Generally you’re right, but specifically there are issues with what you’re saying.
First the grid. Of course the grid can’t be fixed on these time scales. That’s why the data centers are building their own power plants. Some are natural gas. Some are nuclear. They are even buying old decommissioned or moth balled nuclear plants and starting them back up for just the data center. The grid doesn’t need to be upgraded.
The value of a.i. is obviously not in commerce. The US sees it as a potential strategic game changer for both domestic police state management and for peer conflict. The business model is DoD funding.
The construction is also not constrained to the US. US companies are building nearly anywhere within the US hegemonic sphere of influence. There’s a lot of surplus power hiding out there, especially around hydroelectric plants. Canada in particular overproduces electricity significantly.
There are also a bunch of underdeveloped opportunities, like hydrogen and geothermal, that while they couldn’t meet full demand, will be given massive cash injections and any amount of progress will create some market exuberance. A single large scale geothermal project could create a large off-grid campus for dozens of data centers. Do that a few times and the geothermal industry starts getting larger.
I agree it’s a losing battle, but I think each attempt to solve will give them a few more months, and they have dozens of attempts to make.
I can tell you didn’t actually read my article. :) Here’s the reality of building power plants in the USA:
The Vogtle plant in Georgia, for example, just came online after 15 years of construction. It was supposed to cost $14 billion; the final price tag was over $30 billion for just 2.2 gigawatts. That works out to about $16 billion per gigawatt.
Both building data centers on the scale that’s being proposed and supplying them with power are frankly fantastical ideas that have no grounding in reality. It simply cannot be done.
You are right that the reason the US has to stay in the AI race is precisely because there is a chance that it might work. As long as China continues to develop this tech, the US has no choice but to try and keep up. However, that’s forcing the US into diverting a huge chunk of their economy away from productive and socially necessary purposes. It’s exactly the same situation USSR found itself in during the Cold War.
I live in Canada, I can tell you that Canada absolutely does not have spare capacity on this scale. Meanwhile, Canada is suffering from a lot of the same economic problems the US is.
And, you keep looping back to hydrogen and geothermal, but that’s decades away. This tech is not being deployed at any meaningful scale in the US. The demand for energy is already outpacing the supply now. The spare grid capacity is at mere 15% right now, and that’s needed for stuff like extreme weather events such as heat waves. Data centers are starting to eat into this spare capacity already. The Pentagon has already been gaming scenario of mass blackouts caused by the growing energy demand. That’s why the whole Stargate project is has now stalled.
We agree that it’ll likely take a couple of years for all this to come to a head, but I see no way a crash can be avoided at this point. And it’s always possible there the crash will be triggered by some other economic event that’s unrelated to the AI bubble. The job numbers in the US are looking terrible, businesses are shuttering, and the tariff war is putting pressure on the whole economy. While stock markets don’t care about the lives of the working class, we should not make the same mistake. At the end of the day people have to make ends meet. As Lenin put it, every society is three hot meals away from chaos.
Yeah, sorry, I didn’t address the power plant point. These data center aren’t building gigawatts of power plant on site (except for the nukes). They’re building megawatts of power and their using existing gas generators and turbines and feeding the power directly into the data center. They’re not building traditional power plants to enhance the grid, they’re building off grid.
I live in Canada, I can tell you that Canada absolutely does not have spare capacity on this scale. Meanwhile, Canada is suffering from a lot of the same economic problems the US is.
I have it on good authority that in parts of Canada there is a massive surplus of electricity, as there is in a few regions in Scandinavia. They’re selling electricity at less then $0.05/KwH. This is mostly due to over building of hydropower in more remote areas. I’m not saying the whole grid is flush with power. I’m saying there are pockets of overproduction and the data centers are going to find ways of getting there.
And, you keep looping back to hydrogen and geothermal, but that’s decades away
It’s not. There are active projects right now, actually deployed right now. There are hydrogen fuel cells deployed right now. There are multi-fuel generators that take hydrogen right now. There are hydrogen production capabilities live right now. It’s decades away from being produced centrally, stored, and transported at scale, but data centers are capable of being built with modular capabilities completely disconnected from larger nonexistent infrastructure.
There are active geothermal projects happening right now. The US military is investing heavily in it, and data centers are already considered matters of national security. We’re going to see at least a dozen data centers powered entirely by geothermal within 5 years.
The demand for energy is already outpacing the supply now
This is true and not true depending on your meaning. The existing data centers are powered. The data centers being built are able to be powered for the most part AFAICT. The number of unstarted projects? Yeah, there’s a lot of demand there and they aren’t started likely because they are trying to figure out supply. The amount of capital in terms of potential data centers? Absolutely not enough supply.
But this will cause supply to expand. And as you say, vertical is very slow right now, so it will create pressure to go horizontal. A.I. may not be productive, but the race to generate more power and thus make a profit is going to be productive. In fact, it might be the single most important factor in the US developing new energy capabilities and capacities.
The spare grid capacity is at mere 15% right now, and that’s needed for stuff like extreme weather events such as heat waves. Data centers are starting to eat into this spare capacity already. The Pentagon has already been gaming scenario of mass blackouts caused by the growing energy demand. That’s why the whole Stargate project is has now stalled.
And as I keep saying, data center are literally being built off-grid to account for this. There’s a 2 year waiting list to get generators right now because the data center projects have bought up future production lines. But that demand is going to push more manufacturing of generators, and the whole suite is going to consume a ton of natural gas. But again, multi-fuel generators are on the way and that means we’re going to see not only natural gas but biogas and hydrogen as fuels that can be brought in at later dates.
But really, this is happening. Texas announced that no new data centers will be allowed to stay on grid in emergencies, so what’s happening right now is that thousands of acres of data center projects are being built without ANY connection to the grid, entirely on natural gas generators.
Remember, markets aren’t great ways to run society, but they are great at getting millions of people to work on large problems independently and produce a robust suite of options in the face of scarcity that limits profits. That’s what’s literally happening on the ground right now. I wouldn’t be surprised if safety regulations that have been slowing down other energy tech gets destroyed and a whole bunch of industrial disasters start happening as profit-chasers start loading trains up with fuels without proper equipment.
We agree that it’ll likely take a couple of years for all this to come to a head
Agreed
but I see no way a crash can be avoided at this point
I don’t think it can be avoided, I think it can be deferred. Given everything remains the same as it is right now, I think it’s less than 2 years. But I think the current developments I am aware of put us at 2 - 3 years. Within that time, there will be new developments. So it’s possible it could be deferred longer.
And it’s always possible there the crash will be triggered by some other economic event that’s unrelated to the AI bubble
I think that’s entirely possible. I mean, anything that shuts down Manhattan Island would probably do it. Or another major infectious disease catastrophe. Or the food supply being incredibly tight from this harvest (although we’ve got plenty of soy for everyone).
However, usually the way you solve that problem as an empire is you expand your extraction and lebensraum, and the US is actively on that path in Latin America. I think the USA will have real trouble fighting a jungle war in Venezuela, but I think there’s a real risk of the US gaining breathing room through use of military force from Mexico all the way South.
The job numbers in the US are looking terrible
Jobs only matter for 2 reasons: production and money circulation. Most Americans aren’t productive anyway, working instead on paper pushing, marketing/advertising, sales, etc. So that’s not useful. The jobs numbers are a real problem because money is not circulating downward. That’s solvable with government intervention. It’s solvable by hiring more into military roles (ICE, local police, national guards, military, etc). It’s solvable by pushing people into the fields to replace the deported immigrants. Money flowing to the working class is one of the easiest things for a modern government to solve. They can literally just create the money in their accounts and suddenly consumer spending will be up and management jobs will come back. Jobs programs are in the USA’s history and they are part of the fascist playbook as well. I fully expect there to be a jobs program at some point in the next 5 years.
the tariff war is putting pressure on the whole economy
Honestly, I don’t even know if this is the biggest problem. The problem is that the US has lost the economics game globally and is no longer competitive enough to charge a sufficiently high price to maintain profits. We’re going to have reorganize the economy in the US to be a hermit kingdom that eats what it produces. We’re being isolated. All of the West is being isolated. The tariffs can be lifted at any time, and I don’t think it will reverse the macro trend that causing the economy to collapse.
While stock markets don’t care about the lives of the working class, we should not make the same mistake. At the end of the day people have to make ends meet. As Lenin put it, every society is three hot meals away from chaos.
Don’t get me wrong. I’m not saying ANY of this is good. I’m trying to temper your predictions of imminent failure with the factors that will allow this shit show to continue far longer than you expect it to.
Yeah, sorry, I didn’t address the power plant point. These data center aren’t building gigawatts of power plant on site (except for the nukes). They’re building megawatts of power and their using existing gas generators and turbines and feeding the power directly into the data center. They’re not building traditional power plants to enhance the grid, they’re building off grid.
My point was that this is how long it takes to build a power plant in the US. This is what I keep saying here, the fact that they say they will build gigawatts of power plants doesn’t mean it’s actually possible. These are not realistic projections. These companies aren’t going to be magically building power plants faster than anybody else in the country.
I have it on good authority that in parts of Canada there is a massive surplus of electricity, as there is in a few regions in Scandinavia. They’re selling electricity at less then $0.05/KwH. This is mostly due to over building of hydropower in more remote areas. I’m not saying the whole grid is flush with power. I’m saying there are pockets of overproduction and the data centers are going to find ways of getting there.
There is a lot of hydro in Ontario, however it’s hard to call it a surplus since the whole grid has gone down several times under load already. Whenever there’s stress like a heat wave there’s power shortage. So, on the authority of somebody who has to live here and see what the grid looks like, I can tell you that it’s not going to support mass data center roll outs.
The existing data centers are powered.
The problems is that prices are already going up as a direct result of data centers eating up existing power supply. So, that’s negatively affecting the economy as a whole. Meanwhile, I simply don’t see how these off grid power stations can be built at a pace that’s needed to keep up with the projected demand.
I think there’s a real risk of the US gaining breathing room through use of military force from Mexico all the way South
If the US really starts attacking Latin America again, that’s absolutely going to unite everyone there against them. Attacking Venezuela or Mexico could turn out to be the very event that brings the empire down.
That’s solvable with government intervention.
I think you’re wildly overestimating the competence of the policy makers. This is certainly something that would’ve been solvable in the 30s or the 50s, but America today is a different beast. The US is basically at the child emperor stage of the Roman empire where you have an oligarchy that’s mostly focused on political intrigue as opposed to doing any actual governing.
We’re going to have reorganize the economy in the US to be a hermit kingdom that eats what it produces.
The problem is that the US doesn’t produce what it needs to be self sufficient. Nor are there workers specializing in the industries like science, engineering, and trades available because there’s no reason to go into these careers since there few jobs to be had there. Most people go into law, entertainment, service industry, and so on. That’s why tariffs have backfired so spectacularly, there are no domestic industries to protect. So all the tariffs do is just raise prices for everyone.
I don’t think lifting tariffs will reverse the trend either, but tariffs absolutely acted as an accelerant because they made the input costs go through the roof, and destroyed a whole bunch of companies in the process. It’s an ongoing economic shock on an economy that’s already in a bad way.
I’m trying to temper your predictions of imminent failure with the factors that will allow this shit show to continue far longer than you expect it to.
It’s certainly possible they’re going to be able to kick the can down the road longer. Even if a major crash happens in the next couple of years, things will most likely stabilize to a point where the system can keep chugging along. The US could end up starting to look like Brazil where you have the top 10-20% who live relatively well, and the rest of the population ends up in favelas. That’s a very likely scenario in my opinion.
Generally you’re right, but specifically there are issues with what you’re saying.
First the grid. Of course the grid can’t be fixed on these time scales. That’s why the data centers are building their own power plants. Some are natural gas. Some are nuclear. They are even buying old decommissioned or moth balled nuclear plants and starting them back up for just the data center. The grid doesn’t need to be upgraded.
The value of a.i. is obviously not in commerce. The US sees it as a potential strategic game changer for both domestic police state management and for peer conflict. The business model is DoD funding.
The construction is also not constrained to the US. US companies are building nearly anywhere within the US hegemonic sphere of influence. There’s a lot of surplus power hiding out there, especially around hydroelectric plants. Canada in particular overproduces electricity significantly.
There are also a bunch of underdeveloped opportunities, like hydrogen and geothermal, that while they couldn’t meet full demand, will be given massive cash injections and any amount of progress will create some market exuberance. A single large scale geothermal project could create a large off-grid campus for dozens of data centers. Do that a few times and the geothermal industry starts getting larger.
I agree it’s a losing battle, but I think each attempt to solve will give them a few more months, and they have dozens of attempts to make.
I can tell you didn’t actually read my article. :) Here’s the reality of building power plants in the USA:
Both building data centers on the scale that’s being proposed and supplying them with power are frankly fantastical ideas that have no grounding in reality. It simply cannot be done.
You are right that the reason the US has to stay in the AI race is precisely because there is a chance that it might work. As long as China continues to develop this tech, the US has no choice but to try and keep up. However, that’s forcing the US into diverting a huge chunk of their economy away from productive and socially necessary purposes. It’s exactly the same situation USSR found itself in during the Cold War.
I live in Canada, I can tell you that Canada absolutely does not have spare capacity on this scale. Meanwhile, Canada is suffering from a lot of the same economic problems the US is.
And, you keep looping back to hydrogen and geothermal, but that’s decades away. This tech is not being deployed at any meaningful scale in the US. The demand for energy is already outpacing the supply now. The spare grid capacity is at mere 15% right now, and that’s needed for stuff like extreme weather events such as heat waves. Data centers are starting to eat into this spare capacity already. The Pentagon has already been gaming scenario of mass blackouts caused by the growing energy demand. That’s why the whole Stargate project is has now stalled.
We agree that it’ll likely take a couple of years for all this to come to a head, but I see no way a crash can be avoided at this point. And it’s always possible there the crash will be triggered by some other economic event that’s unrelated to the AI bubble. The job numbers in the US are looking terrible, businesses are shuttering, and the tariff war is putting pressure on the whole economy. While stock markets don’t care about the lives of the working class, we should not make the same mistake. At the end of the day people have to make ends meet. As Lenin put it, every society is three hot meals away from chaos.
Yeah, sorry, I didn’t address the power plant point. These data center aren’t building gigawatts of power plant on site (except for the nukes). They’re building megawatts of power and their using existing gas generators and turbines and feeding the power directly into the data center. They’re not building traditional power plants to enhance the grid, they’re building off grid.
I have it on good authority that in parts of Canada there is a massive surplus of electricity, as there is in a few regions in Scandinavia. They’re selling electricity at less then $0.05/KwH. This is mostly due to over building of hydropower in more remote areas. I’m not saying the whole grid is flush with power. I’m saying there are pockets of overproduction and the data centers are going to find ways of getting there.
It’s not. There are active projects right now, actually deployed right now. There are hydrogen fuel cells deployed right now. There are multi-fuel generators that take hydrogen right now. There are hydrogen production capabilities live right now. It’s decades away from being produced centrally, stored, and transported at scale, but data centers are capable of being built with modular capabilities completely disconnected from larger nonexistent infrastructure.
There are active geothermal projects happening right now. The US military is investing heavily in it, and data centers are already considered matters of national security. We’re going to see at least a dozen data centers powered entirely by geothermal within 5 years.
This is true and not true depending on your meaning. The existing data centers are powered. The data centers being built are able to be powered for the most part AFAICT. The number of unstarted projects? Yeah, there’s a lot of demand there and they aren’t started likely because they are trying to figure out supply. The amount of capital in terms of potential data centers? Absolutely not enough supply.
But this will cause supply to expand. And as you say, vertical is very slow right now, so it will create pressure to go horizontal. A.I. may not be productive, but the race to generate more power and thus make a profit is going to be productive. In fact, it might be the single most important factor in the US developing new energy capabilities and capacities.
And as I keep saying, data center are literally being built off-grid to account for this. There’s a 2 year waiting list to get generators right now because the data center projects have bought up future production lines. But that demand is going to push more manufacturing of generators, and the whole suite is going to consume a ton of natural gas. But again, multi-fuel generators are on the way and that means we’re going to see not only natural gas but biogas and hydrogen as fuels that can be brought in at later dates.
But really, this is happening. Texas announced that no new data centers will be allowed to stay on grid in emergencies, so what’s happening right now is that thousands of acres of data center projects are being built without ANY connection to the grid, entirely on natural gas generators.
Remember, markets aren’t great ways to run society, but they are great at getting millions of people to work on large problems independently and produce a robust suite of options in the face of scarcity that limits profits. That’s what’s literally happening on the ground right now. I wouldn’t be surprised if safety regulations that have been slowing down other energy tech gets destroyed and a whole bunch of industrial disasters start happening as profit-chasers start loading trains up with fuels without proper equipment.
Agreed
I don’t think it can be avoided, I think it can be deferred. Given everything remains the same as it is right now, I think it’s less than 2 years. But I think the current developments I am aware of put us at 2 - 3 years. Within that time, there will be new developments. So it’s possible it could be deferred longer.
I think that’s entirely possible. I mean, anything that shuts down Manhattan Island would probably do it. Or another major infectious disease catastrophe. Or the food supply being incredibly tight from this harvest (although we’ve got plenty of soy for everyone).
However, usually the way you solve that problem as an empire is you expand your extraction and lebensraum, and the US is actively on that path in Latin America. I think the USA will have real trouble fighting a jungle war in Venezuela, but I think there’s a real risk of the US gaining breathing room through use of military force from Mexico all the way South.
Jobs only matter for 2 reasons: production and money circulation. Most Americans aren’t productive anyway, working instead on paper pushing, marketing/advertising, sales, etc. So that’s not useful. The jobs numbers are a real problem because money is not circulating downward. That’s solvable with government intervention. It’s solvable by hiring more into military roles (ICE, local police, national guards, military, etc). It’s solvable by pushing people into the fields to replace the deported immigrants. Money flowing to the working class is one of the easiest things for a modern government to solve. They can literally just create the money in their accounts and suddenly consumer spending will be up and management jobs will come back. Jobs programs are in the USA’s history and they are part of the fascist playbook as well. I fully expect there to be a jobs program at some point in the next 5 years.
Honestly, I don’t even know if this is the biggest problem. The problem is that the US has lost the economics game globally and is no longer competitive enough to charge a sufficiently high price to maintain profits. We’re going to have reorganize the economy in the US to be a hermit kingdom that eats what it produces. We’re being isolated. All of the West is being isolated. The tariffs can be lifted at any time, and I don’t think it will reverse the macro trend that causing the economy to collapse.
Don’t get me wrong. I’m not saying ANY of this is good. I’m trying to temper your predictions of imminent failure with the factors that will allow this shit show to continue far longer than you expect it to.
My point was that this is how long it takes to build a power plant in the US. This is what I keep saying here, the fact that they say they will build gigawatts of power plants doesn’t mean it’s actually possible. These are not realistic projections. These companies aren’t going to be magically building power plants faster than anybody else in the country.
There is a lot of hydro in Ontario, however it’s hard to call it a surplus since the whole grid has gone down several times under load already. Whenever there’s stress like a heat wave there’s power shortage. So, on the authority of somebody who has to live here and see what the grid looks like, I can tell you that it’s not going to support mass data center roll outs.
The problems is that prices are already going up as a direct result of data centers eating up existing power supply. So, that’s negatively affecting the economy as a whole. Meanwhile, I simply don’t see how these off grid power stations can be built at a pace that’s needed to keep up with the projected demand.
If the US really starts attacking Latin America again, that’s absolutely going to unite everyone there against them. Attacking Venezuela or Mexico could turn out to be the very event that brings the empire down.
I think you’re wildly overestimating the competence of the policy makers. This is certainly something that would’ve been solvable in the 30s or the 50s, but America today is a different beast. The US is basically at the child emperor stage of the Roman empire where you have an oligarchy that’s mostly focused on political intrigue as opposed to doing any actual governing.
The problem is that the US doesn’t produce what it needs to be self sufficient. Nor are there workers specializing in the industries like science, engineering, and trades available because there’s no reason to go into these careers since there few jobs to be had there. Most people go into law, entertainment, service industry, and so on. That’s why tariffs have backfired so spectacularly, there are no domestic industries to protect. So all the tariffs do is just raise prices for everyone.
I don’t think lifting tariffs will reverse the trend either, but tariffs absolutely acted as an accelerant because they made the input costs go through the roof, and destroyed a whole bunch of companies in the process. It’s an ongoing economic shock on an economy that’s already in a bad way.
It’s certainly possible they’re going to be able to kick the can down the road longer. Even if a major crash happens in the next couple of years, things will most likely stabilize to a point where the system can keep chugging along. The US could end up starting to look like Brazil where you have the top 10-20% who live relatively well, and the rest of the population ends up in favelas. That’s a very likely scenario in my opinion.