In their early days, data centers were largely welcomed by communities who saw their arrival as an economic boost. Today as concerns around noise levels, power prices, water use, and AI grow, they face fierce opposition.In this episode, Shayle speaks to Brian Janous, co-founder and chief commercial officer of Cloverleaf Infrastructure. Together, they unpack the data center backlash and explore how the industry can approach the problem differently.
Shayle and Brian dive into specifics like:
- Communities’ shift from “red carpets to pitchforks” on data center development
- How misleading information about data center energy and water usage has led to the AI sector losing the public relations battle
- Distinctions between data centers’ micro-level benefits and macro-level price inflation on power grids
- The potential, and pitfalls, of “bring-your-own-capacity” and behind-the-meter generation
- The viability of taking data centers off-grid, off-land, or off-planet
- The emerging market for distributed edge compute facilities
Resources:
- Catalyst: When to colocate data centers with generation
- Catalyst: Explaining the ‘Watt-Bit Spread’
- Open Circuit: The off-grid data center fantasy
- Open Circuit: Utilities are in the crosshairs of the data center backlash
- Latitude Media: In Indiana, an anatomy of data center opposition
Credits: Hosted by Shayle Kann. Produced and edited by Max Savage Levenson. Original music and engineering by Sean Marquand. Stephen Lacey is our executive editor.
Catalyst is brought to you by EnergyHub. Peak season puts every grid to the test — and the utilities that pass are the ones that built flexible capacity before they needed it. EnergyHub works with more than 170 utilities to coordinate 2.5 million devices and 3.4 gigawatts of dispatchable flexibility through a single platform designed to perform when it counts most. See what that looks like at https://www.energyhub.com/
Catalyst is brought to you by Bloom Energy. Bloom Energy fuel cells deliver affordable, ultra-reliable onsite power for hospitals, utilities, and data centers – at speed and at scale. Learn more by visiting BloomEnergy.com.
This episode of Catalyst is brought to you by ENGIE, the smarter energy supplier. ENGIE doesn’t just provide the power to run your business — they supply the energy to move it forward, with reliable, flexible solutions built for what’s next. Learn more at engieresources.com.
Transcript
Shayle Khan: I’m Shayle Khan. I lead the early stage venture strategy at Energy Impact Partners. Welcome to Catalyst.
So, you already know that the data center backlash is real. I think the question worth asking is whether it’s the kind of thing you can actually fix, at least in the near term, because I think there are two separate issues kind of wearing the same jacket. One is a set of specific, negotiable grievances, some of which are highly warranted: this substation upgrade that has a cost on electricity rates, that water draw, this rate case, this worry that residential customers end up subsidizing a load that showed up overnight, pollution effects, all sorts of things like that. Those, you know, you can solve by doing it right: site it differently, pay for the upgrade, structure tariffs so the big load carries its own cost. The tools exist, and the industry, I think, is starting to get better at using them.
But the other one maybe isn’t really negotiable, at least not at the individual level, because it’s not really about that data center. It’s about AI and the companies building AI. And that issue would be there even if the facility drew half the power, lowered rates, never touched the water table. You can’t redesign your way out of a values fight.
So, where does this go? When developers can’t get a yes, if we hit a ceiling on our ability to build new large-scale data centers on the grid, do they stop asking the community and start building their own power behind the meter? Do they go small? Do they go off-grid? Do they go off-planet?
Honestly, all of these are pretty clearly suboptimal, but right now, I think suboptimal is not a dirty word in this industry. And so I don’t think we can say how much of them will happen.
Does this get worse before it gets better, and what should data center developers do differently, if anything?
This is an interesting set of questions right now, and it is basically going to dictate the future of power in this country to a large extent. So, let’s talk through them. My guest is my good friend Brian Janous. He’s been on the show a bunch of times before. Brian is the co-founder and Chief Commercial Officer of Cloverleaf Infrastructure, which just a couple weeks ago announced a big partnership and investment from Nvidia. Congrats to Brian on that. Really good conversation. It’s coming up after the break.
Shayle Khan: Brian, welcome back.
Brian Janous: Thank you. Excited to be back.
Shayle Khan: All right. So, the data center backlash is here. It is real. It’s substantial. It is mainstream news literally every single day right now. You know, you’ve been involved in developing data centers for what, 13, 15, something—like, a long time, right? How many years?
Brian Janous: More than 15 years, probably. Yeah. I think I worked on my first data center project in 2006. So, maybe that’s 20 years.
Shayle Khan: Yeah. Okay. Wow. All right. So, 20 years. Paint me the really quick picture, in your perspective, of how we got here. Like, why did we land in this place where there is this absolutely enormous upwelling of opposition?
Brian Janous: Yeah. So, it’s interesting because, you know, we’ve been building data centers at some scale for 20 years in this country. Even a few years ago, you were getting the red carpet rolled out for you when you’d go meet with communities and cities. We’ve somehow in a relatively short time gone from, you 0know, red carpet to pitchforks. You know, and it was—and there wasn’t even like a whole break in between. You know, it was really fast.
And so, I think there’s a number of things going on. I mean, one, of course, like what we’re building today is much bigger, you know, in one fell swoop than what we built 15 years ago. Now, if you look at places like Quincy, Washington, for instance, Microsoft probably has close to a gigawatt of data centers in Quincy, Washington. I don’t know the exact number, but it’s got to be sort of in that ballpark. But it’s happened over a period—I think they built that first data center in 2007, right? So, almost 20 years, you know, they’ve gone to a gigawatt. And there was a fantastic story just in the last week about Quincy, Washington building a $15 million aquatic center, and they have a $150 million state-of-the-art school, and their unemployment rate has gone from like 29% to 6%. You know, so they’ve been able over a period of 20 years to experience substantial benefits from this.
But when you come in and say, “Well, I’m going to build, you know, what they did in 20 years in 18 months,” you know, clearly there’s a larger target because then you start going, “Well, how much water is that? How much power is that going to be?” And so, some of it is just the scale that we’re doing. It draws a lot of attention, understandably so.
Some of it is the—some extrapolations that people were doing around, “Well, if, you know, this old data center uses X amount of water, and I take that times 100, then it’s 100 times that.” But of course, it doesn’t take into account the fact that the technology’s changed substantially, and we’re not using the same types of cooling systems that we did back then. We’re using largely, you know, closed-loop systems because of the heat rejection that’s necessary. So, some of it is just a misperception of how much water these things actually use. And if you look at the actual facts, like, data centers do not use that much water relative to—fill in the blank—almonds, avocados, golf courses, you know, chemical plants, power plants.
Shayle Khan: Right. Yeah.
Brian Janous: Right. And I find that one even ironic, too, because if you were coming into a city and say, “Hey, we’re going to build a new factory, and we’re going to bring all these jobs,” everyone’s, you know, rolling out the red carpet again. It’s like, but that factory’s going to use more water, likely have more emissions, you know, have more externalities than a data center would. But, I mean, unfortunately, like, data centers are about the only thing we’re building in this country at any massive scale right now, so they get the lion’s share of the attention.
And I also think—so, you know, there’s the externality issue. I mean, there’s the scale, there’s the at least perceived externalities. And I also think just the general backlash against AI is also real. I think people look at it and go, the thing happening in that—you know, if I’m building an auto factory, it’s like, “Well, at least I like cars. So, like, sure, maybe I don’t want this factory in my town, but I like cars, so I’m going to accept the fact that we’ve got to build cars somewhere.” But if, you know, one, they have this perception of the externalities of the building itself are negative for the community, which I would contend is not true, but nevertheless, that’s perception. And two, the thing that building is doing, I also don’t like, well, then it’s a double whammy then. It’s like, “I don’t like the building, and I don’t like what they’re doing in the building, and therefore, like, why would I support this project?” So, I do think the tech companies have a real issue around how AI has been presented, you know, and that is causing a lot of this backlash as well.
Shayle Khan: Yeah. The other layer, I guess I’d be interested, because you’ve seen this change over time, but I think the other layer that’s worth throwing in there, too, is the secrecy component. Because it used to be totally standard to be very secretive as you were developing a data center in a community, and nobody knew whose data center it was until very late, and all this kind of stuff. And that falls maybe more into your latter category of generalized opposition to AI—this is more, I guess, generalized opposition to the tech companies, probably, to the hyperscalers. But what has changed, if anything, in terms of how secretive you need to be when you’re developing a data center, and at what stage?
Brian Janous: Yeah. That’s been an issue for sure. I mean, for Cloverleaf, we don’t sign NDAs with government officials. Like, we don’t think that’s good practice. I would say that those that have done that are probably moving away from that, I would guess. I don’t really know the policies of all the tech companies right now, but I think that issue is a lightning rod issue. And we’ve seen that in some of the regulations that have come out. I think that was part of Shapiro’s deal in Pennsylvania, which, you know, we’re very supportive of that. Like, no NDAs, that’s great. Disclose the cooling system. Like, all these things that we’re seeing, most of them I look at and go, “Yeah, I agree with everything they’re saying. Yes, that’s responsible data center development. We should do that.” Now, how it gets implemented is a whole ‘nother story, right? And can you actually implement it in a way that’s, you know, smooth and efficient, and actually enables good projects to move forward, or does it just grind everything to a halt? Like, that remains to be seen, how this all gets implemented.
So, I think that transparency issue is super important. You know, we still sometimes face issues, though, because, as you know, like, we’re an early-stage developer, right? So, we’re developing a project without necessarily knowing who the end users are going to be. So, we end up sometimes in these situations of saying, “Well, why don’t you tell us who the end users are going to be?” And it’s like, “Well, we don’t know.” And then they’re like, “Well, why aren’t you being transparent?” I’m telling you everything I know. So, you know, that’s something we have to navigate as an early-stage developer that, you know, you wouldn’t if you’re Google or Microsoft.
But then, you know, you still have to deal with the same AI backlash. You could say, “Well, you know, the company’s Meta,” and they’re like, “Well, I don’t like Meta.” Well, then it’s like you didn’t really fix anything. So, that’s the general—takes us back to that AI thing. So, I’m not sure how much that really fixes anything.
And we, I mean, we’ve seen this with dialogues with communities, too. It’s like, you can be as transparent as you want. At the end of the day, like, often times the response is, “Well, I just don’t believe you. I just don’t like it.” Or, “I just don’t like it. I don’t like data centers.” So, it’s like, okay, well, I can’t fix the emotions around this. And that’s a lot of what this is, is I think the industry has tried to attack this with facts and really missed the emotional resonance of like—people just aren’t happy about this stuff right now.
Shayle Khan: Yeah. I mean, that gets to sort of—I think the bigger picture question here is, in your first category, there’s a list of things that are concerns that communities have. Some of them are totally well-founded, and some of them are less well-founded. But they’re specific. They are impacts on power bills, they are water consumption, maybe they are secrecy, they are economic impacts on the community. Those things seem very tractable to me, right? And there’s a playbook now that the good developers of data centers—whether it be the pure developers like you guys, or some of the hyperscalers; I think Google maybe is the best at this—have a playbook where you address all of those things as close to upfront as you can, and you offer a lot in exchange to the community. I mean, you gave the example of a town in Washington, but, you know, there are lots of these examples now where it’s like, okay, data center’s coming, and as part of data center arriving, they’re offering, you know, a big community development fund and helping fund, you know, trade school X, Y, and Z, right? That stuff, that seems tractable, and like the good developers seem like they can solve for all of that.
But then there’s the other thing, which is this generalized opposition, whether to AI overall or to the tech companies. And that one at the micro level, Cloverleaf, can’t solve that second one on behalf of the AI industry. So, I wonder how you think about weighing those two against each other. Like, to what extent are we hitting a point where you can do all of the right things, and then the facts will totally be on your side, and yet it’s not going to matter?
Brian Janous: I think that’s very real. Like, I do think that because we see that day in and day out. Like, we come in, you know, with the facts, we lay it out, we put it into statute, and it doesn’t seem to move the needle in a lot of instances. Like, everyone’s already dug in on the issue that, you know, they really don’t like data centers.
And, well, you and I were at an event yesterday where we were talking about this, you know, with a group of industry folks. And, um, you know, that did come up of like, there hasn’t been enough from, whether it’s from the foundation models, you know, any player in the space, that has really sold, you know, the world on like why AI. Like, why is this actually going to make life better for everyone, as opposed to “this thing is a threat that we need to be scared of”?
And, you know, I am a firm believer in that like we will—AI will be transformative for the world, it’ll be transformative for medicine, it’ll be transformative for material sciences. I mean, there’s so many things we’re going to be able to do that we would not be able to do with AI. I’m also an optimist in the sense of I don’t believe the job displacement narrative because it just flies in the face of like every technology that’s—that we’ve ever had since the wheel. Like, anytime we create efficiency in some process, it always creates more prosperity. Like, that just has always happened. And I can at least intellectually assent to that it’s possible in theory there is some technology to come along that would be the first to sort of, you know, prove that to be false. But I’m just not convinced that that’s where we are with AI. I truly believe it will make us more productive, it’ll create more opportunity.
It will certainly result in job displacement on the margin, and I think that’s real, and that’s been true of every technology. I mean, it was true of oil displacing whalers, and you know, I mean, that’s always true. And I think that’s where we have to have good regulations, and we have to have good things, focusing on skilling and job retraining and all of these things that will ensure that, you know, those who are affected actually come out better on the other side.
So, that fear is very real, and I just don’t think the industry has done a good job of sort of educating people on why—why would we want to build these things? Like, what is this technology going to do? It’s more than just chatbots and, you know, slop memes, right? I mean, so that—that has been a big miss so far, I think, for the industry, and we’ve got to somehow right that ship.
Shayle Khan: Yeah. I mean, every time I have this conversation with anybody in the industry, we land at this place, which is the AI industry has not done a good—and I think everybody says this, right? Now, like, Sam says this, everybody says this, right? Everybody agrees on this point. The AI industry has done an insufficient job, maybe a bad job, of selling itself to the general public as something that is going to be good, and everybody points fingers at each other, at the, you know, leaders of the labs and so on, as to why their messaging was the wrong messaging or whatever.
But if you abstract all that away, I mean, the question that I have is, at least for the near term, is the toothpaste out of the tube? Like, is it too late to salvage general sentiment toward AI? You know, could you solve it with like, “We cure a form of cancer with AI” or something? Like, do you need something like that, and would it work even if you did?
Brian Janous: Yeah. You almost need—and it’s sort of the inverse of what I’ve been describing is like the Fukushima event. What I found baffling about the shift to this negative sentiment is how fast it happened without like an obvious catalyst.
Shayle Khan: Right. There was no meltdown. Yeah.
Brian Janous: Right. Yeah. Like when Fukushima happened, it’s like, “Oh, I don’t want nuclear.” It’s like, okay, that’s understandable.
Shayle Khan: Yeah, and in fact, the only jobs that have been lost really because of AI to date are like software engineers at software companies in Silicon Valley, which none of the general public cares about.
Brian Janous: Yeah. So, it’s curious how fast it shifted, you know, the public sentiment has changed without an obvious catalyst. All these other things we’re talking about are true about people concerned about water and all these things, but all those have always been true. So, it just—really, I mean, if you look at—I think Politico’s done a good job on, you know, the public sentiment, and it really was like starting last September. Really, perception of data centers was pretty steady, you know, and on both political spectrums was a little more favorable on the Republican side, a little less on Democrats, but it was like steady. But it started in September, and they’re doing these surveys every month, and it’s just like we’re just falling off this cliff.
So, yeah, I think it’s an interesting question. Could there be a catalyst the other way, right? Could there be something where it’s like, “Oh, wow, like I never thought about the fact that AI could do that.” Would that be the thing that got people to start to go, “Oh, maybe this is a technology that we need”? And hopefully it would be a positive catalyst, hopefully it’s not some negative thing like, “China now has this technology that we don’t have, and it’s threatening to us.” Hopefully it’s something more of, you know, optimistic than that.
But it almost seems like that’s what needs to happen, because I think, you know, the other issue we have right now is we’re obviously in the middle of the midterms and then coming into a presidential election where this is a really top-of-mind issue, and you’re seeing it on both sides of the aisle of either Republican trying to label the Democrat as the pro-data center candidate or vice versa, and it’s working, right? It’s actually affecting perception, and so then that means that that’s a playbook that others are going to try to use. And so we have that also, that sort of headwind of, you know, the political environment right now that is concerning. And so I would hope that we would see some sort of favorable catalyst come along that could, you know, shift the tide a little bit against, you know, those factors.
Shayle Khan: Right. But who knows, and who knows if it would work.
Let’s talk for a minute about the power price part, just because that in particular is both your and my specialty. I don’t know if you saw this, Andy, my colleague, who I think you know, wrote a piece a little while back that I thought made like an interesting point about power prices as they pertain to data centers and community opposition or community feeling about it, which is that there may be this dynamic wherein the objective truth, particularly today, is that a data center coming into a community tends to lower rates because nowadays, you know, the deal that the utility is striking with the data center is basically guaranteed to do that. That’s sort of the whole point. You were like a pioneer of doing this in the first place, right?
And yet, and so like it may be true that at the micro level, rates are lower if you get a data center near you or a bunch of data centers near you. And yet, it may also be true that at the macro level, the buildout of data centers has created a lot of inflationary pressure on electricity prices because it has created shortages of transformers and gas turbines and higher demand for natural gas and all this different stuff, which is pushing the price of all the components of your retail electricity price up. And so like, you know, that shouldn’t—you know, the logical conclusion of that should be sure, if youif you if you care about power prices, then you should want more data centers to be near you because they will lower your rates. But it also isn’t like entirely crazy for you to say data centers are broadly causing an increase in power prices.
I wonder how you think about that dichotomy.
Brian Janous: I think that’s right. And so, because there’s two ways that I can think about this: One, it is true that producing more of something does not necessarily make it more expensive, right? In fact, it often makes it less expensive, and we saw that for 100 years in electricity, right? Electricity prices did not go up for 100 years as we massively expanded the grid. Now, it is also true that we went through a period of no load growth, and hence, you know, we did not—not to mention, you know, COVID-related supply chain disruptions, and so growing, even if we’re growing at the same rate that say we were growing in the ’80s, we’re doing it in a very different environment. We’re doing it in an environment where we have to rebuild supply chains that aren’t there, right? And no one was prepared for the demand to spike like it has, and so, you know, the supply chain issues are real and thus the inflationary issues are real.
So, I don’t think though then the argument should be, “Well, we should just not grow.” It’s like, okay, we have some growing pains that we’re going to have to get through, but long term, that growth is still actually going to be good for rates. Like, it should be favorable. It’s just it’s very—and this is why I think we have to as an industry be very sensitive to the fact that, you know, there are short-term pains that we’re going through, and these concerns that people have about power rates going up and their power bills higher than it was last year is true. Now, is it necessarily higher because of data centers? I mean, maybe partially. I mean, it’s just higher also because we have these inflationary pressures that we have to deal with.
But I think that that is also an opportunity to say, “Well, what can we do as an industry to minimize the degree to which communities are facing those?” And we’ve seen things like that, like your data centers offering, you know, credits on power bills, and like are there things that we can do because I do think there’s enough money to go around in this industry to ensure that, you know, that people are really insulated to a large degree from that sort of macro impact. So, that’s not easy to do, but it is—it is possible, and so it’s something we have to take seriously as an industry of, you know, how are we really targeting—and you mentioned sort of the community benefit programs—like how are we targeting those programs that really address the concerns of a particular community, that aren’t just sort of general givebacks to—
Shayle Khan: Yeah, it doesn’t feel like a—like a, yeah, like a bribe or something.
Brian Janous: Right, right. But it’s like, “Okay, we hear you, we understand this pain point, how can we work together to target that particular pain point?”
Shayle Khan: Yeah. Staying on the power theme for a second. Okay, so one of the emergent results of the backlash and of the political response to the backlash has been states and others starting to really push hard on this bring-your-own-generation, bring-your-own-capacity concept. You see this with what Governor Abbott’s doing in Texas, Governor Shapiro in Pennsylvania, other places as well. And as I understand it, most of those, they’re implied behind-the-meter. It’s not like always entirely clear whether they mean that or not, you know? So, I’m curious your perspective on it. But I know you historically have been generally skeptical of behind-the-meter generation, and so I wonder what you think about that trend that we’re seeing right now.
Brian Janous: Yeah, and I do think the language that people use is often very imprecise, and I always wonder that, too. Like, when you say that, do you—
Shayle Khan: Do you mean that that capacity has to be on site or not? It’s not clear. Yeah.
Brian Janous: Yeah, because the bring-your-own-capacity concept as we think about it is, you just have to make sure that you’re funding all necessary capacity additions. And so, we’re doing a project right now in Oklahoma, and it’s with a smaller municipal, and we are doing like 100% of the sourcing of all the generation capacity, of the wind, solar, storage, and everything that’s going into that, and we are packaging that up and delivering it to the utility to then turn around and deliver it to the data center, right? So, when I hear bring-your-own-capacity, bring-your-own-generation, that’s the world that I think about. It’s, “Okay, go solve it for the utility, be financially responsible for backing it.”
But yeah, I think most people probably think about, “Oh, that just means you’re going to have your own generation behind the meter.” And I’ve heard a number of people talk about that, and I think they even—one, it definitely doesn’t solve the problem we’re talking about. Because if you’re talking about the issue is, you know, if the inflationary pressure on rates is not coming from, you know, more load on the system, but rather it’s coming from higher cost for equipment, well, just buying that equipment and putting it behind your meter doesn’t do anything but provide more upward pressure for the other stuff that you’re going to have to put on the system, right? So, I don’t think that—not to mention the fact that, you know, sort of unit-for-unit, doing something behind the meter actually results in you having to buy more stuff than if you were connected to the grid.
Shayle Khan: Right, you have to overbuild, you got to overprovision, whatever. Yeah.
Brian Janous: You’re overbuilding. So, now you’re making it worse! In fact, I was talking to someone from, gee, Vernova yesterday, and he was talking about this gigawatt data center, they’re having to build 2.6 gigawatts of capacity across generation and batteries and everything else. And so, when you think about that, you’re like, “You’re buying lots of extra stuff that you wouldn’t have to buy if you connected to the grid.” So, I don’t think that, if that’s what people mean when they say bring-your-own-capacity, I don’t think that actually helps the problem. From a public perception it is, but when you actually understand the ins and outs of what’s going on, you’re like, “Oh, that’s actually worse.”
I also wonder, like, if people don’t want a data center next door, do they really want a data center plus a power plant next door? That also seems like not the right solution. Like, I think that’s worse than data centers connected to the grid. So, I do get concerned when I hear about these proposals because I think it is very—well, again, I don’t know exactly what they mean, but my assumption is they mean something that actually is worse than, you know, the right way to solve this problem. So, it does create a lot of uncertainty in, what does it mean for how people are now going to go pursue projects as a result?
Shayle Khan: Yeah. You know, it’s interesting, now that you’re saying it, I think putting a bunch of—okay, in order to go behind the meter entirely, and particularly if you’re trying to go off-grid as a bridge power thing or whatever, you’re definitely overprovisioning if you care about reliability at all. So, you build too much stuff. Now, from the, you know, back to the micro-macro thing about power prices, it probably is fine from the micro perspective, right? You are not causing any new generation or equipment to get built on the grid, and you’re paying for all the stuff behind the meter yourself, so it’s probably fine micro. But it’s definitely worse macro.
Brian Janous: It’s fine micro, but I would even say, I mean, if you believe that a larger denominator is actually good for lowering electricity rates over time, it’s actually negative in that you’re taking that denominator off the grid, right?
Shayle Khan: You’re not adding to the—you’re not, sorry, you’re not reducing the slack, the excess in the system, which is the way to lower rates for everybody. Yeah, that’s a good point.
Brian Janous: It’s not negative, but it’s not positive either. It doesn’t help.
And so, yeah, I just don’t think that that is at all fixing the problem. From a public perception it is, but when you actually understand the ins and outs of what’s going on, you’re like, “Oh, that’s actually worse.”
Yeah, and on your second point, actually, I happen to—you know, my algorithms now, like, all they serve me is these like videos of local community hearings about data—I mean, this is your life, but it shouldn’t be mine! And yet, I watched a video yesterday of some Texas community hearing, some town hearing, where somebody was trying to fight against the data center that was coming, and they were mostly focused on the gas. They were saying things about how polluting the gas stacks are, and all the nitrous oxides coming out of the gas stacks, as they called it. And I mean, I think it’s a good point, right? Like, behind-the-meter generation, particularly if it’s thermal, is—there are reasons people don’t want that in their backyard either.
Brian Janous: Right. So, I don’t think this is at all helpful at this point in the conversation.
Shayle Khan: Yeah. Okay. So, then the next question, of course, is let’s say, and I actually think this is like a—I would bet on this being the case for the next couple of years, that it gets worse before it gets better. Partially because of the political timing that you described. It’s just currently, where we stand today, there is no political advantage to being the pro-data center candidate at the moment. I actually am interested to see somebody really try to take that mantle, because I think there should be a lane there for somebody. But I don’t know, unless you tell me otherwise, I don’t see it right now. What I see right now is exactly what you said: everybody’s fighting to, you know, brand the other candidate the pro-data center candidate.
Brian Janous: Yes. That’s what we’re seeing, too, but I’m hopeful, and you know, we’re certainly having discussions with various folks that we could create a narrative that really ties to that community benefit, ties to the jobs. And I have definitely seen more stories—I mentioned Quincy, Washington, there was a story about Mecklenburg County, Virginia, there was, I think, the Richmond Times-Dispatch did—Loudoun County, I saw one. Loudoun County is now the richest county in the country.
Brian Janous: Yes! Exactly. And the highest concentration of data centers, and their property taxes have gone down every year for the last decade. There was a good story about I think Ellendale, North Dakota, which is where Applied Digital—I think it’s North Dakota—where they have a facility. So, I’ve definitely seen a lot more stories that have been sort of like, “Whoa, hold on, slow down on this like data centers are absolutely terrible, period,” to like, “Oh, yeah, there’s a lot of benefits that come from these things, and there’s a lot of value that we can see for having the jobs and having the tax revenue and lowering our property taxes.” And so, that is a narrative that—it’s certainly not as loud as the anti-data center narrative, but it’s there.
Now, would a politician or a politician that has, you know, significant influence really get behind that story? It remains to be seen, because certainly the voices that have been pro-data centers have gotten a lot quieter leading up to the election. But, you know, we continue to work with the local communities and, you know, we work at all levels of government all the way down to, you know, the county, and we spend probably more of our time there. You know, we’re not doing as much at the national level, but not that that’s not important to us, it’s just, you know, we just tend to do a lot more kitchen-table politicking and spending time with communities and trying to hear what the concerns are.
But I would like to see that happen. I mean, I would like to see some candidate come and say, “Hey, it is important that we build things in this country. Like, we—and data centers are the thing that we are building at massive scale, and they are a huge part of GDP growth, and if we took that away, it would be hugely negative for the country right now.”
Shayle Khan: And also, the geopolitical stuff, like, you know, China’s going to build it either way, like, we probably should. You know, you hear a little bit of that, but it’s not—nobody’s nobody’s really owning it.
But okay, but let me just say, under the assumption that it does get worse before it gets better for a second, there’s two different thrusts of like, “Okay, so what happens then?” that I’ve seen people are thinking. One thrust is the—I’ve been calling it data centers go off-grid, off-land, or off-planet. Right? So, off-grid, on the sea, or in space. That’s one—I’m grouping them together. They’re obviously different from each other, but it’s like, we’re going to continue to build big data centers, they’re just going to be, you know, not connected to the grid or possibly not on Earth. And then the other one is tied more into the, you know, workloads over time shifting from just training to equal parts or maybe even more inference, and that meaning maybe we can go distributed, and it becomes edge computing, and we’re not trying to go build gigawatt data centers in communities anyway.
So, I want to get your perspective on both of those kind of in order. First, off-grid, off-land, off-planet.
Brian Janous: So, we already talked about off-grid. I don’t think that’s really the right solution. The underwater—it’s funny, so we did a—we did this in gosh, it was probably 2016 at Microsoft. One of the guys that was working for me, Sean James, who’s now at Nvidia, did this—it was mostly just like a science experiment. They literally built a Lego of a data center and put it in a fish tank for a think week paper, and was like, “Oh, you can put a data center underwater, and then you use cooling from the, you know…” And Microsoft Research picked it up, and they were like, “Yeah, we should do that.” And we’re like, “Uh, okay. I guess you can go do that if you want.” So, they did. They built one off the coast of California. I can’t remember, maybe it was in the kilowatts scale. And then they did it, and I’m like, “Okay, good, you did it. You proved that you can put a data center underwater.” And then they’re like, “We want to do it again, but we want to do it bigger.” I was like, “Oh, okay.” So, they built a much bigger one. It was a megawatt-scale, you know, off the coast of—it was in the North Sea.
And the theory was like, “Well, you know, you could drop these like all over the world. I mean, most people live close to water, major ports, and, you know, for latency standpoint, you could just drop these all over the place.” Of course, it never went anywhere, where like, well, you could do that, but like you can also build them on land. It’s a lot easier, and then your techs don’t have to be scuba-trained, and there’s a whole bunch of things like why you wouldn’t want to do that.
Now, I know there are some companies that are pursuing that today for real. The—what’s the name of the company?
Shayle Khan: Panthalassa is the one you’re thinking of.
Brian Janous: Panthalassa. Thank you. Yeah. So, I know they’re doing it, and I will say I was initially quite skeptical given my experience with underwater data centers. But given everything we’ve just talked about over the last 30 minutes, like, maybe! Like, maybe there actually is an argument for doing this. I still think the complexity and the scale of doing that is—or getting to the scale you would need to get to make it really worthwhile is super hard. Like, I have—and I haven’t spent a ton of time really diving into it, but I have lots of questions about, yes, you could do that, but how scalable is it really?
I very much have the same question about space. Yes, of course you could do that, but the scale at which, you know, we’re building data centers, and I—I just—I think both of those things will happen. I just look at the pie of total data centers by 2030, pick your date in the future, I’m a big believer of terrestrial data centers are still going to be like the vast majority of the way we build data centers. Like, I just don’t—I’m not sure how much it moves the needle. Are there edge applications and other things that for which it could be useful? Probably. I’m just somewhat skeptical that it’s going to be a big piece of the pie.
Shayle Khan: It’s interesting. I have sort of the opposite intuition from you about where the breaking point is. For you, you said like off-grid’s not the solution, but we probably will do some in the sea and in space. And to me, look, if we hit this impossible-to-penetrate ceiling on data centers connected to the grid—which we both agree is like what should—we should do as much of that as we can do, because that’s the right way to do it. But if that—if that hits some ceiling, to me, I don’t understand why we’re not just going off-grid then. Right? Like then, because you—the land, you go anywhere, right? Like, remove the constraint of the grid, you can put the data center anywhere. You don’t have to deal with operations in the middle of the sea or in space. It’s very scalable.
Brian Janous: But you still—you don’t get around—let me—but let’s say if the biggest barrier today is actually the community acceptance, off-grid doesn’t solve that problem at all. You still need a piece of land that you can permit and build a—
Shayle Khan: But if you remove the constraint of the grid, man, you can go anywhere. I mean, we’ve got some empty places in this country.
Brian Janous: Yeah, we do, but I still—I’m not convinced it takes that much of that barrier away. Like, I really don’t. And I could be wrong. Like, I definitely think—I mean, there is that sort of latency question. Like, if you really don’t care where you are, but I think if that’s the case, I think what you end up seeing is you end up seeing a lot of stuff built in some other countries, too.
Like, I saw a pitch recently for, you know, building a bunch of data centers in Mexico, and I was like, “That’s actually pretty good,” because this group had found a bunch of old industrial sites that had been set up to be big manufacturing facilities that never got built, and they were like, “Oh, that’s actually not a bad idea. Like, you know, bring the jobs there, bring the economic development.” You know, if you don’t care about it, and that goes to like how much you care about it being in the United States and how much you care about data sovereignty and other things, you know. So, we are kind of having a debate about, you know, how much we’re going to build in the United States, but, you know, there are other places in the world to build data centers. Now, there’s other places in the world that are actually harder to build data centers than the United States, too, so it’s not like that’s not a panacea either, just to be like, “Well, I’m just going to go, you know,” because it was like, you know, prior to the, you know, us attacking Iran, everyone was like, “Well, why doesn’t everyone go to the Middle East and build data centers there?” Now everyone’s like, “Hmm, maybe we don’t want to build data centers incenters in a in the middlea middle of a war zone,” right? Or in an increasingly unstable region.
So, so yeah, I don’t know. I think—I will say like I’m a lot more open-minded maybe than I was even nine months ago as to like, there have to be other solutions we pursue because all the stuff we thought we were going to build in this country, or thought we could build, a lot of that’s not going to get built because of all the things we’ve just been talking about. Like, I think that this problem is going to persist for the foreseeable future, and everyone who thought, “Oh, I just got a constraint in 2027 or 2028,” you have the same constraint in ’29 and ’30 and possibly into the early 2030s. Like, I think this will just continue to be a very, very tight market.
Shayle Khan: Yeah. I guess my perspective has just been, okay, if you think that constraint will be fundamentally caused by power generation, then space and the ocean make a lot of sense because you get your energy for free, ultimately. But if you don’t think it’s power generation, you think it’s some combination of community and grid, like transmission distribution and water concerns, things like that, like then that—off-grid to me just feels like a smaller step that’s like more plausible, and, you know, you’ve seen the maps of like all the solar it would take to power the entire country if you put it in the Mojave Desert, right? Imagine all the data center space you would need, even including a bunch of solar and a bunch of thermal generation and all that kind of stuff, if you just put it in the middle of the Southwest or United States in the emptiest part of the country. Like, you could do it, for sure.
Brian Janous: Yeah. Yeah, you could do it, in theory. I don’t know, I think we’re going to learn a lot, honestly. This is such a unique time because, you know, you have to sort of like throw out like everything we’ve sort of understood about how we develop things, you know, for the last 15, 20 years, and it’s like all this stuff that really made no sense, again, going back to Microsoft doing underwater data centers, you know, 10 years ago, and we were all like, “Ah, we’ll just put that on the shelf. Like, we’re not ready to do that.” Now you’re like, “Huh, maybe that’s not as bad of an idea as we thought.”
And I think we’ll get to—I think the next question was more about sort of that edge argument, right?
Shayle Khan: Yeah, and the edge one, it’s related, but I just want to be clear that the rationale for edge is not entirely the same as the rationale for that. Okay, the rationale for edge contains maybe you can do it with less community pushback, but it seems to be equal parts about like, well, maybe you get faster speed to power because you don’t need to go through the big long interconnection queues and, and you know, it’s a lot of things.
Brian Janous: Yeah, and I think that’s—I’m increasingly more bullish on that than I ever have been. I mean, I’ve seen a lot of pitches, I’m like, “That doesn’t make any sense.” And I’ve talked to a lot of investors, and they’re like, “Yeah, I keep people coming to us and saying like people are going to take like one floor of a commercial building and all this stuff,” and it’s like, “Well, maybe, but, you know, you’ve got floor-loading issues, and there’s a lot of…” So, I mean, a lot of those things are like, “Well, in theory, yeah, that ought to work, but when you talk about actually what it takes to build a data center, it’s pretty hard, even if it’s small.” So, you know, it’s not like power isn’t the only thing you have to think about, right?
However, I definitely think that given the—I mean, if you believe that, and just take any curve between now and 2030, if you believe the market is going to be short, in which means the the marginal megawatt that you bring online has real marginal value, then a bunch of things that don’t make sense in an unconstrained world might make a lot of sense in a constrained world. And so then that—it really gets in that price point. It’s almost like the same thing when you think about oil and gas discovery, right? People will say, “Well, like, oh, you know, we only have so much gas because like there’s a bunch of unproductive wells, and people aren’t going to drill in these things,” and it’s like, “Well, yeah, that’s true if, you know, if gas is $60 a barrel. If gas is $150 a barrel, there’s a bunch of things that people would be like, ‘Oh, I’ll drill that well all day long,’ right?” I mean, it’s the same sort of thing when you have that price signal.
And so I think we’re getting to a world where the price signal for that marginal unit of data center capacity is actually really quite high. Like, it’s a lot higher today than it was two, three years ago, right? Which means that a bunch of businesses that maybe you got pitched two or three years ago that you looked at and were like, “Hmm, interesting idea, but it’s never going to work because the break-even point is wrong,” right? Probably, you know, dust some of those off, and be like, “Actually, that works in this environment of really high marginal cost for compute, right?”
So, I am, you know, I still believe that the vast majority of the computation is going to be done in large data centers—large being defined as 100 megawatts or more in my world. Like, that’s how I would think about it. I mean, I think, you know, and definitely the people I talk to today are of that mind, even the ones that are desperate. They’re like, “I got to have capacity.” The lowest I’ve ever heard is like, “But it’s got to be 50 megawatts. I can’t spend a bunch of time doing like some onesies, twosies and things.” But I think, you know, I think there’s probably—in fact, I talked to one yesterday about—I think there’s probably this column like Cloverleaf-type customers that are going to go out and just say, “Well, we’re just going to go do this on spec, and we’re going to go find a bunch of locations, and we’re going to aggregate them together, and we’re going to create a portfolio of these things, and someone’s going to buy them.” And I think that’s probably true.
I just don’t know the scale. I don’t have a good sense of like how big of a business you could build doing that, but I think people are going to build some businesses doing that.
Shayle Khan: So many people are building that business right now, the edge Cloverleaf. I can’t tell you how many of them there are.
Brian Janous: Yeah, there’s a bunch out there. But just like Cloverleaf and and sort of all of our competitors, there’s a bunch of people that are going to try to do that, and it’s not going to go very well. It’s hard. It’s going to be a harder business to pull off.
Shayle Khan: And it’s a spectrum, right? There’s a big difference between 15 kilowatt data centers strung together at people’s houses up to, you know, industrial sites that can house 20 megawatts a piece or whatever, you know? And there’s a full spectrum in between there, and the density that you need, and the volume you need to aggregate, and how you finance the whole thing. There’s a lot to figure out there that’s kind of new and different from, you know, even what you—you’re, you know, when you started Cloverleaf, like, you’re building bigger sites than really people had built much of before, but it was kind of the same thing, fundamentally, that you were doing at Microsoft, right? This is a different thing.
Brian Janous: It’s a different thing. It’s a different sort of product, and, and I just—and that’s why I’m saying like I I think it’s going to be really hard to pull off. Some people are going to do it, and I think the ones that do it are going to do quite well, because again, the marginal cost of compute is just super, super high, and someone’s going to be willing to pay for it. But I don’t. I don’tdon’t I don’t envy your shoes as an investor to try to sort through which of those ones are going to work and which aren’t.
Shayle Khan: I was hoping you were going to tell me.
Brian Janous: Nah, yeah, sorry.
Shayle Khan: All right. Well, in the absence of you telling me what to invest in, it’s good enough to just have a good conversation. Fun as always, Brian. Thank you.
Brian Janous: Yeah, absolutely. Thanks a lot, Shayle.
Shayle Khan: Brian Janous is the co-founder and Chief Commercial Officer of Cloverleaf Infrastructure.
The show is a production of Latitude Media. You can head over to latitudemedia.com for links to today’s topics.
This episode is produced by Max Savage Levenson. Mixing and theme song by Sean Marquand, and Anne Bailey edits the video version of the show. Stephen Lacey is our executive editor.
All of our episodes are on YouTube. You can subscribe to Latitude Media for episodes of this show and Open Circuit. You can also find the audio version of this show anywhere you get your podcasts.
Also, if you haven’t been before, you should come this year to Powerhouse’s 10th annual New Dawn party. It’s on October 22nd in Oakland, California. I’ve been on the host committee for, I don’t know, most of the 10 years, probably, and I’m always there, so if you come, you’ll see me. You can grab your tickets at powerhouse-ventures.co/newdawn.
I’m Shayle Khan, and this is Catalyst.
