn't this just a dog and pony show?
Kind of: IMHO it's to 'simplify' negotiations to certain extent.
Given that Oklo is a publicly traded company, and major financial deals may have material effects on stock, there are procedures for handling those deals as if some folks know about a (potential) deal but others do not, you can get into trouble with insider information.
By public announcing the negotiations and intent then the handling of information—especially with-in Oklo—can perhaps be more flexible and less siloed, which allows more folks to talk about what's going (both internally and externally). This way the negotiating team can (e.g.) pull in any random employee into a meeting to talk about (say) technical details about connections, power factors, support, without having to worry about 'internal embargoes'.
> GPT-6 might need about 10% of the world’s computers, a large power plant’s worth of energy ... Probably this looks like a town-sized data center attached to a lot of solar panels or a nuclear reactor.
> GPT-7 might need all of the world’s computers, a gargantuan power plant beyond any that currently exist ... Probably this looks like a city-sized data center attached to a fusion plant.
> they'll just need to use fear mongering to manufacture consent with the tax base
There is zero evidence this has occured. To the degree Altman et al's doomsday fundraising had regulatory effect, it hasn't been in the form of subsidies but onerous rulemaking.
You lied about something which isn't happening, the person who replied to you pointed out that you lied, and then you called that "a straw man argument" and quickly diverted to paragraphs of unrelated handwaving and character attacks.
I don't think you're a serious person who should be taken seriously on this topic.
You just made a bunch of claims without even citing what you claim I apparently lied about. Please cite where exactly I supposedly lied - and where the person I was replying to pointed out.
And what character attacks are you even talking about, who did I allegedly character attack - please cite my words for that too?
Since you commented on me, I'll comment on your reply that you seem to be very reactive - and I don't think you even understand the contents of what I wrote.
I also wonder if you're somehow biased in regards to why you have such a strong reaction?
If you're a "serious person" that you claim I'm not, you'll actually respond to each of my points to engage in an adult conversation - which will also show you're able to be calm enough to not be reactive such that you rush through to make claims without even referencing/citing specifics to support your claims.
> proving they can provide more value with competitive pricing
Prove to whom?
Since the availability of GPUs is quite constrained (availability of datacenters and electricity is constrained, too) you won't have many who "can provide more value". Free market has no mechanisms to combat abuse when resources are limited.
Prove to the end consumer - and hopefully not to industrial complexes that may somehow lobby politicians for regulatory capture to give the power to a handful vs. the consumer's "voting" power via their distributed-individual buying power.
I highly disagree that the free market has no mechanisms, exhibit A being Elon Musk who had the credibility to rally $44 billion to buy Twitter to take that social media platform out of the hands that the establishment clearly had as a core pillar for their censorship-suppression-narrative control apparatus, of which the Twitter Files exposed the government's illegal suppression of Americans et al for anyone who wasn't toeing the acceptable narrative talking points of the last 3+ years.
> I highly disagree that the free market has no mechanisms, exhibit A being Elon Musk
1. I didn't say "no mechanisms" in general. Re-read what I wrote
2. Elon Musk's acquisition of Twitter doesn't show anything you pretend it shows, it doesn't show the mechanisms I specifically talked about, and since you jumped off of a cliff into deep conspiracy theories, I don't even care what you think it shows.
I understood what you wrote. Your logic and interpretation is off, why you don't understand what I said.
Thanks for the ad hominem and being too lazy to engage except to tell us that indirectly, clearly "not worth my time" either - so let's say it's mutual - if you're arrogant enough to think the unreferenced "conspiracy theories" I've brought up have no truth or validity to them - you must clearly know better, especially with how much effort you put into things. /sarcasm
And if you're referencing the Twitter Files "conspiracy theories" - here you go, enjoy, but I doubt you have the patience to actually "dive" into it, or even bother to lookup the US senate hearings on the issue:
https://twitterfiles.substack.com/ - Matt Taibbi's articles on it, one of the journalists Elon Musk gave access to Twitter's data including their staff communications/emails with the FBI et al.
I realize this headline means nothing, but there is something to nuclear power futures. I live in Northern Virginia and my recent 20 mile bike ride in Loudon county was nothing but massive datacenters and crisscrossing powerlines to support them. The western side of neighboring Prince William County is now being consumed for data centers. The Washington Post recently had an article on how the demand for electricity for these centers is getting so desperate that they are running powerlines deep into West Virginia and bringing coal plants back online to meet the demand. Estimates are that it's going to take multiple nuclear powerplants to meet future demand in just this locality.
I have very mixed feelings about this. I love the possibilities of all these datacenters and their potential, but I'm worried about the massive amount of energy they are consuming. If this is all just so much hype and marketing, then it's incredibly wasteful.
So your solution is to power these datacenters for another 1-2 decades with coal power while you wait for new nuclear plants? Because that's the realistic timeframe for any new nuclear. Vogtle 3 took 14 years to build.
Of course, any "advanced nuclear" that doesn't even have a prototype yet will take longer than that. If it will ever happen.
Don't know why you're saying "your solution" like that guy is in charge of us energy policy, but Virginia already has nuclear reactors. And Dominion has the permits and plans and facility to construct another 1500 megawatts.
How many times in history have we regretted over-building this kind of infrastructure? We have no shortage of good ways to use electric energy. If AI turns out to be a fad, others will benefit.
Pollution is a concern, but I think don't think that anti-growth or anti-consumption sentiments are useful. All this money pouring in is an opportunity to invest in clean energy.
> my recent 20 mile bike ride in Loudon county was nothing but massive datacenters and crisscrossing powerlines to support them. The western side of neighboring Prince William County is now being consumed for data centers.
Is this actually true or hyperbole? Ireland has the same size and population as Virginia is described as a major location for data centres, but all the data centres on the island combined take up less than one square km [0]. By comparison, Dublin Airport is 10 square km.
Sounds like Virginia needs to be two orders of magnitude more dense in DCs, or have them developed exactly in a ribbon along your route, to really dominate the landscape like that.
It's hyperbolic of course, but there is an ongoing NIMBY battle over the next datacenter. And it is a huge datacenter, dont get me wrong. The company bidding recently said it would convert a shitton of the land it would be given into greenspaces that yon biker could bike through. And we can be clear here too, if you can afford to bike in Loudon then you work for one of the companies directly requiring this new datacenter, so not sure why he's "conflicted" (nimbyism)
I don't follow your reasoning. It costs nothing to ride a bicycle in most of America. We have bike paths everywhere that are free for anyone to enjoy. You also don't have to live in a locality to enjoy the bike paths there.
For the record, on this specific outing I was doing volunteer trail-surveying. I live in Prince William County. I think the Data Centers are awesome and I love them, the tax revenues they bring in, and the jobs that come with them. It was clear from my comment that I am conflicted about the incredible energy-consumption, not the data centers themselves.
What does Ireland having the same size and population as Virginia have to do with my comment? Virginia is a state and I was specifically referring to Loudon and Prince William Counties, which are very tiny in proportion to the entire state of Virginia. Look up "Loudon County Data Center Alley" to understand the proportions and density of this project.
Considering that they’d filed the first application for a non-LWR in decades I don’t think they did that bad. Still I thought it was a waste that they went on and on about the risk of supervolcanos and never said what they would do if the sodium coolant caught on fire. (Not that hard to answer, necessarily as you can detect it, drain water if is involved, and put the fire out…. But you’ve got to have some answer for an SFR.)
Gosh am I annoyed by these bullshit energy startups.
They want to sell a PPA? For a technology for which they don't even have a working prototype? That has never produced a single kwh of energy?
That's like... step 200 before step 0.5.
Both AI data centers and new-generation nuclear tech are prestige investments so arguing about need and appropriateness is beside the point. Paying large premium to have a carbon neutral AI is going to go unquestioned.
But there is still a potential snag: Is it insurable?
Slightly off-topic, but what is more profitable: energy generation (power plants) or electricity distribution (the grid)? Would love to hear from those who watch this space.
I think it’s typically an auction system so generally speaking you expect the plants with the cheapest energy to undercut the rest. Off the top of my head I’m not sure how concerns of collusion are dealt with when there’s few participants. There’s probably some sort of regulatory stick in there.
I was under impression that the grid operators would have more leverage over power plants when it came to setting prices, after all, if you cannot send your output t the customers, you are sort of stuck with it.
65 comments
[ 2.8 ms ] story [ 116 ms ] threadDoes this mean anything at all? Sounds like the equivalent of "cool story, bro" from the signing party.
Without committed financial skin in the game, isn't this just a dog and pony show?
https://en.wikipedia.org/wiki/Heavy_ion_fusion
which is like laser fusion but 10-100x more efficient in terms of wall plug to energy delivered to the pellet and also capable of multiple shots per second compare to laser systems that are doing good if they fire a few times a day.
Nobody is interested in developing it because the high energy requirement for the ion beams to transfer energy effectively into the pellet means you need multiple big accelerators so you can’t demonstrate it at less than a full scale power plant. Such a thing would cost about 1/10 of what a certain billionaire spent to buy a washed up social network.
Advanced particle accelerators are an area that could be proliferation sensitive and I think you might get some trouble if you tried to advance the technology. For instance a particle accelerator that drives a spallation neutron source could be used to make Plutonium with a smaller thermal footprint than a nuclear reactor. For that matter, a D+T fusion reactor produces copious very high energy neutrons which would be great for making Pu239 or U233.
No need to wait for some future power source.
The revenue potential for any nuclear reactor in the age of renewables is going to be incredibly slim.
During all other time renewables will be the marginal producer which means renewable costs for the electricity.
This derisks both sides letting both of them build without worrying about the market moving against them and rendering their capital investments worthless.
Not to take this discussion in a different tangent, but I seriously doubt it. In CA, PG&E’s share of renewable keeps increasing, as do our rate. There’re a myriad of operating expenses as well as monopolistic rent seeking, and in this case IP development cost recovery that cost reduction may remain a pipe dream..
https://oklo.com/technology/
https://en.wikipedia.org/wiki/Experimental_Breeder_Reactor_I...
The only way I can see something like fusion become cheap quickly, is if all labor becomes cheap because of AI.
[1] https://en.wikipedia.org/wiki/Chicago_Pile-1
But that MAY be after we spent all the easily extractable uranium and (unfortunately) fossil fuels.
When looking at costs for almost anything, there tends to be 3 main components:
1) Raw materials for construction. (Edit: + energy consumed)
2) Raw materials for fuel/operations. (Edit: + energy consumed)
3) Labor
Fusion may require very significant amount of raw materials for construction, and also (with today's technology) a lot of labor. But it is nearly free in terms of fuel costs.
Now, let's assume (as a thought experiment) that AI will make labor virtually free within 100 years, we're down to fuel costs vs the materials needed for construction.
As long as we live in a world where we expect capital to yield some kind of significant/exponential "return on investment", we can calculate a "present day value" based on what we expect that return on investment to be.
But it's not a given that the assumption of exponential growth will hold true in the future Or rather, it's almost certain that SOME DAY it will cease to hold true. In particular, when we run out of other fuels (and we've covered Earth in solar panels), stagnation is inevitable. That means the natural interest is basically 0.
In such a world, and with the assumption that labor is free, fusion would give "free" energy, since capital costs no longer matter compared to the present day value of a permanent revenue stream.
So, given enough time, it seems clear that fusion will almost certainly take over as the dominant energy source. The question is when.
If our energy needs remain moderate for quite a long time or if we get a 300 year AI winter, it could take 100s of years for fusion to become dominant.
On the other hand, if AI makes labor virtually free 50 years from now (or makes labor only slightly more expensive than the energy used), and we continue to need more and more power to drive our data centers, manufacturing and every other need we may have, we may need it sooner.
Also, if we're not able to progress beyond barely breaking even in terms of energy produced vs energy consumed in the process, that will make progress slow.
But if we continue to make progress in terms of energy efficiency, and especially if we're able to scale fusion power to generate very large outputs (10's, 100s or 100s of GW), it may take over relatively quickly.
If I were to bet, I would guess it will take somewhere between 50-300 years for it to generate more than 50% of our power requirements.
Anybody who says this about any technology isn't thinking about infrastructure costs.
The price of electricity has a high volatility, and having a guaranteed buyer at a given price can stabilize their business. Even if not guaranteed, it can make their business look more attractive to lenders/investors.
Kind of: IMHO it's to 'simplify' negotiations to certain extent.
Given that Oklo is a publicly traded company, and major financial deals may have material effects on stock, there are procedures for handling those deals as if some folks know about a (potential) deal but others do not, you can get into trouble with insider information.
By public announcing the negotiations and intent then the handling of information—especially with-in Oklo—can perhaps be more flexible and less siloed, which allows more folks to talk about what's going (both internally and externally). This way the negotiating team can (e.g.) pull in any random employee into a meeting to talk about (say) technical details about connections, power factors, support, without having to worry about 'internal embargoes'.
> GPT-7 might need all of the world’s computers, a gargantuan power plant beyond any that currently exist ... Probably this looks like a city-sized data center attached to a fusion plant.
Snippet from Astral Codex Ten's article "Sam Altman Wants $7 Trillion".
https://www.astralcodexten.com/p/sam-altman-wants-7-trillion
I guess $10 trillion is the next B.S. number you have to throw out to get PR points.
There is zero evidence this has occured. To the degree Altman et al's doomsday fundraising had regulatory effect, it hasn't been in the form of subsidies but onerous rulemaking.
I don't think you're a serious person who should be taken seriously on this topic.
You just made a bunch of claims without even citing what you claim I apparently lied about. Please cite where exactly I supposedly lied - and where the person I was replying to pointed out.
And what character attacks are you even talking about, who did I allegedly character attack - please cite my words for that too?
Since you commented on me, I'll comment on your reply that you seem to be very reactive - and I don't think you even understand the contents of what I wrote.
I also wonder if you're somehow biased in regards to why you have such a strong reaction?
If you're a "serious person" that you claim I'm not, you'll actually respond to each of my points to engage in an adult conversation - which will also show you're able to be calm enough to not be reactive such that you rush through to make claims without even referencing/citing specifics to support your claims.
Prove to whom?
Since the availability of GPUs is quite constrained (availability of datacenters and electricity is constrained, too) you won't have many who "can provide more value". Free market has no mechanisms to combat abuse when resources are limited.
I highly disagree that the free market has no mechanisms, exhibit A being Elon Musk who had the credibility to rally $44 billion to buy Twitter to take that social media platform out of the hands that the establishment clearly had as a core pillar for their censorship-suppression-narrative control apparatus, of which the Twitter Files exposed the government's illegal suppression of Americans et al for anyone who wasn't toeing the acceptable narrative talking points of the last 3+ years.
The conspiracy theories abound in this thread.
Attack the arguments and not the person by indirectly calling me a conspiracy theorist.
Are you even aware of the Twitter Files I referenced, the specifics for what I'm referencing?
Here's Matt Taibbi's specific work on the Twitter Files: https://twitterfiles.substack.com/
1. I didn't say "no mechanisms" in general. Re-read what I wrote
2. Elon Musk's acquisition of Twitter doesn't show anything you pretend it shows, it doesn't show the mechanisms I specifically talked about, and since you jumped off of a cliff into deep conspiracy theories, I don't even care what you think it shows.
Adieu.
Thanks for the ad hominem and being too lazy to engage except to tell us that indirectly, clearly "not worth my time" either - so let's say it's mutual - if you're arrogant enough to think the unreferenced "conspiracy theories" I've brought up have no truth or validity to them - you must clearly know better, especially with how much effort you put into things. /sarcasm
And if you're referencing the Twitter Files "conspiracy theories" - here you go, enjoy, but I doubt you have the patience to actually "dive" into it, or even bother to lookup the US senate hearings on the issue:
https://twitterfiles.substack.com/ - Matt Taibbi's articles on it, one of the journalists Elon Musk gave access to Twitter's data including their staff communications/emails with the FBI et al.
I have very mixed feelings about this. I love the possibilities of all these datacenters and their potential, but I'm worried about the massive amount of energy they are consuming. If this is all just so much hype and marketing, then it's incredibly wasteful.
Of course, any "advanced nuclear" that doesn't even have a prototype yet will take longer than that. If it will ever happen.
Pollution is a concern, but I think don't think that anti-growth or anti-consumption sentiments are useful. All this money pouring in is an opportunity to invest in clean energy.
Is this actually true or hyperbole? Ireland has the same size and population as Virginia is described as a major location for data centres, but all the data centres on the island combined take up less than one square km [0]. By comparison, Dublin Airport is 10 square km.
Sounds like Virginia needs to be two orders of magnitude more dense in DCs, or have them developed exactly in a ribbon along your route, to really dominate the landscape like that.
[0] https://baxtel.com/data-center/republic-of-ireland
I don't follow your reasoning. It costs nothing to ride a bicycle in most of America. We have bike paths everywhere that are free for anyone to enjoy. You also don't have to live in a locality to enjoy the bike paths there.
For the record, on this specific outing I was doing volunteer trail-surveying. I live in Prince William County. I think the Data Centers are awesome and I love them, the tax revenues they bring in, and the jobs that come with them. It was clear from my comment that I am conflicted about the incredible energy-consumption, not the data centers themselves.
https://biz.loudoun.gov/datacenteralley/
https://www.nrc.gov/reactors/new-reactors/advanced/who-were-...
https://www.nrc.gov/reading-rm/doc-collections/news/2022/22-...
They want to sell a PPA? For a technology for which they don't even have a working prototype? That has never produced a single kwh of energy? That's like... step 200 before step 0.5.
These companies tend to try survive on public money and grants while not delivering anything.
Then when the last deadline finally passes the project is cancelled like what happened with NuScale.
But there is still a potential snag: Is it insurable?
A solar plant is very profitable to run right now. Probably that.