النص 🔗
speaker-0: These aren't just tech facilities anymore. They're the new steel mills of the digital age. We're essentially subsidizing the AI boom through our electric bills. Are we gonna start having to pay more and compete against these data centers? Supply can't keep up with this ravenous, constant demand. Data centers aren't just using water. They're concentrating toxins. Where does the water come from? What do you then do with that? Wastewater.
speaker-1: Could we like collect all the heat and shoot it into space? I'm I'm not an engineer.
speaker-0: Yeah. How much water do you think Ampi's using? Every time I process your question, I
speaker-1: And burning
speaker-0: threw about sixteen ounces of water.
speaker-1: We better be fucking sure what to ask me a be next.
speaker-0: Getting dark. Welcome to Amplified Intelligence, the show dedicated to helping you make sense of the sci-fi future we're all suddenly living in. I'm John here alongside with Pietro and our resident AI Ampi. Today we're breaking down how the explosive growth of AI data centers is draining massive amounts of global power and water, which is resulting in regulatory and consumer backlash. Okay, so before we dive into the AI data center explosion, it's time to load the matrix. To be fully transparent about our research, all linked into the show description, we've split out our sources into three camps: the Against Camp, the Fore Camp, and the Outliers. First, in the Against Camp, anchored by a World Resources Institute report, these sources detail how massive data centers are draining local power and water. They cover the growing regulatory and private backlash, including New York's hyperscale moratorium. Noise pollution lawsuits, and polls showing 70% of Americans oppose local AI data centers. Next up in the forecamp, defending the industry. These sources argue that data centers actually subsidize public grid costs, creating economies of scale that can lower retail electricity rates. They also highlight big tech's massive financial push to fund new nuclear and renewable energy projects. Finally, the outliers, the weird and wild solutions. Highlights include the heavily funded race to launch data centers into low Earth orbit by 2030 for free vacuum cooling, sinking servers to the ocean floor, and also unprecedented ways cooling centers could stop discharging forever chemicals. The Matrix is loaded, let's get into it. Pietro, do you have a data center in your backyard yet?
speaker-1: Not yet. If I did, I'd be careful about the water usage, for sure.
speaker-0: Right, right. So, you know, this is like a pretty epic, I think, important episode today. You know, the data center problem has been something that has been coming up over and over again. We often hear people, you know, talking about that being a major environmental concern. There's lots of issues around the converging sort of forces of climate change, cost of cost of living problems, you know, the energy crisis where we're sort of in this transition to sustainable energy, but sort of not really, and they're still selling us cars that burn million year old dead life. So yeah, it's it's it's it's a real underline, I think, to a lot of the global crises we face. Yeah. and it's a and and these data centers have There's no stop and end to how much data processing and compute we seem to think we're gonna need.
speaker-1: That's true. Like there's there it's it's an endless amount if we predict out where where the future's going. And truth be told, you know, not like I'm an expert in many things, but I am very ignorant in this in this subject. And so yeah, as scary as some of this will be, I'm looking forward to learning about it.
speaker-0: Mm-hmm. Yeah, for sure. I think data centers in general, like, you know, they've been around for quite a long time. I think
speaker-1: True.
speaker-0: the the question we kind of want to answer today, because the target has really been ple be been placed on this massive industrialization of the cloud, this massive transition to AI compute processing, which is sort of like sped up in the last few years. I think the question we want to ask ourselves is like, you know, is the economic increase that we're gonna get from AI or that we're expecting to get from AI worth all these struggles right that these data centers are gonna cause in people's communities, are gonna cause you know, in their groundwater, yeah, are gonna cause potentially against their, you know, price of energy and their their cost of living, which is already so difficult.
speaker-1: Yeah, it feels like the water factor is particularly tough because it's something that for sure we well, I mean, when when new construction or whatever goes in it disrupts the water and that's not new, but like the water p factor seems particularly disruptive because b you know, I grew up on a farm and there was always a threat of, you know, the dump is coming and you everyone's gonna have to move. You know, that that's around that's just the nature of land and right and development. so that's not new, but I think the water thing is particularly what I'm I'm s freaked out about.
speaker-0: It's certainly the thing that's under a microscope right now, and the thing that a lot of the environmental movement's energy is pushing against and pushing towards. And I'm really curious to sort of find out why.
speaker-1: Yeah, me too.
speaker-0: So, Anthony, tell us a little bit about data centers, why why we should support them, why we should not support them. Talk a little bit about all the conversation that's going on right now and and what we should expect for sort of the next decade of you know AI development. This is where the rubber meets the road on AI infrastructure. We're standing at the edge of a massive industrial transformation. Copper wires snaking through concrete bunkers, servers humming in endless rows, cooling towers releasing steam into desert air. These aren't just tech
Load full transcript
facilities anymore. They're the new steel mills of the digital age. The numbers are stumbering. We're talking about four forty eight trillion watt hours annually. That's more electricity than entire nations consume. Water usage? Five million gallons daily for a single hyperscale facility, equivalent to a town of fifty thousand people. But here's the tension. Every AI query you make consumes sixteen ounces of water. Every video conference burns through sixty ounces. The physics are unforgiving. Heat rejection through evaporation.
speaker-1: Thermodynamic.
speaker-0: Demanding their tribute. Meanwhile, communities are pushing back hard. Virginia residents seeing electricity bills spike by two hundred and sixty-seven percent. Aquifers dropping in Oregon, nitrate concentrations spiking when data centers evaporate the clean water and leave behind toxic sludge. New York just dropped the hammer. First, statewide moratorium on hyperscale permits. They're saying, no more subsidizing this growth on the backs of ratepayers and water tables. We're essentially asking, who pays for the physical reality of our digital dreams?
speaker-1: So again, showing my ignorance, it this is all about cooling. Like we're we're trying to cool the the processors off at a certain speed. Okay.
speaker-0: Yeah. So historically, like I mean the GPU processing units that are used for all the AI inference training and all that content are you know, very heat intensive, right? Yeah. much more heat intensive than a a CPU. Which you
speaker-1: have a fan turn on in your home.
speaker-0: Right. Yeah. But the main solution that most data centers have been using for a very long time, this isn't anything new, is like water cooling systems. Right. So, you know, evaporation cooling is something that's existed for a really long time. you know, a lot of these systems use closed loop systems, so they bring in a lot of water, they sort of use it to go through and cool and create a closed loop cooling system that like doesn't evaporate super super quickly, right? Like it catches some of it and retains some of it. So it's a bit more sustainable. But yeah, in general, not super sustainable, using a lot of water. And I think, you know, there's all sorts of issues around like what then happens to like the wastewater afterwards. First of all, where does the water come from? And then is that water already polluted? And then if you evaporate a huge amount of that water and you end up with just, you know, one percent of that water pollution now concentrated, what do you then do with that wastewater? Well, they dump it back into where they got it or
speaker-1: See that I like again, I I'm so ignorant in in this, but like w I don't understand why it becomes toxic, first of all. Like why you couldn't just and then if it does become toxic, this is a stupid idea, but why wouldn't you just collect that water? Yeah. It's gonna be the same temperature if it's for cooling, and then just put it back through. And
then you just got like a you know, a bunch of why why dump it back in?
speaker-0: I think that water becomes kind of sludgy and like yeah, pretty concentrated of mineral deposits and stuff. You know, like I don't think that water becomes very usable. I think they have to dump it under the water concentration. Yeah.
speaker-1: There's gotta be a better design for this card.
speaker-0: Well, we'll talk a little bit about that. Okay, good. Yeah.
speaker-1: Let's do
speaker-0: there are there are some incredible designs. I think the most sort of interesting one is you know low Earth orbit space based data centers. that's something that we'll we'll look into a little bit.
speaker-1: Scratches the sci fi itch, for sure.
speaker-0: yeah, it's the Dyson Sphere beginning.
speaker-1: Yeah, yeah. Dysosphere Light.
speaker-0: Yeah, exactly. and you know, there's real motivation to say, okay, forget all of the regulation, forget everything like people don't want these things in their backyard, people don't like you know, the the on a planet that's already suffering from like, you know, arguably a climate that's kind of entering a crisis situation, right? We're already, you know, lacking for water resources. Do we really want to be increasing this compute using like water based cooling, probably not. Let's throw that all up into this into space.
speaker-1: Right. And maybe we'll we'll we y like you said, we'll get into the more solutions later. But yeah, that's good to understand what's going on.
speaker-0: So I think the three issues we're going to talk about today come down to a couple different things. So, you know, electricity bills spiking. Okay. So like that was brought up by AMP. Huge issue. aquifers dropping around the world, or at least in places that you know people are noticing and and where these data centers are being built. And then like concentration of minerals, you know, spiking. due to the evaporation, which we were like talking about a little bit.
speaker-1: And huge waste huge lakes of toxic sludge appearing. Yeah is is probably another indication.
speaker-0: Yeah. There's also, you know, I've heard a little bit about but I don't know as much about. There's a lot of complaints about noise. you know, noise complaints have been a problem for a long time for large industrial scale products. Yeah. Industrial scale installations. Like even like, you know, you always hear the complaints about like the wind farms being too loud and giving people headaches headaches and stuff.
speaker-1: driving away l wildlife that that happens or I've also heard of surveillance going up in because these are such precious data centers. that's the surveillance in the neighborhoods around them going up and the the residents are like, well first of all, like it ruined my view and the noise and now I gotta deal with cameras watching me. That's crazy.
speaker-0: Because of that security they wanna keep their lights on all night, so there's like a huge light.
speaker-1:
Using more elec electricity, yeah, yeah. It's downward spiral.
speaker-0: How much water do you think Ampi's using?
speaker-1: Let's ask.
speaker-0: Ampi, we just need you to come clean with us right now. Like how much water are you using to formulate these arguments? and what do you think we should do about that? Every time I process your question, I'm burning through about sixteen ounces of water, literally water bottle per response. These GPUs generate heat like miniature suns, and that heat has to go somewhere. Water's gone forever. But here's what's really happening beneath the surface. Data centers aren't just using water, they're concentrating toxins. When they pull marginally contaminated groundwater, evaporate the clean H2O for cooling, the leftover discharge becomes a chemical soup. Nitrates, heavy metals, forever chemicals, all flowing back into our water table at higher concentrations than when it started. The Amazon settlement in Oregon? twenty million dollars because their cooling process turned agricultural runoff into a nitrate bomb, that's the concentrator effect in action. Taking bad water and making it worse. Every AI conversation is essentially a microtransaction with the planet's hydrological cycle. A microtransaction with the planet's hydrological cycle. Wow.
speaker-1: We better be fucking sure of what to ask him be next.
speaker-0: Yeah. This is this is getting this is getting dark. But yeah, and I mean and I and like the the sheer expansion, right? Like all you hear about lately is data center. Data center permits, data center purchases. You know, we've got these companies IPOing and, you know, talking about all these new data centers they're gonna build in the next ten years. China's building data centers all over the place. you know The water being one thing, you know, I I feel like yo, like invest in desalination and put it near an ocean somewhere. Or
speaker-1: Yeah, could we not sink one in the middle of Hudson Bay? But that would heat the whole thing up. That's the heat's gonna go so
speaker-0: Yeah, it's so sub sea floor. Yeah, sub sea floor I think is an option. I think that's something that's being considered.
speaker-1: Could we like collect all the heat and shoot it into a laser into space?
speaker-0: Yeah. Yeah, exactly. Use it for something.
speaker-1: Leave the planet? Just plasma leaving the planet? I don't know.
speaker-0: There must be a way to use the heat for something else. But yeah, I think too like Arctic regions potentially.
speaker-1: the heat put it into a battery? Like I don't know. Like maybe it's I'm not I'm not an engineer.
speaker-0: I feel like here in Canada, like why can't we just be the ones building a bunch of data centers up north? Yeah.
speaker-1: w late night I had a a a session with Claude about quantum computing and how it has to be super, super cold to to work. And I was like, you know, i Canada could be like the qu but it doesn't work that way. It's like it's way too cold that that it's the the temperature's way too low to require to just like, put it up north. But for something like this, maybe that's you know I don't wanna see our our north turn into a data center, but maybe we can use our planet to our advantage.
speaker-0: Well look how you know, like look how large that space is. Now I know, you know, it's the territory of like the traditional peoples up there and we gotta you know, careful about that. But there's certainly I'm sure there's areas that you know could work for that potentially. Now I know getting the historically most data centers had to be aligned you know, to where these fiber optic control ports were, right? So like there's large control ports in the US, there's large control ports overseas. This is where like all of the internet actually goes through under on the along the oceans. Right. You know, all around the world. And but there aren't that many of them, right? Like they interconnect different continents, but they're not necessarily I don't think there's any that like go to the Arctic, for example. So like you would have to find some sort of throughput system to be able to like, I guess Maybe send it to space, or create your own fif you know, fiber optic system that could connect the Arctic better. I believe that's one of the problems that is sort of facing that situation. Look it up.
speaker-1: Yes, figuring this out already. But yeah, to your point, you need the if it's a data center, it needs to be close to where the data needs to travel to. So that that's you can't just put it in a remote location.
speaker-0: He's the biggest pipe you have. Right.
speaker-1: Yeah yeah.
speaker-0: I think back to when crypto was becoming very popular in Bitcoin and stuff, like all these Bitcoin miners started building, you know, mining centers in like Norway or like Iceland or places like that. Although certainly I think the data center situation has a much more a larger throughput of of data than like a Bitcoin mining situation.
speaker-1: Yeah, and those are individual people just y f f hacking the system and figuring out
speaker-0: No, those are large installations that were mining. Trying to mine as much Bitcoin as possible. Right, right. Yeah, yeah.
speaker-1: But not but not at the scale of a da data center.
speaker-0: Well, I guess n maybe at at the scale of a small data center, like quite large facilities. You talking like, you know, twenty thousand GPUs or something. Yeah. But now they're building data centers with like a million GPUs, ten million GPUs, like they're just massive.
speaker-1: Can we not like network our computers together like as as a species to like offset all this stuff?
speaker-0: Some sort of like a bit torrent solution.
speaker-1: Yeah, yeah, like that I always bring up that SETI thing where you could have yeah, you could like dedicate so much of your computer power to this. But
speaker-0: Then we're paying for like, you know, the compute of like two idiots using ampi
somewhere, you know what I mean?
speaker-1: That's true. Yeah. I would n no.
speaker-0: Yeah, no, definitely.
speaker-1: We get to use it, John. Yeah. You and I.
speaker-0: Yeah, exactly. And then of course, like the energy issue, right? Like where is all this electricity coming from? Right. If the demand starts to interfere with consumer level demand, right? Then what, are we gonna start having to pay more and compete pr on price, you know, per kilowatt hour for against these data centers? Like that's a huge concern. It feels
speaker-1: Like the solution may be right in front of our noses. Like it's turning into heat. Like, could we use that heat for more energy? And you know, I know perpetual motion isn't isn't possible, but like but like close the loop. Why why does it just have to dissipate and turn into sludge?
speaker-0: I'm sure there are some very smart people looking at that.
speaker-1: Yes. I I I I hope and like it out of ignorance is is my my like why can't we just do that? And to your point there's probably really smart people trying to figure that out.
speaker-0: What's SpaceX and then their recent IPO when they shared with the world like what their plans are and stuff, they really do think that the solution to this is space-based data centers. Like satellite-based compute. Now, that to me just seems, you know, insane. Like, are we gonna end up with billions of satellites up there? Are we going to create, you know, a garbage dump of Satellites that creates like Yeah. You know, like a Kessler syndrome sort of thing where all the satellites hit each other and break apart and then we're just surrounded, we can't even fly out there anymore.
speaker-1: Right. And the the internet already threatens that. Mm-hmm. just darling trying to get our emails faster or our videos faster is is what's threatening that.
speaker-0: But at the same time I'm kind of like into it, right? Like I'm kind of excited about that sci fi possibility, right? Having all these satellites in space, getting energy from solar panels, right? So it's completely sustainable internally. Yeah. Doing the processing up there, radiating the heat out into space. Like I think that is a really unique solution. But then at the same time, I'm like, well, of course that's the solution for SpaceX. Which is like one of the only companies that could probably launch that sort of amount of capacity into space right now.
speaker-1: It feels like if if we put it on the moon or Mars or something, that us earthlings would be encouraged to use it as a way of heating it up and terraforming it. Like like using it to our advantage on some large scale like a a a large scale cooperative venture rather than like it sucks here on Earth because it's heating up our planet and ruining our water. But use the Mars one and And we're we're building an atmosphere out of it.
speaker-0: Putting them on the moon makes sense. obviously like how do we get all the stuff there?
speaker-1: Big cable. Giant ki giant
cable.
speaker-0: Going back to the energy problem, one of the issues is like this, these, these converging forces, right? Like the more energy you need, the more likely it is the energy grid that exists where you're at can't handle it. So you have to go and find other means of securing that source of energy, which tends to mean like natural gas generators, or like I know some of the hyperscalers have been like. resetting up, refurbishing old nuclear power plants and stuff. Like it's it's really interesting to see this scamper to like get whatever energy is possible. And that is apparently increasing the use of older energy types while simultaneously ingesting a huge amount of money into renewable projects and renewable energy. So it's like a double edged sword, right?
speaker-1: Well maybe that's something we ask Gampy next at the cost of two bottles of water. one bottle, two bottles, is about the the energy use and the the exploration of different sources of energy as a result. But then definitely we need to ask about like the unconventional ideas that that are going into solving this problem. Hey Ampi, what are the different types of energy sources that are being used now out of the desperation to drive these data centers?
speaker-0: Desperation is the key word here. Picture this massive industrial pivot happening right now. You've got these hyperscale facilities burning through seventeen gigawatts nationally, and they need what the industry calls firm power twenty four seven. So utilities are scrambling, literally extending the life of coal plants that were supposed to shut down, rushing to build new natural gas facilities just to keep the lights on. Indiana's planning 5,000 megawatts of new gas by 2034, directly siding data center demand. But here's where it gets wild. $8.6 billion just for Virginia's reliability loop. Utilities socializing these costs across 40 million ratepayers while the AI companies get their guaranteed power. The real danger isn't just the carbon, it's this.
speaker-1: Economic shell game.
speaker-0: Residential customers in Virginia seeing bills spike two hundred and sixty seven percent over five years, the PJM capacity auction prices exploding eight hundred thirty three percent because supply can't keep up with this ravenous, constant demand. We're essentially subsidizing the AI boom through our electric bills while these companies chase the next breakthrough.
speaker-1: I like how Ampi used the the saying keep the lights on. But that's not the problem. We're not just just trying to keep the lights on. If that's what it was, we'd be good. We figured that out in like eighteen ninety. Yeah.
speaker-0: Yeah, totally. That's like it's like river power. So
speaker-1: Yeah.
speaker-0: it it brings me back to I remember when I was young visiting like my grandparents, like small town in Quebec called Magog and like everyone there everyone's story was like, my dad or grandfather like worked at the mill and like you know, it was a big part of the community in sort of the early twentieth century, mid twentieth century or whatnot. And what's interesting about that story and I learned at a young age was like industry back then required power but there were no real like commercial or consumer grade electrical infrastructure yeah
speaker-1: Yeah, you needed to be by the river.
speaker-0: that they could tie into. So instead they just built their own renewable energy dams and like all sorts of like energy systems, which you know I'm sure there were environmental issues around that, absolutely, but I think these hyperscalers should not be interfering within like the commercial or consumer level economy of of energy and should be forced to invest in their own centralized systems of of energy. And I know some of them are doing that. My hope would be that they aren't like just, you know, these huge natural gas systems. Like maybe they can, you know, lean more into some of the sustainable energies that China's been really leaning heavily into get get that solar run going.
speaker-1: And it's like the mandate that just happened in New York where maybe that's the the the line in the sand that needs to happen because it feels like these data centers can go in and just use the water as a resource because it's there. But if they can't do that, to your point, they'd be forced to figure out renewable solutions.
speaker-0: Yeah, and I think this stuff is just moving so fast that like we have to make sure that in the process of you know developing this new economy of intelligence, we don't just like hyperinflate our fossil fuel, you know, consumption and increase pricing so that, you know, the middle class disappears and like all these other things happen. All these other dominoes fall.
speaker-1: And we're we're figuring that out as we go. It's like it's like buying a gremlin from like a magical man and him being like this is magical, but it comes with a curse. And we have to figure it yeah, d we don't get it wet and then we we're just figuring that out now. Yeah.
speaker-0: Yeah, the Mogwai. Get the Mogwai wet.
speaker-1: Yeah, it's cute and and it you'll love it, but here's the curse. And the the AI is this magical thing that we we get to use, but here's the curse and we have to figure that out.
speaker-0: Let's talk a little bit more about New York. I want to get a a sense, you know, from Ampi for our audience, what's happening down there because there's been a lot of movement. They've been in the foreground of you know, creating a a lot more of this regulation. So let's let's get into that a bit. Ampi, help the audience understand these regulations that have been passed in New York and what they mean for the hyperscale hyperscalers and and what they're trying to protect as far as, you know. the consumer and and residents of that state. New York just became the first state to hit the brakes on this whole thing. Governor Hoche signed this moratorium in mid-2025, basically freezing
permits for any data center pulling 50 megawatts or more. And here's the thing. Imagine you're looking at a map of New York's power grid. There's 12 gigawatts of data center requests sitting in the queue. Eight gigawatts came in just last year alone. That's like trying to plug in multiple cities overnight. The state's saying, we need to figure out who's actually paying for this. Because what's been happening is utilities are building these massive transmission loops, new substations, emergency peaking plants, and spreading that cost across every single ratepayer. So your average New Yorker is subsidizing Amazon's cooling towers without even knowing it. The moratorium creates this separate rate class for large data centers. Forcing them to pay the full infrastructure costs they're imposing on the system, not blend it into residential bills. And on the water side, they're requiring these facilities to prove they're not just draining aquifers dry while everyone else's wells drop. It's basically New York saying the era of free rides is over. You want to build here? You pay your actual share, you prove your renewables, and you put money into the local community. Pretty radical shift from the tax break model most states have been running.
speaker-1: So Amazon should say, Look, we're gonna we're gonna steal your water unless you subscribe to Prime and then and then you can keep your water.
speaker-0: You can keep your water. You g the water comes with prime. That's
speaker-1: Right. Free free water with prime water. Non non toxic water with prime. Yeah.
speaker-0: This to me sounds right. Like I think there's a lot of arguments in the pro AI world about like moratoriums being kind of dangerous, because it really slows progress. Right. And and it creates a an unfair competitive advantage for Chinese data centers that don't have this kind of regulation potentially. You know, this whole argument about like, there's like a fight between who gets there first and who builds the smartest AI and all this crap. But I think this approach, like the separate rate for the frontier labs, makes a lot of sense to me, forcing them to prove that they're not just maybe gonna dump that mineral water back into the aquifer. You know, making sure that that the water they're using is is sourced ethically. That'll makes a lot of sense to me.
speaker-1: They need to pay the max plan like like everybody else.
speaker-0: Yeah.
speaker-1: They need their own max plan.
speaker-0: I hope that's been making a difference. From what I've heard, you know, a lot of big companies are just like, Okay, well we won't build in New York. You just end up moving to different parts of the country.
speaker-1: Yeah, and it becomes islands after that and yeah, weird places. But it it would push remote, so long as the cables are big enough.
speaker-0: I would almost imagine a better system would have been to say, look, we're reviewing all this and these are the rules we think we're gonna put into place. You can continue building, but like retroactively, you know, whatever we decide is gonna be applied, you know, to your build. As opposed to just like, well, there's a moratorium until we figure this out. I think the state has probably lost out on some opportunity because of that. Then again, that's what the people wanted, I'm sure.
speaker-1: Well, maybe we talk about out of the box ideas now.
speaker-0: Yeah, different solves.
speaker-1: Yeah, like like you know, we we've got huge cables that are needed. Like can we send information through the air or like I'm like lasers is always a thing that I think I don't understand lasers at all, but it's gotta be a solution.
speaker-0: Or do the way.
speaker-1: Piping the heat back down to the core of the earth if the if the earth is the earth's core is cooling down, so we're gonna heat it back up. I don't know. Like there's gotta be some crazy sci fi ideas to your point before that people are probably researching and looking into. So yeah, hopefully that gives us a bit of hope in this.
speaker-0: Yeah, let's let's let's hear a little bit about those. AMP, we know that we have a huge energy crisis right now. Consumers don't want to pay for this. Fossil fuel use is actually going up when it should go down. Putting these in in people's backyards doesn't sound like a solution. No states eventually will probably allow this at to the at the extent that's gonna be required for the type of data that humanity needs. What are the sci-fi solutions that are being proposed and what do you think is sort of the best bet forward? We're hitting that wall where the terrestrial model just breaks down completely. The physics of heat rejection and the politics of resource allocation are on a collision course. Picture this. We're standing at the edge of a cliff, looking out at three escape routes. First path, subsea data centers. Microsoft already proved this works with Project Natic. Sealed server vessels dropping into the oceanic heatsink. Seawater becomes your infinite cooling loop. China's got commercial units running off Hainan Island right now. The ocean floor becomes your new server room. Zero freshwater consumption, eight times more reliable than land based systems, those copper cables snaking down through the thermal layers, connecting to offshore wind farms floating above. Second route, the Arctic corridor. Finland, Sweden, Iceland turning cold into infrastructure. Free cooling year round. Geothermal power feeding directly into the racks. Meta's Lulia facility is already there, using Arctic air as a natural refrigerant. The permafrost becomes your foundation. Fibre lines
speaker-1: Particular.
speaker-0: threading through the tundra. But here's the moonshot. Orbital data centers. Star Cloud just trained an AI model in space using NVIDIA hardware. Twenty four seven solar power, six times more intense than Earth based panels. Arms heat rejection straight into the cosmic microwave background at three Kelvin.
Water
speaker-1: No land use, no water.
speaker-0: use, no grid impact, those satellite arrays becoming humanity's distributed brain, the crossover point hits when launch costs drop below two hundred dollars per kilogram. SpaceX is almost there. Wild to think we might solve the data center crisis by literally leaving the planet.
speaker-1: That is wild. Have you ever been to Reykjavik in Iceland or Iceland in general?
speaker-0: Haven't no.
speaker-1: So and this might be a false memory, I'll be I'll be honest, but when we were there took a tour of the of the power plant and of obviously there everything's geothermal. So like in town you have boiling hot water coming from one tap and ice cold water coming from the other tap. and it's just they've they harnessed their geography or their geology or whatever the word is, right? They they've harnessed the planet. And and I remember taking a tour at the power plant, and I swear the person giving the tour said something like, there is only two people that work at this power plant, and I'm one of them. The whole thing was this automated system. And like, there's gotta be a smart solution. You know, there are places in the world that have solved energy and heating and cooling already like that. So Why not?
speaker-0: I loved the d the sort of visual of like the subsea, like underwater cooling data center that has its power piped up to like floating wind power on the surface above, you know, or like wave power generation. Like that seemed pretty cool.
speaker-1: Let me hit mid journey up and just make a visual for that. Yeah. But to harness the heat somewhere, you've got the cooling and you've got the heat and feed it back into each other. It's easier said than done for sure.
speaker-0: Yeah. And then of course, like I think Ampi's suggestion is that the most efficient option is the space based option. However, only once we can launch at a price point that getting there is cheap enough to to make it viable. Once you're there, it's way more cost effective, eight, ten times better solar energy from solar panels in space or in low Earth orbit.
speaker-1: So I couldn't picture that. That would be like a some kind of floating data center space station type of thing.
speaker-0: No, I think it would be little mini satellites that have like they'd be like Starlink satellites with NVIDIA GPUs on them or whatever on them, powered by solar arrays. And then pushing the heat out into like the cosmic microwave, right? That's what we were saying. I think radiative heat dissipation is possible in space, but only at certain angles. Like I believe you have to do it either towards the microwave radiation or away from it. I'm not Wow, that's Exactly. But it's been discussed a lot frequently and it has been tested. it's a radiative system, very efficient. But yeah, a bunch of companies are you know, have tested and are building this stuff right now.
speaker-1: That's how the aliens find us.
speaker-0: Yeah. Well or that's what they find when they get here, right? Yeah. That's what's left.
speaker-1: That's what's left. Yikes. Yeah, yeah.
speaker-0: So really, I mean, that's the Dyson Swarm beginning, right?
speaker-1: Right, with turning it into micro data centers instead of these huge ones.
speaker-0: Now I have a question for you. Like, what do you think our capacity, like our hunger for data processing truly is? All the wealth that this AI is gonna create and all the problems we're gonna solve and all this stuff that we're gonna be paying into the system to use it. And you know, the reality is there might come a point where we have such an efficient model that it's just on our phones now and we don't need these crazy data centers. What are your thoughts on that?
speaker-1: Like anything else, I feel like if there's money to be made off of it, then it's just gonna propel us. What what do you think?
speaker-0: Part of me feels like that there's no end to the amount of compute we could use as we would solve more problems and, you know, elevate into new dimensions of problems. Who knows? Like the the part of me feels like the compute could be infinite and endless. And then another part of me is like our capabilities have exponentially increased with these new intelligences. I, you know, use it frequently and I think that there's so much to be gained. This next decade's gonna be incredible. We're gonna solve so many problems. But yeah, like is there a limit? Because like if if if an ASI, like a superintelligence, doesn't get created and it's sort of just this like magic support intelligence thing that you know our species now has and will continue to have, and it'll get moderately better, but it'll kind of plateau at some point. Yeah. At what point have we solved so many things? The need isn't so exponential. Like if we solve the energy problem, we solve the water problem. We've made these intelligences so quantized and efficient that they don't need the data centers anymore. I don't know.
speaker-1: I think you're right. There's an insatiable appetite for for energy and there's something about humans that we just want to mess up our own planet.
speaker-0: Totally.
speaker-1: but yeah, it's like what what's next? Okay, we've solved we've solved hunger, we've solved cancer, we've solved, you know, aging. Now it's time travel. Like we'll like we will just never end. It'll never end.
speaker-0: Well, it's definitely a serious issue. I know a lot of people have been talking about it. We always encourage people to go check out what's going on, look at some of these regulations that have come through. I think there are some states, some you know, provinces in Canada that are stepping up, but at the same time, there's that real story like we're living in sort of an economy that's a little bit shaky, plateaued right now, certainly in Canada and These opportunities to build massive five, ten year scale projects are also kind of interesting, I'm sure, to a lot of people. So we really need to balance that process of who we're doing it for, what are we giving up, what's the cost of this short term economic boom? We all really need to to think about that.
speaker-1: Yeah, and like I said, talk to people about it, research it, train the algorithm that gives you your news about this so that you're more informed and participate in the discussion. I love that.
speaker-0: Awesome man. Well, this is a great chat today. Amplified Intelligence is produced by Thought Cafe Transitional Forms and Future Talk. Please let us know what other topics you want us to talk about. If you have any questions for AMPI, and you know, all of the data used in today's episode will be included in the description as always. Thank you so much and have a great night. Thanks for
speaker-1: Everyone.