Gas pipeline players in talks to fuel AI datacenter demand Proximity to natural gas lines could become just as desirable for datacenter operators as high-speed fiber-optic networks as they scramble to satiate AI's ever growing thirst for power. Speaking to analysts during their respective earnings call this week, executives at Energy Transfer LP and Williams Companies, both of which …
There is so much hype these days of crazily-energy-consumptive computer-applications-of-dubious-value (blockchain-based quasi-currencies, non-fungible tokens, and now generative AI/machine learning) that one almost might suspect nVIDIA, AMD, Intel, et. al., are being given kickbacks from the energy companies to invent and promote this stuff.
The hardware for this just feels so kludgey and inelegant -- in the way of rowing teams being issued new oars with one-square-meter blades, so that they can "go faster".
I couldn't agree more.
Maybe I'm not a "power user" of LLM's, but the notion that we need excessive power provision and consumption to provide answers to, and come on let's be honest here, who are the majority of the users? I would bet it's the average joe/layperson with answers to their school homework/manager's latest hairbrained scheme, seems disingenuous at best, and a down right con at the other end of the scale.
Whilst some of the responses I've had from LLMs have been interesting, most of the time I'm finding it is not providing the responses I need to conundrum's that I pose to the LLM for consideration, so for myself, I'm generally not seeing the huge benefit that I'm being led to believe AI/LLMs are providing.
I spend an equal amount of time delving into the responses, which aren't much better than an internet search IMHO.
Still, as I said, maybe I'm not the best user case!
The power companies. Altman owns one, too.
AI is now the biggest con the World has ever known. Kill it with fire. Stop this LLM nonsense. It is a dead end. Shift now to low-energy r/t neurual networks. Big costs upfront but pennies thereafter.
Will somebody actually think of not the planet, cause it doesn't really give a pooh about us, but of the clever monkeys put here to fathom how to reverse entropy (probably)? Us. This collection of humans alive at the same time. One massive family in time.
SI. Coming soon. You have seen bits of it. Reg will get it first, as always.
Enjoy the sun today for those of us in God's own land of Blighty and I'm locked and loaded to give a shout out to all those n's going to Margate for a swim in the sea whilst it is still possible. F AI and switch all your devices off - except for the Watch and touch SAND and sea. Luvs
I love the LLMs. They are going to revolutionise the way we interact with information forever. But can’t they build mini nuclear plants rather than using gas? Seems like building more fossil powered stations is going the wrong way. Maybe the tech is not ready for prime time yet but I remember reading about a few companies developing mini plants.
"But can’t they build mini nuclear plants rather than using gas?"
Nuclear power inspires fear for a disproportionate number of people (Case study: Germany, other nations have followed). The operators are also on the hook for vast potential liabilities, and to plan, build and get online they're subject to the crushing weight of civil nuclear regulation. if SMR's were as cheap and easy as people would like, then they'd be popping up everywhere.
And the other thing is that nothing happens in the energy sector without government approval, policy support, and some form of financial support. Government policy has been massively behind wind and solar, there's been no support for SMR. All of which is a pity - big nuclear is massively expensive and high risk, we know small nuclear reactors work well and safely (as various navies have proven for many decades).
And the other thing is that nothing happens in the energy sector without government approval, policy support, and some form of financial support.
I think this is also a policy risk for using gas. Sure, if your DC is close enough to a gas pipeline, you could build a CCGT or OCGT and generate your own electricity. You could use that for the DC, and export any surplus. The DC operator takes the risk for building the turbines, the gas supplier just sells gas. But government policies in the West are very much anti-gas, and imposes costs on turbine operators to subsidise wind & solar.
It'll be interesting to see how government policies impact SMR deployments. The Rolls Royce SMR is slowly working it's way through the approvals processes, but in the UK, nuclear doesn't attract the same subsidies or incentives, even though they're essentially zero carbon. The prospects for changing that don't appear great with Ed Millibrain back in charge of UK energy policy, even though SMRs are a good solution to our growing energy crisis.
we know small nuclear reactors work well and safely (as various navies have proven for many decades).
True but we have no idea as to the purchase and running cost for military reactors. If they are costed like most military procurement they will not be a financially viable match for civilian markets.
True but we have no idea as to the purchase and running cost for military reactors. If they are costed like most military procurement they will not be a financially viable match for civilian markets.
I think some of it can be deduced, ie costs of UK's Astute program or other nations operating nuclear vessels. Or cost arguments and budget proposals around things like nuclear propulsion for the UK's QE2 carriers where it chose not to on cost grounds. Or there's the AUKUS deal to get Australia operating nuclear submarines, but that involves a whole bunch of other technology transfer, training, joint development etc etc.
Main difference I think though is they're very different operating environments to civilian. So civilian versions derived from naval reactors should be simpler & cheaper, but then gets into political issues like the levelised cost comparison, FOAK costs vs serial or parallel production and any regulatory red tape that might be cut. Key point of SMRs is having a standard design that can just be rolled off the production line with a lot of the regulatory overheads already done.
I think one of the most overlooked aspects is safety, so perception vs reality of submariners living & working in very close proximity to a reactor, usually with no ill effects. Also having watched a bunch of naval channels, just how small the actual reactors are.
... is there because nuclear facility operating companies have historically taken shortcuts and, from time-to-time, massively fucked up, with horrendous consequences.
I see no reason to expect such companies behave better were they unregulated. Do you?
"I see no reason to expect such companies behave better were they unregulated. Do you?"
Where did I say that these companies/activities should be unregulated? I AM a regulator myself, I fervently believe that done well, regulation is central to having a civilisation (you know, law courts, weights & measures, not too many dangerous products, utilities at least being monitored and sometimes held to account, etc).
The problem with nuclear regulation is that AS IT OPERATES NOW it is inordinately expensive and painfully slow, it hinders innovation at a time when we really need it.
Sorry, AC 12 Aug 2024/19:47, I incorrectly extrapolated from your phrase, "crushing weight of civil nuclear regulation" that you favored deregulation.
As a child, I was a science hound, and in first grade, I understood the very basics of fission reactors, (theoretical) fusion reactors, and nuclear fission bombs. In my teens, I toured two teaching reactors (I love seeing the visual effects of Cherenkov radiation; it is just so beautiful!) and a commercial power reactor.
But ... I and my classmates were unknowingly-at-the-time exposed to strontium-90 and cesium-131by playing in the grass at our local high school. How did that happen? A government agency, the Bureau of Mines, had offices and labs next door to the school. Instead of properly disposing of their used radionucleides, they just dumped them into the ground. A high local water table assisted the migration of said materials. They were sloppy, internally, as well. One of their buildings, after "remediation" went from being a three-story building down to being just a two-story building.
My point here is that in any group of intelligent, educated people ("scientists"), there always will be some who act irresponsibly, even reprehensively, with the powerful things they work with.
Thus the need for (painfully) detailed regulation, and the resulting inhibition of experimentation and development.
But can’t they build mini nuclear plants rather than using gas?
I'm fascinated that this response comes before What a pity LLMs appear to be economically unsustainable. If mini nuclear plants were genuinely viable, should we not be using them instead of gas for our present energy consumption? Seems like everyone is being blinded by the reflection of dollar signs in other people's eyes.
"They are going to revolutionise the way we interact with information forever."
Ok this "silver bullet" thinking is how we got here in the first place. LLMs are extremely neat and for specific applications can be powerful. But they're also like turing machines. Technically, yes, you can solve any programming problem, no matter how complex, with a turing machine. Practically...
Entirely understandable. A few years ago there was quite a bit of talk of installing microturbines for home or small industrial uses, in lieu of getting a grid connection. At 40p/kWh, a microturbine capable of burning gas at 10 or 20p/kWh assuming efficiencies of about 60% favours installing the gas burner. (Decent mid-sized CCGTs can achieve 60%).
The downside is the capital and maintenance cost of the microturbine; compared to paying for a grid connection.
If you can get a grid connection is a serious question today (for a variety of reasons), and so gas might well be favourable in terms of delivery timescales.
Whether you think gas prices will remain 10 to 20p/kWh is probably the other factor in your equation. Given the direction of the world it is reasonably safe to assume electricity WILL eventually beat this; once enough wind farms come on to stop gas prices from defining electricity prices as it does at the moment.
Now, whether the demand is justified is a whole different question. But that's not how capitalism works.
The downside is the capital and maintenance cost of the microturbine; compared to paying for a grid connection.
There's also Sterling engines, but those have the same issues. I've been looking at those for a new build and trying to balance the energy/cost equations. Which is the challenge for Sterling engines, microturbines, fuel cells etc. If you can design around them, they kinda work, but we have a lot of housing stock that just doesn't have the space. Microturbines in apartments could be.. fun, but in a basement serving those apartments could make more sense. But that needs developers to do it, and there isn't really any incentive. Well, there could be, if CHP could attract things like 'Renewable Heat' incentives, but there's an awful lot of paperwork involved.
...once enough wind farms come on to stop gas prices from defining electricity prices as it does at the moment.
But windmills are why we're dependent on gas. It's really the 'levelised cost' scam where the cost of their intermittency is excluded. If UK capacity auctions were changed to be based on firm delivery, there would be no economic case for wind. There is also the problem with the 'renewables' industry lying, eg telling us it's cheap, but no bids last auction and they're demanding ever larger subsidies. It'll be interesting to see how SMRs are treated, especially RR given that's a 300MW unit so provides a decent chunk of power.
I've been in the energy industry for a good few years and followed the various energy innovations. Sterling engines are certainly a thing, but in practice have never been economically viable. Throw in shitty subsidies like RHI and maybe they are, but the problem with subsidies is that they always shove the average cost of all energy upwards. That's not good, when our energy costs are 3-5x those of Chinese companies.
Throw in shitty subsidies like RHI and maybe they are, but the problem with subsidies is that they always shove the average cost of all energy upwards. That's not good, when our energy costs are 3-5x those of Chinese companies.
I think it's how and where subsidies are applied. Renewables are a good & bad example where that industry claims costs are falling and it's the cheapest form of generation. That lie is revealed in the capacity auctions and demand for ever increasing subsidies. Wind power has been around for thousands of years, wind turbines since the late 19th Century. So they're arguably a mature technology and subsidies are only necessary because they'd otherwise not be economically viable.
Subsidies for newer technology like SMRs or Stirling engines may make more sense. Like the wiki article shows, there have been attempts in the past to produce small, practical engines for microgeneration but those never took off. Now with our looming energy crisis, it could be beneficial to subsidise R&D or get production ramped up to help get SMRs or Stirling engines to market. It's happening to some extent with SMRs in the UK, but the subsidies are tiny in comparison to the billions being thrown at windmills.
Perhaps if you switch to Stirling you might save a few.
Doh! I first got interested when the mad Finn who runs the Hydraulic Press channel did a vid showing his home installation, but can't find that now. Even when I spell it correctly. But that seems the challenge for Stirling engines in general. Search for them on YT or popular search engines and results are dominated by models rather than anything practical. Or there are units like this-
https://en.wikipedia.org/wiki/Stirling_engine#/media/File:STM_Stirling_Generator_set.jpg
which at 55kW is more than I need. Maybe. Depending on how crazy I go with the workshop I also want to build and if I really need (ok, want) an induction forge etc. But an advantage is they also produce 'waste' heat that could obviously be used. Which is then back to the need to really design around the system and challenges with trying to retrofit into existing buildings.
A lot of these places get huge grants, or tax kick backs, to be built. If at some stage the market, bezos forbid, the market shrinks and these power generation points end up not being used, could they be made to supply the people?
Yeah I know, snowballs and hot underground caverns.
I just wish there was a bit more evenness or push back on the demands. Right now its, I scratch your back, you leave with the backscratcher and money.
The Register 
Biting the hand that feeds IT
Copyright. All rights reserved © 1998–2024
