Episode 394 – The Leviathan that Eats the World: AI’s Hidden Hunger for Energy, Water & the Planet with Erald Kolasi

Steve Grumbine:
00:00:42

All right folks, this is Steve with Macro N Cheese. Folks, AI has been ever present in all of our lives. It is probably one of the most contested ideas out there and yet everybody uses it.

When you’re living in a capitalist world, the capitalist tools become very, very hard to choose not to use. They’re in everything. You do a Google search, you got AI hitting you up, okay? But most of us think of AI as more of a cloud thing.

We don’t really understand the entire footprint or maybe we’ve reduced it simply to data centers or no data centers. But the footprint of AI is huge.

And if you follow it, if you trace through from the start to the finish, all the different components that make up the ecosystem of AI, you begin to realize that it’s a whole lot more than data centers. So one of my returning guests and very good friend of this program, Erald Kolasi will be joining me today.

Erald, for those of you who do not know you, should know Erald, He’s prolific on Substack and just a real well-rounded guy that knows a whole lot about a lot of things.

And to be more specific, Erald Kolasi is a visionary complex systems theorist, exploring the nexus of economics, politics, energy networks and ecological dynamics. He’s the author of The Physics of Capitalism: [How a New Political Ecology Can Change the World] which we’ve discussed on this program before. With that, let me bring on my guest, Erald Kolasi.

Welcome to the show, sir.

Erald Kolasi:
00:02:11

Hey Steve, glad to be back.

Steve Grumbine:
00:02:14

Man, it’s been a while and you know, we did a couple bang bang and took a little bit of a break but my God, you just cover so much stuff. It’s almost impossible to in good conscious not bring you back as quickly as I can. So I really appreciate you doing this.

You and Jesse Damiani, I believe is how you pronounce his name, recently wrote a really, really deep paper on the ecosystem of AI and really understanding its footprint. And I hope my intro didn’t bastardize this too badly, but it was truly, truly fantastic.

Can you tell us the name and basically what the essence of that paper was and then we’ll dive into the interview?

Erald Kolasi:
00:02:58

Yeah, absolutely. So the Substack piece was called “The Global Energy and Emissions Footprint of AI Diffusion”.

And so the central point of the paper is sort of contained right there in the title.

We were trying to calculate the global material footprint of the AI industry using some new methods and assessments that you don’t see in a lot of other, you know, media reports or even in some more limited studies that have been done in the past, right? So this was a global analysis, meaning we weren’t just looking at, you know, one industry in one country.

It was a life cycle analysis, meaning we were not just looking at, let’s say, the operational electricity or water consumption of a particular data center. We were looking at all of the cradle to the grave upstream factors that are required to produce the physical infrastructure for AI, right?

And that physical infrastructure includes data centers, water reclamation facilities, robotic centric warehouses, and even supporting infrastructure like natural gas power plants, renewable energy systems like solar farms and wind farms, which you need to build out the data centers and supply them with power. So we looked at a cradle to grave life cycle analysis, right?

We looked at how much energy goes into producing the, you know, the minerals and the metals that go into these things, right? What about the steel and the concrete? There’s so much embodied steel and concrete that goes into these facilities.

The semiconductor devices takes a lot of energy and a lot of water and a lot of electricity to produce all of the chips, right? The chips that go into these things, the GPUs [graphics processing units] and the HBMs [high bandwidth memory] and the network switch chips and the optical transceivers, right?

All of these devices have to be produced and manufactured. And the global supply chains are highly entangled and extended for all of these industries, right?

The semiconductor industry, but the logistics industry, there’s so much of logistics is involved in making this world possible, just shipping things from one place to another.

The statistic that I’ve given before is all of the logistics cost of shipping just one extreme ultraviolet lithography machine from the Netherlands to Taiwan, you know, that it takes like three cargo planes and like 20 trucks and like, you know, 40 freight containers. So it’s just enormous global energy footprint of the AI industry. And that is what we were trying to calculate.

And the final headline that we came to was that for 2026, the AI industry would be responsible for about 1.1% of global energy consumption and about 1.2% of all global greenhouse gas emissions. [Wow]

Now, you may, yeah, you may think those numbers are high or low, but if you don’t have any context, a lot of prior assessments that have been done on this, and again, very limited assessments, but they would give you answers that are often like four times lower, right, four or five times lower.

And the reason why is because a lot of those assessments, they’re just looking at the operational electricity consumption in data centers.

Even some of those assessments are only looking at operational electricity consumption in some of the biggest data centers, like hyperscale data centers, colocation data centers, but they may not be including in the analysis a lot of smaller edge data centers, micro data centers, modular data centers, and things like that which are also involved in the AI build out. So this is essentially the analysis that we did. We did a global life cycle analysis for the entire AI industry.

One of the things that we emphasized in the piece is we should be thinking of artificial intelligence as an industry with a wide spectrum of technologies and not just as large language models like ChatGPT or Gemini or Claude or other things that people are familiar with- Copilot. Those are obviously cornerstone of what we’ve come to culturally understand as AI, or what many people have come to understand as AI.

But AI systems really extend far beyond large language models, right? I mean, they go into, you know, self-driving vehicles and autonomous vehicles and industrial robots and autonomous robots and so many other things, right?

All of these things, if you’re willing to have a broad enough definition of artificial intelligence as we do in our piece, we define it as any system that partially or entirely depends on symbolic logic or neural networks or both, then you encompass a wide spectrum of technologies in this industry that go far beyond just large language models, which I think is probably what most people think of when they think of AI. Our popular understanding is sort of like, well, AI is like ChatGPT. That’s what most people think of it. So we go a little bit beyond that.

We have this industry view, we go away from the product level view. And a lot of prior analyses and assessments that have been done in this area, they’re mainly focused at the product level, right?

So they’re focused on like, you know, “Hey, what’s the water footprint of a ChatGPT” query? Or something like that. And we point out that that’s a very narrow way of approaching this problem.

You have to analyze the industry not just like one particular product and one particular action within that product. Because even for so many of these large language models, there’s so many different ways to use them.

Some of them are actually very complex, especially the paid or subscriber versions, and they consume much more energy than like simpler free versions.

So there’s orders of magnitude of differences even within large language models and how you use them in terms of their energy and emissions footprint. So these are some of the things that we emphasized.

And you know, I’ll pause right there to sort of get your take on it and see if you have any follow up thoughts.

Steve Grumbine:
00:08:41

Well, absolutely. You know, it’s clear that it’s not a. We’ve got a, I don’t want to say a flawed, we’ve got an imperfect understanding ourselves.

Regular people are looking at this incomplete… when we analyze what we’re seeing out there. Some of it, I’m sure is industry jargon and industry talking points that keep us from seeing the big picture.

And I think part of it also is a little hysteria, not misplaced necessarily, but maybe not accurate in terms of what it’s, you know, trying to get our attention to focus on. You guys have taken a much more complete, dare I say, materialist view of things for us to really understand how things came to be.

So what I want to ask you, I guess as a follow up to this is we have this growing debate not only within our organization, but you see it all over the social media world, people at work. I mean, I work in an industry where we’re dealing with AI all the time.

I guess the question is, is AI’s growing energy footprint really a technology problem or is this far more about capitalist growth and accumulation here? And to add a little carrot to this, I have no idea what Jevon’s Paradox is.

I know you guys talk about it in your piece, but I think it’s related to this. Can you give us a little bit of background on that? Because obviously I think most people are just like, “Hey AI’s stealing our work.

This is bad for labor, bad for workers.” And a lot of things are probably bad for workers when it comes to a capitalist environment.

I mean, that’s the whole contradiction of capitalism, right? Labor and capital are always at war. So within that space though, it seems like maybe the focus is maybe on the wrong thing.

So are we talking technology or is this an outpouring or outcome of capital?

Erald Kolasi:
00:10:34

Well, so there are many questions sort of embedded in that one question. It’s, you can say it’s a mix of both, but obviously I think the wider story is a story of capital accumulation. Let me answer the…

Before I finish that thought, let me answer the part about the Jevon’s Paradox. So yeah, Jevon’s Paradox has been a fairly well known concept in the history of ecological studies.

And then in this decade it somehow hit Silicon Valley after they, you know, looked it up on Wikipedia or something. And now they’re. They’re mentioning it a lot. So the…

In the context of ecological studies and economics, which is how it’s traditionally been understood, Jevon’s Paradox is simply the observation that efficiency gains under capitalism tend to lead to more aggregate energy use overall. And that’s because efficiency gains tend to lower unit production costs.

So they make things cheaper, and therefore they spur greater downstream demand in the economy, meaning more people want to buy the cheaper things.

And so because more people want to buy them, your unit production costs are lower, you start producing more and more of the things that are in question. So what are some examples? Well, William Stanley Jevons was the British economist in the 19th century, after whom the effect is named.

So he observed that in the 19th century, steam engines in England, they were getting more and more efficient, and yet somehow England was using vastly more coal over time, right? So coal use exploded even though steam engines were getting more efficient. You see this in a lot of other places. So the Bessemer process.

In the 19th century, the Bessemer process came along for producing steel, and it made the production of steel much more efficient. What happened to total steel production after that? It skyrocketed. You see it with the Haber Bosch process for producing fertilizers, right?

Agricultural fertilizers. When the Haber-Bosch process came along, it became a lot easier to produce ammonia. And so ammonia production, fertilizer production, skyrocketed.

You see it with things like, you know, trends in global agriculture in the second half of the 20th century and the green revolution.

As water efficiency increased for irrigation, the agricultural industry started using a lot more water, because making it more efficient just meant that they could go to, you know, further outfield farm new places. And so the end result was higher aggregate use of water. I could go through a million other examples, right?

And, you know, people have read my book or they’re interested. It’s covered in more detail there. I’ve also written about it on Substack for free, so people can check out those posts, some of my earlier posts.

But that’s essentially what the Jevons paradox is. That’s the paradox, right?

It’s that on the one hand, you’re making efficiency gains which should lead to lower consumption, but the efficiency gains actually spur greater consumption. And in the context of this world, you know, it’s highly relevant, for example, for understanding why data centers have exploded.

So in the year 2000, data centers were like 0.1% of national electricity consumption.

Now, in 2026, I don’t think anybody’s published any, like, you know, very recent numbers, but they’re probably around 5 or 6% of national electricity consumption now in 2026. And so they’ve just, they’re just using so much more electricity. Even though data centers themselves have become more efficient, right?

Have become more efficient at taking the energy and the electricity and running the IT equipment, the computing equipment, right? So that efficiency has increased, but yet we’re building just so many hundreds of data centers, right, across this country under construction, right?

And their total electricity load is increasing and their energy and emissions footprint is increasing. So that’s sort of like Jevon’s Paradox.

In a more economic and ecological context… in the AI world, it also often comes up in the sense that, well, as costs go down for AI, like, you know, more people will use it, it’ll diffuse widely, more people will adopt it, right? So it’s sort of seen as an argument supporting, you know, further AI adoption.

But again, it’s kind of a very recent thing and it’s something that they borrowed from, like, these other fields.

And then to the first part of your question there, Steve, about should this incredible growth be understood as, is this a story about capital accumulation or is this story about technology? So at the end of the day, this answer will not shock you. Everything in capitalism is about capital accumulation, right?

So even, even the way technology diffuses and develops is fundamentally tied into the process of capital accumulation, right?

And so, you know, the recent example I gave in another interview I did was I talked about, you know, the soaring computing costs that OpenAI is facing. And because of those soaring costs, they shut down some amazing technologies like Sora.

Sora was their video generator, which came up a couple of years ago, and Sora is now gone because it was really, really expensive to run a video generator. So even the development of AI itself is very much tied into the process of capital accumulation. That applies to all technologies ever.

I mean, look at, look at 3D printing. 3D printing, about a decade and a half ago was seen as like the next big revolutionary thing.

Even Obama mentioned it at the State of the Union in 2013. He said, like, 3D printing was a technology that had the potential to revolutionize the way we make almost everything.

Today, 3D printing has worked out in some industries where it was cost effective. Like, so it has certain niche-use cases. Some companies use it, like Rocket Lab to make their engines.

But for the most part, it didn’t really make a dent in consumer goods or in many other industries simply because it’s not cost effective relative to injection molding or other technologies that exist. And so that’s another example of a technological pathway that got shaped by the wider constraints of capital accumulation, right?

Capitalists, what do they want to do? They want to make money.

And there are various ways they can do that, from micro-ways to reducing costs, increasing revenues to more macro-ways like controlling the political system, controlling the supply and demand cycles of our economies, and so on. So absolutely, the growth in this industry is absolutely a story about capital accumulation.

Take all the ridiculous financial speculation that’s happening. You know, just a few months ago, Steve, you may have heard there was a. There was this shoe company called Allbirds.

And Allbirds was known for making like eco friendly shoes. And all of a sudden they sold themselves in the spring and they announced that they were going to invest in AI data centers, right?

They were going to buy computer chips like GPUs and then basically rent them out to whoever wanted to use them. And the funny thing is, after they announced that, Steve, the stock price of the company went up like, what was it, 7 or 800% in a single day.

So it was, yeah, it was, it was a very successful marketing gimmick at convincing investors that this company is now heading in a new direction because its shoe business was failing. And so they made this announcement. The stock was up 700, 800% that day.

Since then, the stock has come back down because everybody realized this company has absolutely no technical expertise whatsoever to do anything with data centers or semiconductors or AI or anything like that, right? I mean, this was a shoe company that just randomly decided it wanted to enter the data center space.

And it’s been done by other players who are sort of closer to the action, like the crypto industry. You may remember a few years ago, the crypto industry was actually doing crypto mining, right? Like bitcoin mining. Uncover new bitcoins.

But, now the crypto industry itself has heavily shifted towards building data centers, right?

Or towards actually, you know, buying GPUs and putting them in their crypto facilities, converting the facilities, in other words, from bitcoin mining to, you know, AI related workflows. So it’s kind of happening everywhere. I mean, there’s ridiculous speculation. There’s obviously huge amounts of debt fueling this thing.

You have companies like CoreWeave, you know, that have like 8, 9, $10 billion of annual revenues and like $50 billion in debt or something like that. 40, $50 billion in debt. Just like companies taking on enormous amounts of debt.

The level of financial speculation has gotten so bad that even companies like Google, established giants and monopolists, they’re doing public offerings for like $80 billion, which was shocking when Google did that. They’re just, yeah, they’re just basically creating $80 billion of new stock to sell, right?

And so from the public offerings to the stock buybacks, to the bond issuance, right? There’s so many corporate bonds going out now to fuel this infrastructure boom, you know, related to AI, especially companies like Oracle, right?

Oracle right now is taking on enormous amounts of debt as part of the Stargate project with OpenAI and SoftBank. But they’re building out all of these data centers, including a massive hyperscale campus in Abilene, Texas.

But Oracle’s taking on enormous, enormous amounts of debt. And you know, this is, this is how all bubbles in the history of capitalism pop, right?

People take on enormous crazy levels of debt that they will be unable to service because the debt is just vastly greater than the revenues they’ll be able to bring in the future. And so eventually this house of cards will collapse, just like all the others have in history, right?

Whether you’re talking about the dot-com bubble, the railroads back in the 1870s and the panic of 1873. So this one too is going to go that way, right?

It’s fundamentally, you know, a lot of people talk about the capabilities of AI and what it can do on your screen. This is fundamentally a story about capital investments, right? Capex, right? Physical infrastructure.

And that was one of the main themes of the piece as well, right? Stop thinking of AI as something that sits on your screen. This is a story about industrial manufacturing and logistics and mining and refining.

This is a story about the capitalist system as a whole, the political economy that sustains it, and all the laws and regulations that are being skirted, broken, blasted, so you can build all of these things out. And that’s how this story should be seen, right?

It is absolutely a story about class power, capital accumulation, political corruption, and fundamentally manifested in this massive material industrial scale.

Steve Grumbine:
00:21:22

So I want to ask you another. That’s fantastic. I had so many little anecdotes I wanted to add in there.

And one of them was I used to be a systems engineer for Verizon during the dot-com bubble.

And I remember selling huge amounts of backhaul, all kinds of network to these companies that didn’t have even a single customer, not a single customer. I remember flying to Florida to sell this guy a 25,000 user DSL resale arrangement. This guy was literally In a dorm room at Florida International.

He wasn’t even like a real… He decided he was going to start an ISP, and I don’t think he fulfilled 100 of the actual seats that he had leased for three years.

I mean, the entire market space here is insane. And one of the things I wanted to ask you, and this is going to seem really remedial, but I think it’s really important to nail this down.

We talk oftentimes about, you know, labor and capital and stuff like that, but frequently I see so many people mistaking capitalism and capital as just markets. It’s not it at all. Can you please describe, in your own way, what it is that capitalism means to you? How would you define capitalism?

Erald Kolasi:
00:22:45

Sure.

So, I mean, the way I would think about it at a, as a sort of first order approximation working definition is any economic system dominated by the concentration of private property in the hands of a few powerful people. That’s a working definition, right?

You could define, and people have defined capitalism in a million different ways over the years, but that’s sort of how I think about it at a high level. And even capital, right? The fundamental term behind it, it’s controversial in the history of economics.

I mean, you can think of capital the way neoclassical economists do, as like a factor of production, as like a, you know, a factory or a piece of industrial equipment or, or something like that, right? You can think of capital the way Marxists typically do, which I think is a better way of thinking of it as like value in motion.

And so the going through the different stages of production. So you can start out with a certain amount of money and end up with more money in the end, right?

And so Marx emphasized that capital is never a static thing. It’s never just one thing. It’s never like a piece of equipment or just like one worker.

It’s the whole social relation sort of in motion, the whole social and productive relations in motion that drive capitalism. You can think of it like that. You can think of it. Other people have thought of it as more like a symbolic expression of social power.

That’s also a view that I find useful in many situations, right?

So in this view, the stock market or the world of money in general, whether it’s wages or interest rates or debts or prices or whatever you’re talking about the nominal domain in general, it is a reflection of the social power dynamics that are at the heart of our civilization, right? So for example, Steve, to just give a few examples. Right now, AI, these models, they’re often very cheap or free to a lot of people, right?

Why are their costs so low when they should be a lot higher?

Well, that’s because they’ve been artificially subsidized by big tech, including companies like Nvidia, which, as you know, has enormous vendor and circular financing deals with a lot of people, right?

So, you know, Nvidia will give OpenAI, like billions of dollars, and then as an “investment” and then OpenAI will take that money and buy Nvidia’s chips.

So there’s a lot of circular financing and vendor financing, a lot of subsidies happening right now, which are in effect depressing the cost of this thing. And so therefore, you could argue that the low cost of AI right now is a fundamental reflection of the power dynamics in society, right?

That like really powerful people have decided that this product should be relatively cheap because they want it to diffuse widely, because they want it to be widely adopted. And if they had decided that anything else should be cheap, we could have that too, right?

So if the right people were in power, healthcare could be cheap, right? Or any other thing could be cheap.

But the people who are in power right now, they really think that, you know, it just goes to your fundamental point that you were making earlier, Steve, which is that, yeah, capitalism is at the heart of this, right?

So the concentration of economic power in the hands of a few private individuals, in the hands of a few monopolists and oligopolies, and that is ultimately the root problem.

And the fact that once a new technology comes along that they think is going to change the world, forget for a moment that it’s not going to do that in many ways, right?

It’s, yes, it’s going to change many aspects of the economy, but it’s not going to change a lot of fundamental problems that we face in the world, like warfare or poverty or disease or inequality, because AI is a technology that serves capitalists. It is going to become another elaborate tool or another elaborate set of tools for capital accumulation.

So these people are just going to use it to get richer and to get even more powerful and to try and sabotage labor even more… so.

Steve Grumbine:
00:26:55

So with that in mind, that’s a really important point you bring up, and it leads me right to the next one, right? So when I think about capital accumulation, people always just think it’s about making money. And yes, that is the end goal. That’s what they want.

They want that. But they also use these things as a means of ensuring that they can gain more power and more money.

And that’s why we’re seeing some of the surveillance and things like that, but in the handful of enormous corporations, what does that tell us about monopoly ownership and power? I mean, for real, what, what, like I want people to have a better understanding of the world they live in.

It’s not just, I mean, me and you may differ on this one, I know we do a little bit, but may think that we’re just going to go down slap and I voted sticker on their forehead and voila, they’re going to be able to change society, baby, right there at the ballot box. In reality, these powerful capitalists, you can see the revolving door from Wall Street to government and back again. You can see it in every sector.

I have my good friend Bill Black, who is a genius when it comes to the legal stuff, and he’s been a, he calls himself a serial whistleblower. But even Bill and I strongly disagree on this.

He believes that we’re just going to vote our way out of this and you know, hey, we can, we just got to get the right party in there or whatever. But the reality I’m seeing is that government and industry, they’re not separate entities, they are one.

And the powerful, which again, I’m going to keep harping on this until someone tells me I’m full of shit. And if I’m full of shit, I’ll change on the spot.

But that Gilens and Page study in 2014 from Princeton showed that the desires, the hopes and dreams of the average everyday voter, even not just the voter as the individual, but collectives of interest groups, have a statistical near zero impact on anything. Not some things, anything.

And so when I look at these powerful groups that are not only getting seed money from government contracts and the military, et cetera, the fusion of AI industries in Silicon Valley and the Department of War and so forth, we’re not talking about like, well, shit, we don’t like that. We’re just going to vote that away. That’s not happening. That is not happening.

So when I think about power and I think about AI and I think about surveillance, they’re not doing that to protect us from, you know, some poor guy trying to steal an apple from the local grocer. They’re doing that to ensure that they can maintain control. Because this is what capital in crisis does, it turns fascistic.

It uses the state to discipline labor and AI has been serving in that facet… I mean, you see the open air test grounds in Gaza for a lot of these ideas.

And then you see the police using it now with ICE so what do you think about the concentration of power not just in capital, but at the government level?

Erald Kolasi:
00:30:09

Yeah, yeah, absolutely. Steve, I think, you know, I agree with most of what you said. I think you hit the nail on the head. That is the fundamental problem, right?

And look, there’s always a fundamental synergy in capitalism between dominant corporations and the state. So that has always been the case, and that will always be the case as long as we have capitalism. And you’re seeing it right now with AI as well.

Even though at the local and state level there’s a lot of opposition, you know, rising against AI and against, against the building of new data centers, and some states now are putting like pauses and moratoriums. Nevertheless, the federal government, which is run by Donald Trump, at least the executive branch, but it feels like he runs all of it now.

So the federal government is very much in favor of pushing for this build-out and very much aligned and united with Silicon Valley in, you know, and all of its accelerationist nonsense about going as fast as possible, possible towards the future, towards ASI… artificial superintelligence, towards making sure that we can beat China, right? This is always, they always hit on China is everything, right?

Even Mark Zuckerberg in his latest propaganda piece or manifesto as some people called it, you know, the future is for everyone, he made sure to emphasize that we want America leading the AI race, right? We want America to be first. You know, America should be the one innovating. America should be out front.

And so, yeah, you’re, you’re absolutely right.

This isn’t, you know, when I talk about, like the concentration of power and wealth in the hands of private individuals, of course that concentration fundamentally depends on their control over the state. The state is often the major dominant conduit, the method for supplying a lot of that capital accumulation, right?

For breaking down resistance to labor. It was the mining companies, you know, with the Battle of Blair Mountain in West Virginia. How did they get out of that little pickle?

Well, the federal government sent over, you know, some forces and the National Guards, and so they sent over the military, right? And that’s sort of what, like, brought the Battle of Blair Mountain to a halt, right?

So, yeah, of course, capital fundamentally depends on the state to discipline labor, right?

Intermission:
00:32:20

You are listening to Macro N Cheese, a podcast by Real Progressives. We are a 501c3 nonprofit organization. All donations are tax deductible. Please consider becoming a monthly donor on Patreon, Substack, or our website, realprogressives.org. Now back to the podcast.

Steve Grumbine:
00:32:42

The Washington Post, which I stopped subscribing to some time ago, but they did some good work and they found out literally over Musk, you know, had pulled in a total of 38 billion in total government funding. Think about that for a minute. 38 Billion. And this didn’t just happen on Trump’s watch. Just so we’re on the same page, this was bipartisan fuckery.

Biden was big time on it. And I wanted to state it’s not just Donald Trump screaming China and MAGA screaming China.

At the State of the Union address, Joe Biden, I don’t know how he said it because he could barely tie his shoes by that time, said that China was the number one enemy that we had. I mean, he called it out, right? So this is not a partisan issue. This is a capital issue. And both parties live and die by capital.

I’m just curious, you know, when we talk about this, I want to take it a step further because, you know, we often talk about colonialism here and anti-colonialism and anti-imperialism, and you think about where all the rare earth metals come from and all the real resources that come from this stuff.

And we frequently, even not speaking of AI, but speaking of capitalism as a whole, notice the absolute decimation of the Global South and the pillaging of their real resources and the debt slavery that the long arm of empire with the IMF [International Monetary Fund], etcetera, has imposed upon them to facilitate this kind of resource transfer.

You know, with that in mind, if you think about energy, water, minerals, labor, productive capacity, those are the real constraints we talk about in the MMT world. But in reality, we the people literally have zero say in this. I have never, ever seen real say…

In whether we kill people in Gaza, whether we rape and pillage the Congo, or whether we go down to Venezuela and snatch a democratically elected person, I mean, it doesn’t matter whether you like them or not.

I mean, the fact of the matter is, is that when I think about the real resources and who controls them, I mean, obviously you and I share a similar belief in the sense that we don’t believe that the real resources should be controlled by capital, but today they are in fact controlled by capital. How do you, how do we look at that?

I mean, can you explain the real resource snatching, how this whole system of snatching up these real resources actually happens? I mean, the ugly side.

And we can look at, you know, the confessions of an economic hitman we had him on not too terribly long ago as well, just sort of discussing these things. But what is your thoughts?

I mean, obviously it’s not democratic control of real resources, private property and capitalist relationships are baked into all the legal documents that control this country. So no matter how socialist you are, you’re governed by laws that are 1 million percent designed to be controlled by capital. Your thoughts?

Erald Kolasi:
00:35:39

Yeah, absolutely. I mean, I think there are a lot of good threads there, Stephen. I don’t know that I can follow all of them, but just to give.

Steve Grumbine:
00:35:46

Sorry about that.

Erald Kolasi:
00:35:48

Just to give a bit more context, right.

So, you know, the angle on imperialism and neo-colonialism and, you know, neo-imperialism, whatever you want to call it, I think is absolutely on point, right?

I mean, a lot of these, you know, resources that, for example, are going into this infrastructure boom, yes, some of them might come from the United States, some of them might come from China, but a lot of them are coming from places like sub-Saharan Africa, right, or other parts of Asia. And those are very much countries in the periphery of the world system. They’ve been exploited for a long time.

Western corporations routinely bribe government officials in these countries to give them subsidies and sweet tax deals so they don’t actually have to pay taxes locally.

So as a result, a lot of these, like, fundamental resources are being extracted from many of these countries, but the wealth from those resources is not really flowing to these countries, right? It’s just the natural resources and the raw materials are being extracted from these countries.

And then, you know, the upstream manufacturing is happening abroad and then the final finished goods are sold in various markets around the world. And it’s Western corporations profiting from those sales, right, as part of the global value chains that that many of them control.

So I think there’s that angle and dimension to it. And I think you were right to point out that the new cold war that we seem to be starting with China is supported sort of by both parties.

I mean, you could find even Obama was placing tariffs on Chinese tires when he was in charge. And so I think this is definitely a bipartisan effort, right?

The ruling classes, they see China as a fundamental threat to American power and to America’s position in the world, right?

And that’s part of all of the things that we talked about, all of the, you know, the global trade wars and the tariffs and the remilitarization of East Asia and Japan rearming and all of those things. And now with AI, right? You know, this is a major, major concern right now.

There’s a lot of open source Chinese models which are much, much cheaper than a lot of the frontier models from the, you know, leading AI labs here in the United States. So those Chinese models are nearly as good as the frontier models here in America, but they’re much, much cheaper, right?

And so as a result they’ve diffused very widely, right?

There’s a lot of people using them even all over the Western world, certainly in China obviously, but even in the United States there’s a lot of people using these models like Kimi K3 and so on, you know, DeepSeek-R1 and even in Europe and other places, right? And this is a huge threat to Silicon Valley and I would say it’s probably the leading threat right now to the AI buildout in general.

It’s if people switch en masse to these much cheaper or many cases free Chinese models, then what’s going to happen to the AI buildout, right, which is now the core engine of economic growth in the United States. And what’s going to happen to the huge debt overhang, right, to the $1 trillion plus in capex [capital expenditure] that’s been poured into this.

So I think that is definitely a threat.

And Silicon Valley, you can tell just again from what people are saying publicly and privately, they’re very concerned about these open source Chinese models. And again, that ties into the broader geopolitical rivalry between us and China right now. And I also wanted to mention on Elon Musk.

Yeah, you’re right. Something people don’t realize about Elon Musk is the reason why we have this corrupt capitalist today.

The main reason why, unfortunately, is the United States government, right?

So SpaceX was about to go bankrupt and then in like late 2008 it gets like a one and a half billion dollar grant from NASA to supply the International Space Station, right? So, so NASA essentially saved SpaceX, otherwise there would be no SpaceX today. Even Tesla.

Tesla also really wasn’t going anywhere until in 2010 it got a $500 million loan from the Department of Energy and it got that loan at a very low interest rate that it would not have been able to find if it had to go to like a private commercial bank. And through that loan, Tesla was able to complete the Model S factory in California. And the rest, as they say, is history, right?

So yeah, I love that you mentioned Elon Musk. Musk is obviously a capitalist pretender, you know, but really a government leech.

Steve Grumbine:
00:40:10

Yes.

Erald Kolasi:
00:40:12

Who would be nothing without public support.

Steve Grumbine:
00:40:14

So, you know, I kind of took us a little bit out of the paper, but I wanted to give you a chance. I have a follow up question after this one, but what would you say is the most important things to take from your paper?

Like we’re going to put this in the show notes, we’re going to make sure everybody has access to this because it’s really important. But what would you say is the most important things out of your paper? Perhaps even things that we didn’t cover yet.

Erald Kolasi:
00:40:41

Yeah, thanks, Steve. So I’ll highlight a couple of really important things.

One is the fundamental importance of using a life cycle assessment for analyzing the ecological or environmental impacts of industries and economic sectors or even countries, right?

And the reason why that’s important is because right now there’s a lot of debates happening about AI and energy use and water use and electricity consumption and all of that. And to me, a lot of these debates feel fundamentally misguided because they’re often focused on, you know, the operational on-site consumption, right?

So people will say things like, “Well, you know, most data centers, they don’t consume a lot of water on site.” And then people will say things like, “Well, some data centers have closed-loop water cycles, right?”

Where they just recycle the water and they don’t consume it from, from like, you know, the local environment, like a, you know, public utility or a lake or a river or anything like that. So people will say a lot of things like this, right?

And to understand why that view is fundamentally misguided and why a life cycle assessment is mandatory, not just optional or nice to have, but mandatory, is because you would reach very naive and pseudo-scientific conclusions if you apply that standard to many other industries. So take a look at renewable energy systems, right, like solar panels or wind turbines.

If you just look at their operational emissions footprint, in other words, their carbon emissions while they’re actually running and generating electricity, well, the answer to that is it’s practically zero. They don’t have a carbon footprint while they’re running.

The only way you would know that the renewable energy industry actually has a major ecological impact is by studying its life cycle dynamics, right?

Is by studying the mining and the refining and the processing and the logistics and the manufacturing and the installations that are required to create and build solar panels and wind turbines in the first place, right? In other words, you need a life cycle assessment to fully understand the complete picture.

And so that’s one of the major points that we make in the piece, right? And this applies to anything, even the people who talk about water, right?

Yes, the on-site operational water consumption of many data centers isn’t really that big at all and it’s dwarfed by other industries. But if you go upstream, right? And you look at the water consumed by power plants during the process of cooling to help them generate the electricity.

The water consumption there from power plants is enormous from the power plants that are producing the electricity supplying the data centers. So they actually have a significant indirect water footprint at the power plants.

And then you can’t stop there, right, because you have to do a life cycle assessment. So then you have to look at, you know, the life cycle and embodied water of everything that goes into the data centers.

It takes enormous amounts of water to produce steel, cement, concrete. It takes enormous amounts of water to produce semiconductors.

So all the semiconductor devices that end up in data centers, huge, huge quantities of water, right? The cooling systems, the pipes, the tubes, the supporting infrastructure, like the natural gas turbines, right?

Aeroderivative gas turbines, the batteries, the switch gear modules, so many things, all of these things have a water footprint.

And so when you broaden the analysis like that, you realize that the total water footprint of data centers is actually, it’s actually pretty significant, a pretty significant share of national consumption, probably between 1 and 2%. So again, dwarfed by agriculture and other bigger industries, but much higher than other sort of naive conventional estimates would suggest.

And that’s because you need to look at this life cycle embodied footprint. And the other thing that I would emphasize there is a lot of this life cycle footprint is right here in the United States, right?

So for example, a lot of the components, devices that end up in data centers, they have copper in them. Well, a lot of that copper is mined right here in the United States. And it takes enormous amounts of water to get that copper ready.

And then it might be shipped abroad to East Asia for final manufacturing and then it comes back as a component and it ends up in a data center here somewhere. But that’s the kind of point I’m making, is that a lot of the lifecycle and embodied footprint is domestic, right here in America.

So it’s resources that are being consumed right here in the United States, and then they might enter the global value chains and then they’ll eventually come back and enter our data centers. So that’s one of the major takeaways of the piece, is that you have to do a global life cycle assessment.

If you’re going to understand the true ecological impact of this industry.

You can’t just stop at a ChatGPT query or what this one data center in your backyard is doing, because if you do that, you will necessarily get a ridiculous incomplete picture. Just like you would conclude that renewable energy has no ecological footprint whatsoever.

Even though it obviously does, because it’s pseudoscientific to say that. Anything material must have some kind of ecological impact because you have to get it from somewhere, right? You have to produce it somehow. That’s what the lifecycle assessment gives you. It gives you that broader complete picture.

So that I would say is probably the biggest takeaway and then other takeaways that I want to highlight is we had some new analyses in this piece that other people hadn’t thought about before. So for example, there are huge emissions that happen every year worldwide as a result of land use changes.

So Steve, when you clear away land, you’re clearing away forests and grasslands, or maybe wetlands and peatlands and marshes, you’re clearing away natural habitats that are already there in order to build real estate properties or commercial developments or data centers.

And likewise, you have to clear away land to build the power plants that are supporting the data centers, like 80 gigawatts of natural gas power plants under construction right now in the United States to support the data center buildout. All of those things emit… So there’s huge emissions associated with all of those land use changes.

Because once you expose all of that soil to the atmosphere, you know, there’s a lot of oxygen powered bacteria, so aerobic bacteria that start, you know, consuming a lot of that organic material, so it releases carbon, combines with oxygen, carbon dioxide, and then that floats into the atmosphere, right?

So even just land use changes, just changing what the land is being used for from like a natural habitat to a data center, there’s enormous ecological costs associated with that. And of course not just with data centers, but with all kinds of developments in general.

In fact, damaged peatlands, which are an important critical kind of wetland, they’re responsible for 5 or 6% of all global greenhouse gas emissions. Just the peatlands that we’ve damaged because of all of the developments that we’ve done, right?

So that’s a major issue that a lot of people hadn’t really flagged before, because a lot of people doing this analysis in the Silicon Valley world, they’re computer scientists, they’re programmers, they have an interest in sort of downplaying the ecological and environmental impact of AI and telling you it’s not a big deal. But they don’t really have a lot of experience with ecology or physics and sort of shows in a lot of the analyses that they’ve done.

And so our piece was aimed at sort of giving this broader view to this industry so people understand its material footprint on a global level. Instead of getting lost in the weeds about what’s happening at this particular point or there.

So those are some of the main things I could go on and on, obviously, about this piece, but to hide from sort of technical enhancements that we made and other interesting things. And then we sort of talked about it at the beginning.

The biggest conceptual leap is getting away from thinking of AI as just a product and understanding that this is an industry. It’s an industry with a broad spectrum of technologies.

It’s an industry with enormous levels of investment behind it, enormous levels of financial speculation behind it, enormous levels of political power behind it. And so those would be the main takeaways from the piece.

Steve Grumbine:
00:48:40

That is absolutely fantastic. I want to also add that I’d love to see if I can get Jesse [Damiani] on here and talk to him at a future time, and I will reach out to him as well.

The final one, I’m hoping we go out slightly more positive note of sorts, is to kind of envision an alternative view of this. You know, I talked to. This was my take, and it’s, mind you, it’s very rough around the edges.

So pardon me, but my thought was, you know, the difference between capitalist AI and what I would consider to be a socialist AI would be somewhat radical in its nature. It would be publicly owned. I mean, all the people would benefit from it.

If there was a way to make work less stressful, less grueling for people, it would be there to benefit people in general, but it would also not have the scourge of this race, this ridiculous amount of groups going and building the same exact thing over and over and over again, each one, you know, exponentially increasing that footprint. There’s a lot of slop in there that goes with competing against each other. We’re fighting against each other.

A lot of, you know, bad things, bad externalities that maybe aren’t thought of just sort of like your whole piece describes.

Can you kind of paint what a vision of a publicly-owned AI might look like and what the ecological impact might be, how it might differ from this capitalist version?

Erald Kolasi:
00:50:12

Absolutely, Steven. And first, let me start with an actual practical case before we move on to more theoretical stuff, right, and that case is China.

China has hundreds of data centers. We have thousands. China’s level of investment in AI is far, far, far, far, far, far lower than America’s level of investment in AI.

And yet with far fewer resources, they’ve been able to build AI models that are nearly as good as our leading frontier models.

So already, if you want a practical example of how you can do this, without upending everything, without destroying our ecosystems and polluting our air and just setting us back to the days before, you know, the EPA and the Clean Air Act, then China provides like a relatively good example. I’m not saying it’s a perfect example, right? It’s definitely not.

But there are ways of enhancing and developing AI and even having it diffused broadly across society that are not so energy intensive and that don’t have such a huge emissions and ecological footprint, right?

That aren’t so damaging to the global biosphere at a moment where we’re struggling with record global warming and all of the massive consequences resulting from that, right? So China, I think provides like a good practical example. Again, not perfect, but in terms of like what are ways to make this better.

So yes, I think bringing a lot of this infrastructure, a lot of this buildout, a lot of these companies under public control is absolutely an essential first step. There’s so many other things you can do, right?

One I think is, you know, we as a society, and then this needs to get enacted at a government policy level, but we in a society need to understand that AI is a great technology, but that’s all it is, right? It’s just another technology, it’s just another form of software optimization.

It’s not something that’s going to fundamentally change the world and fix everything, right?

And I think this is a lot of the silly propaganda that is often repeated about AI to justify the enormous levels of investment, financial investment and infrastructure investment and capex that are happening right now in the United States.

You’re talking about people like Google’s DeepMind’s CEO Demis Hassabis, saying that like, you know, the rise of superintelligence is going to cure all disease. He said this in a 60 Minutes interview, right, that you know, once we have ASI, [artificial superintelligence] it’s going to lead to the end of disease, right?

And of course it’s not going to do that, right? It doesn’t matter how many protein structures AlphaFold discovers, It’s never, we’re never going to get rid of disease.

You know, biological systems will always find a way to evolve. I always bring up the example of penicillin and antibiotics.

After we flooded the world with penicillin and antibiotics, we had the rise of drug resistant bacteria and they kill millions of people every single year. So there’s a lot of these fundamental problems that the AI boosters and the Silicon Valley accelerationists are saying AI is going to solve for humanity.

Even we’re going to fix climate change and have space colonies is what, you know, Sam Altman said, none of that stuff is going to happen, right? And so this technology, it’s not going to take everything by storm.

I think right now there’s just a ridiculous level of overinvestment in it because we’ve kind of become addicted to it. It’s the core engine of economic growth, and the ruling classes don’t know what else to do.

But, yes, in a different society, we would roll back a lot of this slowdown, right?

Like cancel a lot of these data centers and actually invest in things that would make people’s lives much, much better, like free and universal health care or child care, or, you know, public education, public housing. You know, we talk about the housing crisis in this country and how expensive housing is, let’s invest in that.

Instead of investing billions in data centers, can we invest some billions?

Can the government invest some billions in building houses for people and in building clean, habitable communities for them with parks and shopping centers and schools and other things like that? Why don’t we invest? Why don’t we do that, right? Why do we invest in our people instead of investing in a few corrupt capitalists in Silicon Valley?

I think that’s the way to go, Steve.

Steve Grumbine:
00:54:32

Well, I agree with you 100%.

I mean, like, at the end of the day, I think it’s important that people, especially for the people that listen to this podcast that are socialists at heart and are wanting these things. It requires more than hope, right? It requires more than putting an I voted sticker on your forehead.

We’ve got to organize and we’ve got to work together, and we’ve got to really become committed to making socialism happen as opposed to just kind of putting I voted sticker on, talking about how, ah, we fought the good fight, door knocking and phone banking.

I mean, there needs to be much more focus on what we have in common and understanding the struggle, the material reality of working class struggle today that is being absolutely exacerbated by this capitalist race to the bottom with AI. I mean, it is right there for all of us to see.

And if you don’t allow your lying eyes to deceive you, you’ll realize that whether it was Obama crushing dissent at the Dakota Access Pipeline, whether it was W saying “victory,” you know, whether it was Bill Clinton throwing “welfare queens” off of benefits, whether it was Reagan destroying the air traffic controllers, or whether it was Jimmy Carter saying, “We gotta tighten our belts”. Folks, we gotta do more than thinking an I voted sticker is doing the job.

And, you know, as it is I hear people already pre-capitulating saying, “You know, Trump was so horrible, if Gavin Newsom runs, you know what? “I’m… count me in.” And I’m saying to myself, “Can I show you this video of him celebrating destroying homeless encampments?

Can I show you video of him celebrating putting spikes in benches in parks to keep homeless people from sleeping on them? Can I introduce you to this guy?” Because this is not the answer. This is not solving these problems. He is business friendly, too. All of these folks.

Elizabeth Warren, “I’m a capitalist to my bones.” And this is the problem. Capitalism. It’s not just the great man theory. Yes, Donnie, tiny hands, bad whatever. But Joe, I can’t tie my shoe Biden, was doing the same stuff!

Like, wake up, Peter Pan, Count Chocula, like, grow up! This is to me a really important time.

And, you know, as a father, and you’ll appreciate this, I hope, Erald, as a father of a special needs child, I look at these things and I say to myself, are we creating a kinder, gentler, more inclusive society that will build structures of support? Are we building things that are creating capitalist accumulation and destroying the biosphere and making it crueler and more difficult to survive?

And I think the answer is quite clear. I’ll give you the last word.

Erald Kolasi:
00:57:25

Yeah, I mean, yeah, couldn’t agree more. Capitalism is the problem and socialism is the solution. If we can find a way to get there, and eventually I think we will.

So, yeah, I mean, I think you summarized it pretty well. Right now, we’re not heading in a good direction. Not just with AI, right?

AI, at the end of the day is a symptom of the rot and corruption that’s at the heart of the system.

And obviously, if we had a different kind of system, it would be producing better outcomes that actually serve the people instead of a few corrupt billionaires.

Steve Grumbine:
00:57:54

Amen. All right, thank you, Erald so much. I appreciate this. And I can’t wait to have you back on.

I’m already thinking about things to have you back on for. Thank you so much for doing this. I’m going to go ahead and take us out. Folks, my name is Steve Grumbine.

I am the host of Macro N Cheese, and I am the founder of the nonprofit Real Progressives. We are a 501c3, not for profit, and we are looking for your donations, man. We survive by them. We don’t paywall anything.

I know it’s probably, you know, one of those things where if you pay for something, you’ll take better care of it. You’ll treat it more precious.

We try desperately to make the information available to all and so if you consider the work we’re doing valuable, we need your donations. We do. We desperately need them. To help keep the lights on. You can go to our website, realprogressives.org you can find the donation link there.

Become a monthly donor. You can go to patreon.com/realprogressives, become a monthly donor or a one time donor there as well.

And you can go to Substack, where we are nowhere near as prolific as Erald, but we are definitely out there and we would love to have your support again. Nothing we do is paywalled. Everything is free for the public because we believe the information is that important.

So with that, on behalf of my guest Erald Kolasi, who I hope you will go find his work on Substack. It’s worth every second of your reading. He is prolific. Go read it. Go check it out. Subscribe to him, but with that. My name is Steve Grumbine.

On behalf of the podcast Macro N Cheese, my guest Erald Kolasi. We are out of here.

End Credits:
00:59:35

Production, transcripts, graphics, sound engineering, extras, and show notes for Macro N Cheese are done by our volunteer team at Real Progressives, serving in solidarity with the working class since 2015. To become a donor please go to patreon.com/realprogressives, realprogressives.substack.com, or realprogressives.org.