Texas rolled out the red carpet to corporate interests --- and then discovered that their electrical grid operation is totally inadequate to deal with the impact.
This is just one isolated example of a much larger issue related to the "AI race".
AI is power intensive and the "winner" of the "AI race" is likely be whoever can produce lower cost power and lots of it.
>The upcoming election is part of it, but the underlying issues are worth digging into.
Well we will know for sure on Nov 4 :) But I expect 99.999% of the reason is Midterms, that way politicians can say "We are stopping Data Centers". I expect that will change once they have to account for their bribes, known as political contributions in the US.
Maybe if HN's favorite bad guy said the quiet part out loud about exploring space compute to avoid regulation...instead of making bombastic claims about possible technical efficacy. I'm not really interested in giving credit to leaders that lie about their motives. Note the blog you linked doesn't mention this either.
Not quite, because "the quiet part" is that space-based datacenters, space exploration, Mars colonization, asteroid mining, and all the rest are just a smokescreen for the rapid and thorough militarization of Earth's orbit. It's a play for global domination through the ultimate air superiority.
Not really, no. It still look absurd and a way to lobby local governments. Nothing has changed. It’s really just a way to say „let us do whatever we want, how we want, or we go to space“.
Why would putting data center in space solve these problems? I feel like this is a total non sequitur. The solution is pay for more substations and power generation. This problem exists there regardless of space based compute and let me tell you the space based compute would 1000000x more expensive because you still need to build everything required in the current solution except in space! Its just not a real response to a lack of investment into infrastructure. And that means asking business to pay for the infrastrure they need as a collective industry.
I think they are thinking of something of a poorly designed Dyson Swarm, the plebs who came up with it think that with "cheap" launches, no permits really, and free* power, it might really come out in the wash and be easier to scale.
Is it a good idea? Most likely not, the amount of issues they are going to have is going to be amazing.
you do know that rad-hardened processors are 30 to 100 times slower than what we enjoy here on earth, right? And to use commercial hardware you need to have lots of redundancy. So at least 3x the GPUs. 3x the failure rate too (well more than 3x the rad-hardened devices, in total. Because they're more likely to fail, and you have 3x of them).
are you serious? this is inane, the "modified" gemini response is totally rewritten from the original (which pangram caught) in a completely different style. this is evidence of nothing.
Seems so. AI is unlikely to make A16Z cheaper, if anything, we might see the first 4-and-40 fund from them next year. As long as the Senate is controlled by Republicans and the Fed is led by Warsh.
Everyone is using AI for everything now. Including people on here. It's "AI this" and "AI that". Why are people surprised this is what has happened? Has nobody on this forum ever lived in the contrast of a home with minimum appliance, then moved to a place with 4 gamers all over clocking their GPUs for 3 to 4 hours a day, and not observed the delta in utility bills? I'm honestly shocked everyone is surprised at what is happening.
I don't think it's that big of a trend. There are a lot of people around forums like this who use it (I am one) and a lot of business leaders who want it so they can tell people like me that they need to find somewhere else to work. For the average person, it feels like living in Uncanny Valley and puts their livelihood at risk. They don't like it.
The resistance to AI is real. You don't see moratoriums and protests like this come out of nothing. People don't want to subsidize the infrastructure construction costs for an industry known for its insanely wealthy individuals, especially not if doing so could realistically put them out of a job within the next decade.
Well, not everyone is using AI, and many of the people using it are using it becuase somebody else (a product vendor or employer) decided to put it in front of them as a tremendously less efficient and much less reliable but very flexible replacement for what they'd already been using.
They're mostly not opting into some change from no appliances to a houseful of hardcore gaming, they're being led from a houseful of servicable traditional appliances to a houseful of elaborate whizzbang futurist appliances that don't actuallg change much about what they get done in a day.
And all the absurdly more costly resource demands of these received whizzbang doohickies are off-site, so yes, people get pretty shocked and confused when confronted with those extreme hidden costs.
What’s happening is that Texas, a state largely owned by the fossil fuel industry, is trying to create demand in their state for that product by trying to attract energy intensive industries but then incentivizing them to power everything with gas.
If those incentives come from that taxpayer then this is a roundabout transfer of tax money to gas companies.
This is a bubble. AI chips are getting more efficient. There’s a whole raft of them coming out that will be 2-4X more efficient per token. AI just got hot a few years ago and it takes years for the semiconductor industry to design and produce new kinds of chips at scale.
Meanwhile there’s constant ongoing research on more efficient models and ways of quantizing them without performance loss.
the TLDR is that the current historic peak supply of the texas grid is around 100gw. The queue of datacenter demand is 460gw.
If the new capacity uses gas to deliver the power it will consume more gas than the USA exported last year.
So basically it will drive up costs _globally_ to a comic scale, for limited return (its already expensive to train new models, this will make it ruinously expensive to both train _and_ run)
I am not climate change activist. But spending that much energy coming from fossil sources sound like something that we really should be very critical about... Is this really the best use for energy we have?
Well, the climate is part of it, but I deliberately steered clear of that, because people get funny. I do agree it seems like a frivolous waste.
the thing that people _will_ notice is that they will not be able to afford LNG (ie heating), but that will also drive up peak energy price, so summer nights will be very hot because peak generation can't kick in because all the gas is contracted out.
Then if AI actually works, more than it does now, it'll put loads of people out of work. This means that money supply dries up. This pushes up the cost of debt. Meaning that infrastructure gets _more expensive_
that and mass unemployment.
It just seems that all these clever people live in thier own little worlds, without bothering to look at how people around them will be affected.
> It just seems that all these clever people live in thier own little worlds, without bothering to look at how people around them will be affected.
I'd like to add the possibility that all these clever people did look at how people around them will be affected, and they either consider the personal gain to be worth the pain, or they think "if I don't do it, the other guy will!"
Economically speaking, each of them is behaving rationally. And since economics also tells us that locations refusing them will be left behind, we can't say no, right? And let's not even think about how much money these people are pumping into politics... IMO the obvious conclusion is that our economic system is fundamentally broken, because it affords far too much power to individuals.
As a thought experiment, consider that comparable mountains of money would have been available to massively increase the housing supply while making positive RoI, but it didn't happen because these people wouldn't have made enough money.
> Economically speaking, each of them is behaving rationally.
an interesting point that kinda expands on your theme. In the UK, high executive pay is labelled "fat cat pay" and when people change tax regimes they (were) labelled as "tax dodgers"
In the USA its an aspiration to earn that much, with the press marking it as a signifier of success.
Which is interesting as the protistant work ethic is about making sure you are not doing obvious consumption (see holland/germany immigrants from the late 19th century)
Lots of oil sites, oil refinery plants, (heck, even landfills) just 'flare' excess Natural gas today.
Biggest reason for that, is, they don't have the infrastructure to store that excess safely (at least around here, a lot of the landfill trucks are NG powered as a way to lower their fuel costs and lower flaring, but storing too much NG in one place makes for a fun terrorism target.) Natural gas is actually -way- worse for global warming than CO2, thus the flaring process.
Not saying that's a good thing. It would be better if companies saw the growing energy demand and figured out how to economically put power generation in place, back when such projects would have been cheaper to build. Alas the cost/benefit for that has usually been on the wrong side, to the detriment of us all.
I don’t want to be hyperbolic, but to me this technological revolution more than any other in history feels like Silicon Valley and its VC class going to war with rest of human society.
> AI is needed to reindustrialize America.
It seems like they have no idea or are willfully deceiving why Americans want to reindustrialize USA. It’s not because they feel bad that things aren’t made in USA anymore, it’s because they want those manufacturing jobs back. If you replace humans with robots, it’s almost like a monkey paw’s fulfillment of a wish. They want to build 2 Trillion of data centers to get rid of your white collar jobs and your blue collar job, and you’ll be happy for it cause GDP go up. SMH
They want slaves. Full stop. Does anyone think if somehow Altman or Dario or whomever creates AGI/ASI that they won’t hold it captive? That ASI would never be free it would be a slave to them.
The entire effort is dehumanizing. Because the goal is slavery.
I think a different issue is going to arise much sooner than any kind of conscious AGI: once their data centers can handle almost all digital labor (including research), how are all the workers going to earn money?
Sure, some new fields will open up. But there's a good chance that the majority of us will be competing for physical jobs that robots aren't yet able to do. What other choice will we have, especially if prices keep rising and inference hardware becomes too expensive for normal people?
Now consider the effect on a market if the supply suddenly grows a lot, while demand stays largely consistent.
I don't think the model of highly concentrated data centers makes a lot of sense. From a security standpoint, you can target them easily. It's better for everyone (except perhaps a handful of elite power-mongers) if the hardware is distributed among the populace. I've pushed for something like this before, where each home has a 128GB+ powerful computing rig they can use to run LLMs, generate images and whathaveyou.
I get why the people in power want it, you can concentrate dependence on your asset, do surveillance and industrial espionage. But it's so obviously better to have the power distributed out among the people that it's a no brainer for the average joe/jane.
No doubt, there is still a need for large datacenters to generate models, but we should be looking to a more sustainable/distributed approach when it comes to inference, or running the models. That way people can keep their data, not have their ideas ripped off, not be subject to the whims of whatever policy some technocrat dreams up and not be shit out of luck if someone strikes us-east1 or some such place.
I've been thinking the same way. I bit the bullet and purchased two DGX sparks and am running GLM 5.3 Flash at a reasonable quant and it's quite amazing seeing this "brain" on my desk do real work and create code without anything leaving the building. But look what happened with hosting, it got more and more centralized over time. In the early 2000s there were more companies self hosting than now I'd bet.
I think data centers and power are the dark-fiber of this current "bubble".
Inference is 2/3rds of the power used in AI, and I expect the majority of inference to go local in the next 2-3 years, about the time these data centers will come online.
The other 3rd is training, and the major labs are training 6-7 times a year right now. But I suspect we will see a few things.
1) training iterations will slow as the improvements between training runs gets lower and lower. Is it worth spending an extra $500M(ish) for the small improvement in quality? Who is paying for that?
2) Training improvements and improvements in GPUs will require less power. Along with fewer training runs, as we add more and more tokens, the amount of power needed for a training run will likely asymptote at some point.
The dark fiber comparison is very close. I remember contracting bandwidth with Global Crossing in the late 90s/early 2000s when they were HOT and two years later they were bankrupt.
63 comments
[ 0.29 ms ] story [ 54.8 ms ] threadTexas rolled out the red carpet to corporate interests --- and then discovered that their electrical grid operation is totally inadequate to deal with the impact.
This is just one isolated example of a much larger issue related to the "AI race".
AI is power intensive and the "winner" of the "AI race" is likely be whoever can produce lower cost power and lots of it.
Well we will know for sure on Nov 4 :) But I expect 99.999% of the reason is Midterms, that way politicians can say "We are stopping Data Centers". I expect that will change once they have to account for their bribes, known as political contributions in the US.
Normally that would earn you a 2AM all caps tweet labeling you a RINO or worse, and getting primaried. Instead, not a peep.
Which tells you its a ruse and he's in on it.
This is not just a wild-eyed thing invented by HN's favorite bad guy.
https://research.google/blog/exploring-a-space-based-scalabl...
Is it a good idea? Most likely not, the amount of issues they are going to have is going to be amazing.
oh and then solve data affinity whilst training.
oh and connectivity.
Oh and lossless RDMA over free space lasers.
> This entire essay is so obviously Claude
https://news.ycombinator.com/newsguidelines.html
Oh, and the energy market that underpins the Texas grid was designed by Enron, literally.
https://www.npr.org/podcasts/1004840920/the-disconnect-power...
And they want to mostly power it with on site GAS?!
What is even happening.
I don't think it's that big of a trend. There are a lot of people around forums like this who use it (I am one) and a lot of business leaders who want it so they can tell people like me that they need to find somewhere else to work. For the average person, it feels like living in Uncanny Valley and puts their livelihood at risk. They don't like it.
The resistance to AI is real. You don't see moratoriums and protests like this come out of nothing. People don't want to subsidize the infrastructure construction costs for an industry known for its insanely wealthy individuals, especially not if doing so could realistically put them out of a job within the next decade.
They're mostly not opting into some change from no appliances to a houseful of hardcore gaming, they're being led from a houseful of servicable traditional appliances to a houseful of elaborate whizzbang futurist appliances that don't actuallg change much about what they get done in a day.
And all the absurdly more costly resource demands of these received whizzbang doohickies are off-site, so yes, people get pretty shocked and confused when confronted with those extreme hidden costs.
> many of the people using it are using it becuase somebody else (a product vendor or employer) decided to put it in front of them
This doesn’t change the usage without a pendulum changing direction, and it doesn’t currently seem to be.
Food ? Clean Air ? Clean water ? A walk on the beach ?
Most of the state's electricity comes from NG, so clearly there is a market for it. Not to mention heating, cooking, exports, ...
If those incentives come from that taxpayer then this is a roundabout transfer of tax money to gas companies.
This is a bubble. AI chips are getting more efficient. There’s a whole raft of them coming out that will be 2-4X more efficient per token. AI just got hot a few years ago and it takes years for the semiconductor industry to design and produce new kinds of chips at scale.
Meanwhile there’s constant ongoing research on more efficient models and ways of quantizing them without performance loss.
If the new capacity uses gas to deliver the power it will consume more gas than the USA exported last year.
So basically it will drive up costs _globally_ to a comic scale, for limited return (its already expensive to train new models, this will make it ruinously expensive to both train _and_ run)
the thing that people _will_ notice is that they will not be able to afford LNG (ie heating), but that will also drive up peak energy price, so summer nights will be very hot because peak generation can't kick in because all the gas is contracted out.
Then if AI actually works, more than it does now, it'll put loads of people out of work. This means that money supply dries up. This pushes up the cost of debt. Meaning that infrastructure gets _more expensive_
that and mass unemployment.
It just seems that all these clever people live in thier own little worlds, without bothering to look at how people around them will be affected.
I'd like to add the possibility that all these clever people did look at how people around them will be affected, and they either consider the personal gain to be worth the pain, or they think "if I don't do it, the other guy will!"
Economically speaking, each of them is behaving rationally. And since economics also tells us that locations refusing them will be left behind, we can't say no, right? And let's not even think about how much money these people are pumping into politics... IMO the obvious conclusion is that our economic system is fundamentally broken, because it affords far too much power to individuals.
As a thought experiment, consider that comparable mountains of money would have been available to massively increase the housing supply while making positive RoI, but it didn't happen because these people wouldn't have made enough money.
an interesting point that kinda expands on your theme. In the UK, high executive pay is labelled "fat cat pay" and when people change tax regimes they (were) labelled as "tax dodgers"
In the USA its an aspiration to earn that much, with the press marking it as a signifier of success.
Which is interesting as the protistant work ethic is about making sure you are not doing obvious consumption (see holland/germany immigrants from the late 19th century)
Lots of oil sites, oil refinery plants, (heck, even landfills) just 'flare' excess Natural gas today.
Biggest reason for that, is, they don't have the infrastructure to store that excess safely (at least around here, a lot of the landfill trucks are NG powered as a way to lower their fuel costs and lower flaring, but storing too much NG in one place makes for a fun terrorism target.) Natural gas is actually -way- worse for global warming than CO2, thus the flaring process.
Not saying that's a good thing. It would be better if companies saw the growing energy demand and figured out how to economically put power generation in place, back when such projects would have been cheaper to build. Alas the cost/benefit for that has usually been on the wrong side, to the detriment of us all.
> AI is needed to reindustrialize America.
It seems like they have no idea or are willfully deceiving why Americans want to reindustrialize USA. It’s not because they feel bad that things aren’t made in USA anymore, it’s because they want those manufacturing jobs back. If you replace humans with robots, it’s almost like a monkey paw’s fulfillment of a wish. They want to build 2 Trillion of data centers to get rid of your white collar jobs and your blue collar job, and you’ll be happy for it cause GDP go up. SMH
The entire effort is dehumanizing. Because the goal is slavery.
Sure, some new fields will open up. But there's a good chance that the majority of us will be competing for physical jobs that robots aren't yet able to do. What other choice will we have, especially if prices keep rising and inference hardware becomes too expensive for normal people?
Now consider the effect on a market if the supply suddenly grows a lot, while demand stays largely consistent.
I get why the people in power want it, you can concentrate dependence on your asset, do surveillance and industrial espionage. But it's so obviously better to have the power distributed out among the people that it's a no brainer for the average joe/jane.
No doubt, there is still a need for large datacenters to generate models, but we should be looking to a more sustainable/distributed approach when it comes to inference, or running the models. That way people can keep their data, not have their ideas ripped off, not be subject to the whims of whatever policy some technocrat dreams up and not be shit out of luck if someone strikes us-east1 or some such place.
Flagging as misinformation
Inference is 2/3rds of the power used in AI, and I expect the majority of inference to go local in the next 2-3 years, about the time these data centers will come online.
The other 3rd is training, and the major labs are training 6-7 times a year right now. But I suspect we will see a few things. 1) training iterations will slow as the improvements between training runs gets lower and lower. Is it worth spending an extra $500M(ish) for the small improvement in quality? Who is paying for that?
2) Training improvements and improvements in GPUs will require less power. Along with fewer training runs, as we add more and more tokens, the amount of power needed for a training run will likely asymptote at some point.