The Stack Overflow Podcast

Where we’re going, we don’t need fossil fuels

Episode Summary

Ryan is joined by Kieran Furlong, CEO of Realta Fusion, to talk about the future of fusion as a safe and sustainable energy source, the computation and scientific advancements that have made fusion possible, and how fusion technology innovations will address data and AI’s rising energy demands.

Episode Notes

Realta Fusion is making fusion real and tackling today’s clean energy issues with innovative technology. Follow them on LinkedIn.

Connect with Kieran on LinkedIn

Shoutout to Lifeboat badge winner zathura, who was awarded the badge for answering Type of triangle in MySQL.

Episode Transcription


 

[Intro music]

AD ANNOUNCER: AI's evolution transforms the way we now integrate information and analyze data. Raymond Yin of The Tech Between Us and Nemanja Jokanovic, Head of Sales at Snap Magic discuss how engineers are incorporating this technology into their designs and its impact on the future of engineering. Listen from your favorite podcast platform, or visit mouser.com/empoweringinnovation

RYAN DONOVAN: Hello everyone and welcome to the Stack Overflow Podcast, a place to talk all things software and technology. I am Ryan Donovan, your host for this episode. And today, we talk a lot about AI and data and what we don't talk about is the amount of energy that they're gonna need. So today we are talking to a company that is trying to solve those energy needs. My guest today is Kieran Furlong, CEO of Realta Fusion. So welcome to the program, Kieran. 

KIERAN FURLONG: Thanks very much, Ryan. Glad to be on.

RYAN DONOVAN: Top of the show, we like to find out how our guests got into software and technology, and I assume since you're in a sort of peripheral technology, it's gonna be a little bit of a different journey.

KIERAN FURLONG: Yeah, sure. Look, energy, it's definitely something I've been interested in for the bulk of my career, but I started my career as a chemical engineer in the chemical industry, but was also from an early age, passionate about our environment and, hence became very concerned about climate change and so on.

And so, even during my time in the chemical industry was always very focused on, how do we make economic growth sustainable, right? So I grew up in Ireland, a country that was relatively poor by European standards when I was born, now it's on the upper end of the scale for income in Europe.

And that's a great transition to see, but how are we going to do that for the next 8 billion people who don't yet have the lifestyle of an advanced economy? And really when you think about it as an engineer, I look at what are the base needs from Maslow's hierarchy: food, water, shelter, clothing, mobility– all of those things need energy, right?

So energy is basically the underpinning of our entire economy. And how can we generate and utilize energy in a sustainable fashion is something that I've been focused on for a long time. So after spending some time in industry, went back to grad school out in California, got plugged into the startup scene there.

This was just in time for the first clean tech boom. And so I was involved in that through a number of startups. And then after that kind of balloon had deflated somewhat, I found myself on the other side of the table in venture capital, making investments in sustainable technologies.

Took a reverse sabbatical for a few years at the University of Wisconsin, which because it's Midwestern, it doesn't trumpet its success as much as it should, but it's one of the top universities in the country doing a wide variety of research. And so I spent two years there looking at everything from quantum computing to advanced crop protection, but was always drawn to energy technologies. 

That's where I first met one of my co-founders, Professor Cary Forest and we had a discussion about fusion fission hybrid reactors, but ultimately I joined up with Cary and the team on the WHAM project, which was an RPE backed project at the University of Wisconsin. Ultimately, that led to the formation of Realta. I came to a realization that, yes, wind and solar are fantastic and we need to keep building them out as fast as we can, but we're already beginning to see limits on the growth for those technologies, and they will always have this intermittency challenge, right, that needs to be plugged with energy storage or on our other solutions. 

Fusion is a clean, always on energy source. And when I saw that, it's just like, this is something that humanity needs. We need to work on it. So that's when I decided to put my shoulder to the wheel with Cary and Jay and the rest of my co-founders, and we established Realta Fusion at the end of 2022.

RYAN DONOVAN: Yeah. When I think of fusion, a lot of people– I've been hearing about cold fusion as a dream for forever, and I know we've been making advances, but how close are we to the Back to the Future, the Mr. Fusion powered fusion car? 

KIERAN FURLONG: That is one of the great things about fusion when I'm trying to talk about it to a lay audience, or sometimes I'll go into schools and so on. There's no shortage of cool movie clips to be able to show whether it's yeah, Mr. Fusion from Back to the Future or Tony Stark and the Iron Man fusion core. That's if nothing else, I think those are helpful for getting people excited about fusion and realizing that there's this huge potential there if we can just figure out how to do it in a controlled  power plant.

RYAN DONOVAN: Right.

KIERAN FURLONG: So yeah, cold fusion, there is a bit of a hangover still from that in the nineties which was one of these scientific scandals. Essentially, I think just misinterpretation of data that then went a bit too far. So the old horry chestnut of a joke, infusion is the–it's 30 years in the future and always will be, but I always tell people that's– it's a multi-variable problem and you're only looking at one variable when you talk about time.

At a minimum, we need to talk about time and dollars or the resources that go into accelerating fusion. And that's been a huge change in the last four or five years on the private sector. We're seeing increasing numbers of venture capital investors and high net worth individuals and private investors coming to the fusion space, which is fantastic because now we've got this competition going on with a wide variety of different technologies and different approaches.

Some of them like ours, stemming from a long legacy of academic and very detailed research and some others that are at the cusp of something completely new and crazy, but still worth a try. So there's definitely been an influx of dollars, which is one of the things which I think is different from “Why now?” for fusion. There have also been a few major technology advances as well that are just unlocking and making it possible.

I often like to use the example of powered flight, right, so the Wright flyer. Two of the enabling technologies for that, like people have been trying to do powered flight for a long time before that.

But in the late 19th century, we had the advent of the oil industry, right? So I think 1871 or something was the first oil well drilled in the U.S. and so that industry came on initially to provide light kerosene, but ultimately started providing fuel for the internal combustion engine. 

RYAN DONOVAN: Right?

KIERAN FURLONG: Which was another major advance in technical innovation at the late 19th century. So by the time the Wright brothers were building their flyer, they had access to a very energy dense, liquid hydrocarbon fuel, along with a very light, high power to weight ratio internal combustion engine. You weren't gonna do power flight with a steam engine, but when they had that innovation, that was an enabler for them.

For our approach to fusion, we're doing something called magnetic confined fusion. Where essentially we create this magnetic cage to confine the plasma that the fusion happens in. So that's essentially a gas at the very elevated temperatures. To do that, we've had this major technical step forward with very powerful magnets. So the advent of high temperature superconducting magnets, we can now get to much higher fields still in a compact space ideal for magnetic, confined fusion power plants. 

RYAN DONOVAN: Yeah. And let's take a step back. Fusion is the opposite of the classic nuclear power. Right? That's fision power.

KIERAN FURLONG:  Correct. That's fision. Yeah, exactly. So yeah, what most people will think of as conventional fission is you take very heavy elements like uranium. A nice step of uranium and you split that. And again, same equation though  E=mc². If you've got a small amount of mass that gets converted to a massive amount of energy. But that's a splitting process. And it's also essentially, the first nuclear fission reactor was called a pile because it literally was a pile of fissile elements stacked up in the squash court at the University of Chicago by Enrico Fermi.

So you basically get a pile of hot rocks, put them together, and a reaction will occur. With fusion, what we're actually doing is taking much lighter elements and we are fusing those or basically smushing them together and the two isotopes will fuse, and again, a small amount of mass and that will be converted to a massive amount of energy.

But it has many advantages over conventional nuclear fission, one of them, for example, is there's no spent fuel, right? So there's no spent fuel rods that need to be confined and handled and stored somewhere for an impossibly long length of time. The trick though with fusion is it's much harder to get to work.

That also means it's much inherently a very safe technology. If we're not actively trying to control the plasma, it will instantly cool down and the reaction stops. So there's no possibility of a runaway fusion reaction either. So that's one of the main differences, or, again, yes, they're essentially opposite processes. 

They're both going after the power of the nucleus of the atom which is many orders of magnitude more energy dense than the energy that's released when we do chemical reactions like burning natural gas or oil, for example.

RYAN DONOVAN: Right. Yeah. It's essentially the process that happens instead of stars and the sun. When I think of that and I'm like, oh, you have a little sun happening in this, in a box or whatever, my first thought is “Oh, if that breaks open, man, there's gonna be a big hole somewhere.” But like you said. That's not gonna happen, right? 

KIERAN FURLONG: Yeah. Exactly. And you do get this kind of worry, which is again, one of the reasons why I think we need to get out and educate people and talk to people about fusion, get them excited about fusion, the potential that it holds, but also address any of these concerns that people may have around the safety of a new technology.

I think that's one of– through my career, something I've always fought against and been, just like we can't do it this way is when you have the experts in the ivory towers who think they're beavering away solving a problem for humanity, and then pop their head out in 10 years, I'm like, well done. Congratulations everyone. You can thank us. We've solved it all. It's like that's not a way to win the public support. What you want to do is get people to come along on the journey for you. 

In terms of, yeah, creating a mini sun on earth and what, what's that gonna happen? This goes back again to the nature of the fusion reaction. Part of the reason we've known about fusion since at least the 1920s, right when it was postulated, that's how the sun was fueled or powered. But to be able to do it in a controlled fashion has eluded us, thus far, right?

So we're almost a century on. But we're approaching that now. It is really difficult to get going and it's really difficult to keep going. An analogy I often use in explaining to folks who are– the difference between the two is I think you spinning a basketball on your finger, right? Try to do that. It's difficult. It can be done, but you've gotta like constantly actively control it. If you stop for a minute, it'll drop off. With fission, the reaction, once it gets going, the problem is it's harder to stop. So I view that as that's when you push a basketball down the stairs or down a hill and it just rolls and rolls, right? And you've gotta try and work hard to get it back or to keep it under control. So that's just one mental model of thinking about fusion: It’s difficult to get going. You always have to actively control it. We're talking about a tiny amount of mass in the fusion reactor at any one point in time.

So if something does go wrong, if there's a process upset, typically that would happen, like if the plasma wide touch the side wall of the vessel and it will instantly cool down and stop. So the real challenge of fusion is not just can we get it to occur, how can we keep it going right and keep it going in steady state in a way that we could then use that to power to do a power plant. 

RYAN DONOVAN: Right. I imagine with this sort of high physics approach probably expensive machinery, testing out ideas, probably done in a lot of simulations, right? 

KIERAN FURLONG: Yeah. Yeah, so I mean, both. At Realta,, we are one of the few fusion companies that has an operating fusion experiment. So we spun out of the WHAM experiment at the University of Wisconsin. We're now working in close partnership with the university researchers on that experiment, and so that's how we're demonstrating the conditions that we need to get to for an operating fusion plasma. But in parallel to that, yes, we have massive amounts of computational physics going on. So massive simulations, we call it our real twin.

And it's a fully integrated plasma solver where we've basically got all these different codes that you're trying to track, billions of particles on what they will do under certain conditions. So there's massive amount of compute going on. 

RYAN DONOVAN: Yeah. And it came to you all through Elizabeth Kaufman from AWS.

KIERAN FURLONG: Right. Yes.

RYAN DONOVAN: I talked to her on the podcast. How much compute is needed for these simulations? 

KIERAN FURLONG: Oh, yeah. This is where I would need the head of our computational physics team in here. But we are talking about thousands of hours on many different cores to run some of these simulations.

I think, yeah, and Elizabeth's been a great supporter of us and we've benefited from our partnership with AWS and being able to utilize their cloud computing service. But, yes, there's a few different levels of fidelity of the models. So the one that I was referencing where you're trying to track billions of particles, that's something known as particle and cell.

And essentially you're trying to track individual particles or really a cluster of particles in a cell and all of their subsequent interactions with the other particles as well. So you can see how rapidly that multiplies into a massive amount of computing. And so it is this massive parallel compute problem.

I know that the ability to do that, you need supercomputers or you need the equivalent of a supercomputer to do these things. But it's, again, it's parallel processing. When you get to that scale of compute, you can run a lot of these in parallel and that will get you to your answers.

So the amount of compute is obviously massive. The time is important as well, and our ability to do that when we need it was one of the other things that we benefited from with AWS. Previously these kind of simulations had only been done on large supercomputers at the Department of Energy, right?

And we still do run some stuff on NERSC, on the DOE supercomputer. But that's a– you have to apply, get in line, wait in the queue, right? Get your hours, run the models and so on. And also because it is a public resource, there's an obligation for publishing the results and so on from that.

So for a variety of reasons to be able to do this in the cloud enables us to do it. We can still publish if we want to, and we are taking an open science approach. But we may not want to publish everything. And as well, it's an on demand service as well. We don't just have to wait in line for access to a supercomputer. We can spin that up and run our simulations when we need them to be run. 

RYAN DONOVAN: Yeah. With the rise of AI there's a lot of parallel compute being built out at these data centers, right? 

KIERAN FURLONG: Yeah, exactly. And I think there's that's– you've got this virtuous cycle, hopefully with AI where AI is going to be a huge energy user, right? And we aim to be able to provide that quantity, massive quantities of on demand electricity hopefully right on site for these data centers, kind of a behind the meter type application, but to get there, we need the help of, the computational help to optimize how to fly the plasma, essentially, how to design the equipment, and even just sifting through the data. And I think some of those more mundane tasks are actually gonna be some of the areas where AI is gonna be really helpful, but it’s not very sexy. We're not gonna talk about it as much. But like just like one of our computational scientists, he always likes– he mentioned, and I just think this is hilarious, he talks about the experiment. So we run an experiment on the WHAM device, University of Wisconsin. We're doing those in short bursts because of the limitations of power supply and ability to take heat out and so on. So we're talking of the order of milliseconds, right? And Kai, one of our computations says, “Yes, like every experiment is like three months worth of compute, right?”

So you generate a massive amount of data with which each of these shots, and we've been running up to a hundred experiments per day, right? So that's generating a huge amount of data. The last thing I want to have are, very skilled PhD physicists doing is spending most of their time sifting and sorting data, and tagging it and storing it and so on. 

I feel that's the kind of task where we should be leveraging and can leverage AI to then provide the data for the physicists to dig into and come up with, solutions for optimizing the plasma conditions and ensuring that we get to net energy producing conditions.

RYAN DONOVAN: Yeah, that data processing labeling has seemed to really come out as a good use case for AI. Do you know what the size of that three months of data is? Like petabytes? Exabytes?

KIERAN FURLONG: Not quite yet. I think we're on terabytes for the amount of data that we're generating at this point. And look, it's also a factor of how many diagnostics we would have on a device. That's going to obviously then continue to extrapolate the amount of data that we'll be generating. And we're still effectively in the commissioning phase of the WHAM device. We will be installing new diagnostics as well. So I anticipate we'll continue to see more data generated from those. 

RYAN DONOVAN: Yeah. So after you have the three months of data, what's the cycle of updating, getting bugs out of the system, manufacturing new pieces, what does that look like? What's your development life cycle? 

KIERAN FURLONG: So some of the things– yeah, and I think development life cycle, I'm not, in terms of  a typical, software version, I'm not sure that’s as relevant for how we're doing things. So obviously we're using codes, we're optimizing codes, we're plugging them together and so on. We're building our own research tool versus building a software product that we're gonna be pushing out the door and so on. We do have this cycle where the links between the experiments and the computational physics, right?

So with the computational physics, we're able to test out things before we do them in real life on the experiment. And we look at what would we expect to see and then we're going and running the experiment, looking at that data and measuring up how good was the model at predicting what we were going to see, but also providing extra input to optimize the model as well.

And so we have this kind of virtual cycle going on where we can continue to improve our model. Our model is then used to identify, okay, what conditions do we really need to be trying to get to? So a big challenge, for example, in any plasma reaction is stability. How can we stabilize the plasma? And that's when we're already seeing examples of that. We run simulations. We are like what if we pull this lever, what's gonna happen, the stability of the plasma? And we then go and run it on the experiment, record all that data and look at how they match up to each other. So that allows us– and we've got a number of these different levers that we can use to stabilize the plasma, already on the experiment, the WHAM experiment. 

We can see which of those are providing the most promising results and then further optimize those in the model simulation as well. So we've got that virtuous circle going and essentially where we wanna get to– and we just recently raised our series A, and as we work through this investment period with those proceeds where we want to get to at the end is we are ready to go and build our own purpose-built fusion prototype, which we call Anvil. And we will have de-risked it as much as possible, right? So we'll essentially say, 
Hey, we've done the engineering design on this Anvil machine. The engineering design is on a physics basis that came out of our simulations. And our simulations have been validated against real world data in the WHAM machine. So basically we've derisked as much as possible without actually building the thing, and that's the next step, and that's what we go and raise our series B to do. 

RYAN DONOVAN: Oh, that's great. With these simulators and these models, does that all have to be custom or is there like actual off the shelf stuff you can use for that?

KIERAN FURLONG: Yeah, actually there's a fair amount of off the shelf stuff. So in the plasma physics community, there's a fair amount of codes that are widely used by the research community. And in many ways it's preferable for us to utilize those because it allows academic researchers to check our work, right?

And validate, again, back to when you mentioned cold fusion at the start, we need to make sure we're maintaining credibility. There's a lot of skeptics out there who still think, you know, fusion is just the mirage on the horizon, right? And we need to prove that no, we're making progress.

So utilizing these open-source codes is one way we're doing that. Now, how we stitch them together and how we integrate those, that's some of our proprietary software and the research tools that I was talking about earlier. But we do want to use tools that the research community is also using. It helps us when we're recruiting into our computational team as well, to bring in people who are already familiar with these codes. And we want to have the research community using the same code. So when we see advances made by the researchers in places like the University of Wisconsin, MIT, Princeton Plasma Physics Lab, University of Washington and so on, or Lawrence Livermore National Lab, or any of the national labs that we'll be able to say, “Hey, that's a great advance. Let's port that into the work we're doing on the computational side and see if we can, advance our ability to stabilize and confine our plasma.” 

RYAN DONOVAN: Do we're big fans of the open-source community here. Do you ever kick some of those advances back into the open-source community?

KIERAN FURLONG: Yeah. So in terms of do we do that, yes, absolutely. We do keep some things on the proprietary side of the fence, but our goal is to ensure that some of the optimizations that we've done for some of these standard codes, some of these things are still written in Fortran. That's how old they are, right? But to be able to push them back out to the research community so they can use the same tools that we're using. And again, it gets back to– there's benefits to both sides, right? We want to ensure that they're using the same tools that we're using so they can check our work, but also potentially find new solutions as well as when we're recruiting from universities and labs and so on, have them come and join us already up to speed on the stuff that we're using.

RYAN DONOVAN: And you talked about the signpost of the series B funding ahead of you on the map. What are the little steps you need to get to? What's the prove out before fusion is actually something you can put in a box and press a button. 

KIERAN FURLONG: Yeah. And again, we definitely, what we're marching ahead is advances on both the computational side and the experimental side.One of the kind of tangible things that's easy to communicate on the experimental side is the ion energy. In some ways you can think of it as the temperature of the plasma. Can we continue to heat up the plasma? Because there's lots of what sounds almost fantastical, orders of magnitude in the fusion world.

So we need to get a plasma to a temperature that's 10 times the hotter than the center of the sun. So it's, bonkers when you think about– we are talking about a very tiny amount of mass in the plasma because we're operating in a vacuum vessel. But showing how we advance on that temperature is one of the critical metrics that we're measuring on.

For fusion to occur, you need three things to happen, right? It's called the triple product. Basically you need to get the stuff in a plasma hot enough, dense enough, for long enough. So those are the three things that you need to measure. And by measuring our temperature, we're also measuring some of the other attributes as well, because we can only keep pouring energy into the plasma. One of the ways we heat it, for example, is essentially a giant microwave, right? So we have to keep heating the plasma. You can only heat the plasma if it's there, right? And you're not losing it, at the ends of the reactor or something.

So by showing we can increase the temperature, we're showing that we're also hanging onto that plasma and stabilizing it. So that's one of the things that we'll be working on. Additionally, we're part of the U.S. Department of Energy's flagship public private program for fusion it’s known as the milestone based fusion development program.

Within there, we've got milestones that we're aiming for as well. And some of those are on the computational side. So it'll be marching towards, okay, can we identify equilibria where we have the plasma and the heating systems, everything set up so that it's, this is a possible stable state for the plasma?

And then we want to show that, okay, identifying equilibrium is one thing. We need to go to the next step now and identify, do we have an equilibrium, which is also a net energy producing state, right? So if we've got all the right components in place and we're actually generating more energy out by fusion, then what we're pouring in with our heating systems.

RYAN DONOVAN: Yeah. With all those crazy temperatures I imagine there's no earthly material that could withstand them. Is this just magnets– like how do you have anything near that kind of temperature and have it not just dissolve? 

KIERAN FURLONG: Exactly. No, that's the key thing, and that's why we're doing magnetic confined fusion. So I'm sure many of your listeners are probably familiar with maglev trains which I first came across when I was eight or so, and thought were the coolest thing ever. You've got these trains floating on a magnetic field. We're doing something similar with the fusion plasma. So essentially you create this magnetic bottle, so a magnetic cage.

You have magnetic field lines, the plasma is a charged mass of stuff and so that cannot escape the magnetic field. So you're confining that within your magnetic cage, and that's at a standoff then from the actual, kind of material vessel that you have the plasma within. So the plasma first wall still needs to be, you know, some pretty interesting materials that could withstand process upsets and disruptions and so on. But again, we go back to, yes, we're talking about very high energies, very high temperatures for this plasma, there's only a very small amount of it. You can think of it like if you splash yourself with boiling water when you're cooking. A drop, not gonna do too much damage. The whole pot, not a good situation to be in. 

We're talking about tiny drops of plasma, if you like, within the plasma cage, within the magnetic cage. Having said that, the materials for the rest of the vice, we are going– we're doing deuterium tritium fusion. So those are two isotopes of hydrogen. You can essentially think of them as heavy water and even heavier water. 

But when we fuse those, the product is a helium ion and a fast neutron. The neutron will carry most of the energy that's neutral, so that will escape the magnetic cage. We capture that in a blanket that's wrapped around the fusion reactor. The helium ion will ultimately come out either random– we can capture the energy that's carrying as well through direct energy conversion.

 So the materials and all of those kind of components are made out of is still another interesting challenge for the overall fusion industry, not just for us and another area potentially that we could see AI and machine learning playing a role. As we've seen in things like protein discovery, there's definitely a number crunching advantage that AI has over human researchers when it comes to looking through the myriad of millions of options for how we combine different atoms in different compounds into the materials to construct a fusion device out of.

RYAN DONOVAN: And we've been talking around the AI thing. I wondered, do you have any sort of data or projections on the energy needs of AI? 

KIERAN FURLONG: Yeah, so look, I think the first thing I'll say is we are going to see continued increase in demand for data centers with or without AI. One way I make it real for people, if you just think about how many photos exist of your grandparents versus your parents versus yourself. And if you're a parent like I am, of our children, and it's just like there's an exponential curve for sure.

RYAN DONOVAN: Sure. 

KIERAN FURLONG: Right there. When you think of the hundreds of pictures we take of our children every year. So that's just one example. And all of those photos have to be stored on a server somewhere in a data center. You've got the really, technophile lean forward folks now in, in Silicon Valley and so on, who are tracking their every bodily function with sensors or small cameras and stuff. There's a massive amount of data to be generated there.

 Just that alone, the advances we could make in healthcare if instead of just seeing a primary physician once a year, who does a, you know, checks your blood pressure and so on, that once a year visit, instead you've got a base load measurement of a daily attribute, for example, right? Those can be huge advances in human health, but it's gonna require massive amounts of data storage and data processing. 

This is before we even get to AI which is particularly data hungry. It was a surprise five or six years ago when people began to realize that Bitcoin mining, just one cryptocurrency alone, was consuming 1% of global electricity. Everyone was kinda like, oh, hang on. Where did this happen?

RYAN DONOVAN: Yeah, I think if it was a country, it would be like number 40 in consumption.

KIERAN FURLONG: Right. Exactly. And I think AI is heading in the same direction for sure, but it's– at least people seem to be aware of it.

We've been involved in many conversations. Obviously, our partnership with AWS that Elizabeth mentioned on your show in March is a great example of that. One of our investors, Titletown Tech– that's a Microsoft partnership as well. So we are seeing interest from the hyperscalers in terms of where are they going to get all of their energy from in the 2030s and beyond.

And they recognize that not just from the commercial side of things of like to run this data center that we've invested 3 billion into, we need this much energy. They also recognize there's a need to maintain and earn a social license for all of this compute by saying, “Yes, we're looking into sustainable approaches for managing the electricity we need the water, we need the cooling that we need on the data centers.”

And I think there's so much scope to integrate a power plant with a data center, with something that can utilize the spent heat coming off the servers and so on. And that's before we even get into making the chips themselves more efficient, which has been a big drive in recent years as well.

 So I think we will see massive new energy usage for data centers, not all of it will be ai. There've been some, very high estimates out there. I heard Eric Schmidt recently and in testimony to Congress on some kind of very high numbers. We'll see. It's definitely going to continue to consume more and more of our energy.

My home country of Ireland. Yeah. Over 20% of the electricity there goes to data centers. Now, I think it's a great use of electricity. It's a great way to generate income. Very small piece of land. As long as you've got the electricity, which Ireland is ample wind power, for example, you can generate many, many dollars for a given acre of land, but you need the power supply. Wind power is great when it blows, it's not gonna be there all the time, even on the windy west coast of Ireland. So we need to have firm, always on based on energy as a compliment to wind and solar as well. 

RYAN DONOVAN: Right, and you don't want that increased energy consumption to raise the prices of everyone else's washing machine and lights.

KIERAN FURLONG: Right. And that's again, I think where there's a hearts and minds battle to be won. Where you can actually– the U.S. military is now looking at this, right? Where they're looking at locating data centers on military bases, right? On the one hand you think this is a good deal for the data center, for the company operating the data center.

They get all this additional security for free, if you like. And it's a win for the military because they're like hey, if we just allow these folks to build a data center and a power plant in the middle of the base, we could get our electricity for free. So there's a good win-win symbiosis there that I think you could replicate that in communities as well.

Particularly, you might look at remote communities with our approach, right, what we call our cosmo approach, compact, scalable, and modular fusion. We could go to remote communities where there's still sufficient people who are looking for employment, and that's one of the things that hyperscalers need in their data centers.

But also you could now justify a power plant to be built in that community and generate electricity. Maybe 90% of it is running to the data center, but the balance of it is going to the surrounding town and village. 

How we process data is going to be one of the massive global energy users the way steel production was in the 19th and 20th centuries, for example.

[Outro music]

RYAN DONOVAN: All right. Thank you very much for listening, ladies and gentlemen. It's that time of the show where we shout out somebody who came on to Stack Overflow, dropped a little knowledge, shared a little curiosity, and earned a badge. Today we're shouting out a lifeboat badge winner. Congrats to zathura for dropping an answer that was so good. They came to “Type of triangle in MySQL” when it had a score of negative three or less, and their answer got 20 or more. 

I am Ryan Donovan. I edit the blog, host the podcast here at Stack Overflow. If you want to reach out to us with questions, concerns, topics, or trends, you can email us at podcast@stackoverflow.com and if you wanna reach out to me directly, you can find me on LinkedIn.

KIERAN FURLONG: Thank you very much Ryan. So I'm Kieran Furlong, the CEO and Co-founder of Realta Fusion. People can find our website at Realtafusion.com and most of our social activity is on LinkedIn, and so you can navigate there to either my own account or our Realta Fusion company account on LinkedIn as well.

RYAN DONOVAN: All right. Thanks for listening and we'll talk to you next time.

[Outro music]