Can data centers be good grid citizens?

Greetings, everyone. This is Volts for July 2nd, 2025. Can data centers be good grid citizens? I'm your host, David Roberts.

If you have paid attention to the energy news in the last few years, or any news at all really, you will be familiar with the growing dilemma of data centers. The powers that be have decided that AI is the future, and we need an immediate, vast expansion of our computing capacity, so enormous data centers are lining up to connect to grids across the country.

But data centers are giant power consumers and, at least at the moment, extremely inflexible, which means they impose enormous burdens on the grids to which they are connected. Often those grids need upgrades, the costs of which generally fall on ordinary ratepayers. Lots of people are addressing this dilemma from lots of different directions, including new ways to power data centers, new interconnection practices, and special rates for large loads. One approach is to go after the

Inflexible part. Could data centers be made more flexible so that they can ramp up and down, move on and off the grid as circumstances require? Could they become good grid citizens?

That is what Verus, a sidewalk infrastructure partner's company, is trying to do. Develop a flexible data center that can help rather than burden the grid. The company recently partnered with the National Renewable Energy Lab. to put a 70 megawatt flexible data center through the paces. You can read the results on NREL's website. Now Verus is looking for commercial clients.

I am excited to talk with Varus's head of energy, Jeff Bladen, about how a flexible data center works, the extent of the flexibility, how much more it costs, and how these technologies might evolve in the future. With no further ado, Jeff Bladen, welcome to Volts. Thank you so much for coming. David, thank you. And it's a pleasure to be here. And I absolutely love the intro, Good Grid Citizen. It's exactly the way we describe it. And certainly in my career dating back to...

earliest days designing markets for PJM, it is something that we've been looking for, frankly, across the grid. The ability to have the demand side of the market. actually participate in and support grid services. It's, in some respects, the holy grail of the grid for a long time. This is coming for all large.

consumers, I think, not just data centers eventually, like all of industry, I think eventually is going to face the pressure to be more flexible. But let's, before we jump into the details of the data center year.

designing, maybe just say a little bit about, generally speaking, the kinds of sort of pressures and forces facing data centers today. What is the sort of state of play? It's obviously an extraordinary and transformational time. two to two and a half years with the emergence of AI and the enormous need to grow our data center capacity to support both training and eventually inference as a key.

Improvement in how the internet serves all of our day-to-day needs. We're on the precipice of something phenomenal. And data centers that really support all of our day-to-day lives are having to grow accordingly. But the strain, as you called out even just in your intro, is real. The power grid hasn't been built out to support anywhere near the growth of the data center capacity that we're talking about. And it hasn't had to, mainly because we've had this extraordinary run of gaining efficiency.

in our electricity consumption over the last 20 years that we've been able to essentially reuse for other growth purposes. There's been plenty of electricity demand growth over the last 20 years, but it's been offset in large part by... phenomenal improvements in the efficiency of other parts of our electric economy. Just thinking about lighting as one example, we've all transitioned from incandescent bulbs to LEDs.

So this is a key and extraordinary moment of change for the grid and for the tech world. I sort of wonder, I mean, right now, there's a bunch of data centers jumping on grids anywhere there's... Anywhere there's capacity. They're out looking at the states. But, you know, we're seeing more and more sort of grid operators or even legislators in some states kind of saying like, whoa, whoa, whoa.

Hang on. Like we need some capacity for ourself. You know, the citizens of our state don't necessarily want to take on a bunch more expenses just to accommodate these. data centers so i wonder like and you know data centers are now looking around they're like well if you can't rely on grids can we self-generate you know they're sort of groping for

Nuclear plants now. They're pretending they're going to build a bunch of nuclear plants really quick. They're looking around desperately, and I'm sort of just wondering, the way things are currently going, are we heading toward a moment? Where the grid is just going to say no more. You know what I mean? Like is demand for hooking these data centers up going to outstrip the supply of grid available?

I just wonder if we're coming toward kind of a crisis point. I think you've hit upon something important, which is to say it takes a long time to build grid infrastructure. It takes a long time to build power plants. It takes even longer to build high voltage power transmission. But what I think we're finding is that, like in many things, the key to driving innovation is a constraint. And that's actually what we're finding now. And so the innovation is going to, I think, result.

innovation in the way in which we power large customers, but I think even more so the way in which we more efficiently use our grid. I like to tell folks that the challenge in front of us isn't how fast can we build the grid. Of course, that will be important. But moreover, how are we going to get more out of the grid we have while we're building the grid we need? People do not appreciate the extent of capacity.

Slack capacity on the grid today. I think people do not understand just because it's not intuitive. Like if it's Slack capacity, where would we not use it? But it's like, you know, it's like, think about the auto fleet. It's kind of the same thing, right? Like it looks like cars are driving around all the time, but most cars are sitting most of the time. Indeed. We build our grid to serve.

everybody's needs at all times and that means when everybody's using power at the same time we have to be prepared to meet that need and we have to be prepared to meet it with some room for air. Power plants do go down unexpectedly at times, and so that room for air only adds the available capacity that's not used nearly as much as it could be.

And it applies to the transmission network as well. And so that's really a big chunk of what is driving the Veris Data Center business forward. And, you know, in my 25 plus year career in the power industry. The opportunity to be part of this team and to deliver what we're going to be delivering soon is extraordinary. It really is the opportunity to extract far more efficiency out of the grid, but not just the grid.

to start with this new design for data centers that actually extracts quite a bit of efficiency out of the data center itself in new and innovative ways. This is the logic for demand-side stuff in general. Basically, we build the grid to satisfy peaks. So you can either build more grid to satisfy higher and higher peaks or you can start moving some demand off the peak. That's sort of the flip side of this. And that's what's driving.

And we should say, you haven't said, I haven't said yet, but you were the head of energy for Meta for years. So I take it you have some experience like going out and looking for data centers and looking for places to put. data centers. Is that sort of what brought all this to your attention? You know, it's a real great opportunity to be part of the Verus data center team and certainly appreciated the chance to support a phenomenal team at Meta for a number of years.

My career really has spanned the energy business much more than the technology world, though. And that's really, in many respects, why I'm so excited about being able to be part of what we're delivering at Veris is that... From my time at MISO and PJM, it's certainly been apparent that if we could find a way to...

activate the demand side of the grid, we would be far more efficient and frankly, more reliable too. Yeah, yeah. Do you, and this is the final question before we get into the actual, the nuts and bolts, but I'm just sort of curious. You've been in the energy side now and a little bit glimpsed into the tech side. One of the things that's on the mind of a lot of energy people as all of this is happening is do the people inside these tech companies, they understand tech.

Do they understand the grid and how the grid works and grid constraints, et cetera, et cetera, in the way that someone like you does having been in the trenches of MISO, et cetera? What's the level of sort of savvy about the grid in that world? It's a very well-informed set of organizations. They have deep expertise that they've built on their energy teams. I supported a phenomenal energy team at Meta.

with lots of folks that came out of the grid world, out of the utility world, they understand the challenges. And frankly, it's why they're investing in new ways of approaching these challenges. It's also why they've been investing in clean energy solutions. You're seeing lots of announcements about large technology companies investing in the grid, and I applaud that. What Verus has the opportunity to do is to really start with...

a white sheet of paper design that allows us to unlock even more innovation. And that's really a unique opportunity for a new entrant. It's usually something that comes from new entrants too, frankly. Yeah, I sort of joke with people having been doing this for a long time now. I regularly these days have the experience of just thinking in my head like XYZ is a need out there right now.

I want someone to come along to my pod and talk to me about it, and it manifests. Within days, I'll get an email about it. One of those things is data center flexibility. It's been on my mind a long time. People just...

People have been assuming the inflexibility of data centers as they think about this problem. And I have been thinking, why are we assuming that? Why is that a fact of life? And so along you guys come. So very exciting. Let's talk about... the data center then so there's not a lot of public information about what you're doing but i've got this sentence it's got a novel grid aware power flow management system

with long-duration battery energy storage systems into a microgrid-controlled medium-voltage power distribution architecture. So that is a mouthful. That's three mouthfuls, and I want to talk about all of them. in turn. So as you say, this is a fresh sheet, but you're building a data center out of nothing and you're building it to be flexible. So let's start with that first piece, a novel grid aware power flow management system. What does that mean?

Well, let me start at the highest level to describe what we're doing. So first and foremost, we are building new data centers. This is not a retrofit to existing data centers. That's important. And we are going to be building, owning and operating them ourselves. And so we are.

a service provider to large technology companies. So at the most basic level, you can think about us as an outsourced data center capacity solution for big tech companies. They will pay us to lease buildings on our data center campuses. common solution used by tech companies historically as a complement to their self-build sites. So there are a lot of these third-party owned data centers out there now? There are, there are. But they generally resemble the self-builds.

that have historically been put in place by the large technology companies. And the reason for that is, and you alluded to this in one of your descriptions, You know, our grid hasn't really been at a place where we needed to have large customers be as flexible as we're talking about now. And that's because we've had all of this excess capacity that I was describing earlier has come about from a whole.

bunch of efficiency. So yeah, there are a whole bunch of third party data center companies out there. We are stepping into a very well understood, very well established business model, but with a new and differentiated product. And so let me get at what you are asking, which is what is all this jargon and how does it actually like work?

So what differentiates us is first, our new design will allow us to unlock maybe as much as 30% more compute capacity inside of a data center campus versus our competitors. And that comes from this medium voltage. electrical distribution network that's alluded to in the in the words you were reading and essentially allows us to move power around more seamlessly within a data center campus a bit more like a microgrid might now now this is like why

What are the other ones doing? How else would you run? I mean, you have to distribute power to the servers, to the racks. Something is distributing power. Is this just sort of like sticking in some AI so it's smarter? No, it's actually more straightforward. And it obviously does have some proprietary elements to it that we have to be careful about. But at its most basic, the way traditional data centers work.

is that you have a high voltage substation with a utility feed or a couple utility feeds usually and that high high voltage substation then has low voltage feeds to each of the data halls and those low voltage feeds are you know a fairly straightforward and simplistic circuit, much like if you really wanted to boil it down, much like you would find in your basement or garage, which has a bunch of circuit breakers, right? So each of those circuits goes back to the high voltage substation.

What that means, though, is that within a large data center campus, if one circuit is down, maybe it's down for maintenance, maybe you're replacing servers and upgrading them. There's all kinds of reasons why a circuit might be down. The power on that circuit is essentially unused. It's stranded in a sense. And so what we're doing is we're designing a new kind of data center that allows for underutilized power, underused power in certain parts of the data center to essentially be...

taken up and used in other parts of the data center. So I like to describe it as it's essentially like getting four buildings for the price of three for our customers. It's a huge benefit. Is there significance in it being medium voltage rather than low voltage? It is in the sense that it allows us to efficiently distribute the power to anywhere on the data center campus, to any of the buildings or data halls. Again, very much more like a microgrid would. Got it.

You've got power coming into your microgrid and it's controlled and distributed better than traditional. Better and more efficiently. And so in super simple terms. Our customers will be able to put more servers, AI or traditional cloud servers in the data center. And then anytime some part of the data center is using less, which is pretty much all the time because they're such big.

campuses with lots of things going on, the power doesn't end up being stranded in the underutilized part of the network. So you're making full use of all the power that's coming in. Absolutely. At all times. That's exactly right. It's also a good thing from the utility standpoint, because that actually improves the efficiency of the interconnection, that we're getting more use out of the actual interconnect itself. Right. Okay.

And that's different than the power flow management system? Or is that both at the same thing? You got medium voltage power distribution architecture and grid aware power flow management system. Are those both referring to the same thing?

They are in a sense, but there's more to it. So the second big innovation is that we're also installing utility-scale battery energy storage on our campuses. And what that allows us to do is to first and foremost provide campus-wide backup. So if there were a disruption in the utility supply of electricity, we have these large batteries available to immediately take up and supply.

a power to the servers on the campus. Let me ask about this. So take your 70 megawatt data center that you built for NREL. Grid power goes out, you shift to batteries. How long? What is the size of these battery installations? How long can you run your 70 megawatt data center on pure battery power?

On pure battery power, it's about four hours, although it depends quite a lot on what's going on on the data center campus. It's probably a bit longer than that. So short-term outages, basically, you can cover.

It is, but it's important to note that the design that I described earlier, this medium voltage distribution network, actually gives us... a huge amount of flexibility to plug any kind of backup technology in and essentially plug it in anywhere on the campus into this network and then it can supply energy to anywhere else on the campus using that same network.

So we have quite a bit of flexibility in terms of the type of backup we provide. It could be batteries. It could be battery plus some thermal type of generation if the customer needed some longer duration supply. Yes. I was going to ask about that in a minute, but just on the batteries, this is something else when I encountered it. I'm just like, well, why? Like batteries, it's something you come to.

learn if you study grids long enough. It's just that batteries are magic. They're helpful in all ways, everywhere. Like there's scarcely a grid problem you can think of that you could not help solve by just throwing a bunch of batteries at it. So why aren't data centers already...

installing big battery banks. That seems like the no-brainerest of no-brainers for data centers. It gives them a little bit of flexibility. It vouchsafes their reliability. It just helps in every way. Why are they not doing this? Are they? I certainly don't want to speak for other data center developers. I think there are probably some that are doing some of this. It really does.

required to make full use of them. It does require a different approach to the design of the data center. And that's one of the advantages that we're able to bring with the new white sheet of paper design is that we can integrate batteries and utilize them. more seamlessly and across the entire campus in a way that you can't with a retrofit. Back to the prior description of a legacy data center design with all of these separate low voltage circuits.

You really can't move the backup amongst the circuits. Oh, so a backup is attached to a particular circuit? Yes. Ah, so you would need, theoretically, if you wanted total backup. to have battery installations on every circuit if you really wanted full. You would, and it's less efficient. You would need to do quite a bit of retrofitting and re-engineering to the point where you're almost essentially creating.

something brand new. And the other nice thing about batteries as our primary backup source is that they're actually a lot more reliable than the traditional diesel generators. Diesel generators will have a degree of failure to start. It's not that uncommon for a diesel generator to fail to start. And I've often wondered what is...

Like the straightforward cost comparison. Like why? Because these backup diesel generators are ubiquitous in data centers. And I've just, again, so puzzled why they're not using batteries instead. Are batteries just more expensive than? Cheap diesel backup generators, is that why they keep using those? I mean, ultimately, what's important isn't the upfront capital cost, although that...

that plays into it. It's about the total cost of ownership over the life of a 30-year asset, which is what these things are, which is what a data center really is. And ultimately, on a total cost of ownership basis, We are very confident that we are cost competitive with all of the

alternatives that someone might look at. And so when they come to us... You mean on like a total data center performance level? For sure. And so when we talk to our customers about what it would cost to lease space, lease a building or the like...

We're offering essentially at market prices. We're not asking necessarily for a premium today, although we do think we're delivering a premium product. And so we think there's real advantages. So your idea then is you are selling data centers at comparable costs to other data centers, but with these improved reliability and improved flexibility, basically.

improved reliability, flexibility, and really importantly, dramatically improved efficiency. Like I said earlier, it's a bit like getting four buildings for the price of three. And that is because of this. power distribution system. This ability to move power in between circuits enables you to have sort of a full microgrid network such that you can plug either generation or storage.

anywhere on that network and have it serve the whole network basically. Yeah, absolutely. And I love that we're nerding out on the energy side of this, by the way, because I'm always stuck in this.

tech world talking about stuff that I'm only marginally qualified to talk about. But when we talk about the energy side of things, I feel so much more comfortable. Yeah, that's absolutely it. We're getting the most out of the power interconnect in a way that most other data centers simply can't just because of the physical infrastructure design. What is the sort of limit on the size of the battery installation? What are the sort of considerations there? Like from a performance perspective...

I sort of suspect that it's just like more is better kind of period. But like presumably there's some limits of like the efficiency you get with an additional bit of storage. Like what is the sort of limiting principle on the size of the battery installation? The key focus for how to size the amount of storage that you put on the data center is really trying to get us to the proverbial five nines of reliability from the power supply. And what's not well known is that...

The data centers at this scale that we're talking about are all connected to the highest voltages of the power grid. And at the highest voltages on the power grid, it's actually extraordinarily reliable. I mean, we're talking about four nines of reliability just by being connected to the grid. And so what we're looking for is to exceed five nines of reliability.

by having sufficient battery energy storage on site. There's also an element of this, and I haven't gotten to it yet, but a big part of what we're doing is we're not just serving the compute customers, we're also serving as a partner to utilities. And so we want to make sure that we're sizing the batteries so that we can meet the needs of utility partners. And what kind of flexibility do they need to drive efficiency and reliability on the grid? And four hours tends to be a sweet spot.

The NREL demo that you mentioned earlier, we were partnering with both Excel Energy and MISO. in the design of the testing regimen such that we were making sure that our flexibility was aimed directly at the flexibility needs of an RTO. of a utility that happens to run its own control area. And we proved that out during that demo. So to figure out the size of the battery installation, you need to sort of triangulate the data center's needs and the grid's needs.

So like for your 70 megawatt data center that you just built for NREL, what is the battery? What's the size of the battery installation there? In that case, we're talking about four hours of building scale battery energy storage backup. Do you have any idea what that is on a megawatt basis? It's 70 megawatts times four hours. So you can do the math. It's 280 megawatt hours. Interesting. And so...

One thing I'm really curious about is the role of on-site generation here, if any. So as you described, you have basically built a microgrid. And in a microgrid, you can just plug generation into it. anywhere and serve the whole microgrid. And I'm just kind of wondering if we could walk through, like, what are the considerations for a data center owner contemplating whether to add some generation or not?

My intuition is that batteries are doing most of what a data center owner would need unless they want to go fully off-grid. Is that right? So why would a data center add on-site generation? Would they? And if so, why would they? Well, it's a great question that I think is rightly placed to those that have. technology compute workloads that they want to back up. I think it will depend heavily on the kinds of workloads that they're backing up. There are certainly some mission critical workloads.

that folks may feel like they need more than four hours of backup. But again, when you look at the likelihood of an outage, you do get to five nines of reliability pretty quickly when you add the amount of backup. that we're talking about. Having said that- And even if you needed more than that, wouldn't you just add more batteries?

Well, you could add more batteries or it's actually quite easy for our design to just plug in generators later, right? There's really nothing that stops us other than reserving some space in the campus for future. addition of some kind of thermal generation, whether it's a fuel-based or a gaseous-based thermal generation. There is nothing that prevents us from adding it later. Again, because we have designed... Why not solar?

Well, you could add solar as well if you had space for it, if it was efficient to add it. It's usually more efficient. to do central stations solar that's grid connected just from the value of the panels themselves and getting the most out of them, putting them in exactly the right locations with the best solar gain.

grid connected at the right voltages in the right place. It's generally more efficient from a cost standpoint and an output standpoint to have them have solar grid connected than to put it where you happen to build a data center. Do you know of, like in your experience with data centers, is there any reason for one to want to be fully self-sufficient, to be able to run independently of the grid, sort of indefinitely? Because that's what you would get.

from adding a bunch of generation, basically. You alluded to this in your opening, and I think it's important to note that one of the key reasons that some are considering this island from the grid approach is simply to get...

power faster that is certainly in the equation for some but that's really the beauty of what we're doing here is you know we've got an efficiency too for that gives us the ability to extract a lot more capacity from the data center itself and allows us to go into places on the power grid.

that utilities can't serve right now because they can't meet a peak need or they have a transmission constraint. And so we're actually solving that problem as well by partnering with utilities to identify where in the grid they could serve. a data center like ours most of the time, but because we're able to be offline during the most extreme periods to solve for both peaking generation and some limited transmission constraints, we can actually solve the speed to power.

challenge as well. So that's really what would drive somebody off the grid is speed to power right now. Well, let's talk about that a little bit because this goes back to this paper. I think I've referenced it on the pod before, Tyler Norris.

It's a lead author. It's for the Nicholas Institute over at Duke. But basically just showing that it would not take much flexibility to get a lot more data center capacity. online, like minutes a year of flexibility can get substantially more data center capacity online. So maybe just talk about that. Is that paper part of what it's kind of sparked this for you? I'm certainly bolsters the case.

Well, the great thing about the Veris team and the work that Sidewalk Infrastructure Partners has done is that this is two years in the making. I know Tyler and I've talked to them. In fact, they asked me to speak at their conference earlier this year. about exactly what we're working on and how it lines up with the report that they issued. But the significance of it is real, that there is this un...

tapped potential for the grid if you can reduce demand by significant amounts, dozens of hours of the year. I don't think it's minutes. It's certainly not a single digit number of hours. But it's, you know, a few hundred hours of the year would be extraordinarily impactful. And that's what we're offering utilities.

We've had great success at partnering with a number of utilities so far and exploring and even beginning to focus on where we're going to site our first few data centers. Well, one other consideration here. And this is kind of what I keep circling back to is like when you take that paper into account and you take the needs of the grids into account, really you just need a few hours of storage attached to your...

data center to get that amount of flexibility to help grids. But there's one other reason you might want to go further. And this is also something I wanted to ask you about. A lot of these companies, including Meta, Amazon, Google, all of them really are on record.

with renewable energy commitments, basically with decarbonization commitments. Lots of those decarbonization commitments. were made before all this kicked off really in earnest, before the AI thing kicked off in earnest, before the load growth. thing kicked off in earnest before the new restrictions around grid and energy availability kicked in. So one of the things that is on the minds of people like me is just how well are those.

decarbonization commitments going to hold up in this frantic scrum for power. And so it seems to me that one of the things that your data center enables is... If the grid cannot provide you clean enough power to satisfy your commitments, you can put some clean power on site and meet those commitments. Is that even...

a consideration here. How deep are those commitments? How firm are those commitments? Are people really going to make, you know, multi-million dollar CapEx decisions on the basis of those commitments? Hey there, everybody. Don't worry, I'm not going to tell you about a new mattress or push a credit card on you. This isn't an ad. There are no ads on Volts. It is supported entirely by listeners like you.

If you'll indulge me for a second, I'd like to ask for your support. I started Volts because we're all surrounded by depressing news about climate change and misinformation about clean energy. And it's never been more important to share the stories of the real people on the ground. doing the real work of transition and all the ingenuity and courage and public spirit they are bringing to it

People are hungry to hear these stories, to learn from and find inspiration in them. I've heard from people who change majors or careers after hearing episodes of Volts. People using it in classrooms and community groups. Even state legislators who have passed bills inspired by specific episodes.

Sharing these stories matters. It makes a difference. If you have found value in it and want to help me continue doing it, I hope you will join the community of paid subscribers at volts.wtf. It's about the cost of a cup of coffee a month. If you don't like subscriptions, you can make a one-time contribution, leave a review on Apple or Spotify, or just tell a friend about Volts. I am grateful for any and all support. If you're already a paid subscriber, thank you. And now, back to the show.

Is that even a consideration here? How deep are those commitments? How firm are those commitments? Are people really going to make multi-million dollar CapEx decisions on the basis of those commitments? Candidly, they have. And so I think the proof is in what we've seen just even recently. Meta made their announcement that they're helping Constellation keep a nuclear power plant online. So that's exciting news to see.

What I think is important to circle back to, though, is that the cleanest megawatt is the one we don't use. Of course. And it's also important to recognize that. So not only are we getting more efficient with this approach to data center design, but... The fact that we can build, in theory, fewer data centers with a much more efficient approach like we're taking actually dramatically reduces the carbon impact of the data centers themselves.

carbon intensity of data centers, and any major construction project, frankly, is the scope three emissions, all the emissions that go into the concrete, that go into the... You know, the steel that's used, right? All of those things take energy and take carbon intense energy in some cases to produce. And so, as I said earlier, if we're getting four data halls worth of servers.

In three, that means one data hall didn't need to get built. Maybe a whole data center at some point doesn't get built. And so that allows for a far more carbon efficient approach to building data centers as well. Yeah, but don't you think, I mean, this came up before when... deep seek came out the chinese ai came out and it's much less energy intensive and then a lot of people were saying similar things like oh look it looks maybe the ai revolution is not going to take all our power

But then the response is just like, we're not going to take those efficiencies and just bank them. We're going to just plow them into more data centers. In other words, the demand is so high here that any possible efficiencies you get... are just going to get swept up in greater demand. I mean, we're still going to be building data centers as fast as we can for the foreseeable future, right? Even if they're 75% more efficient or whatever, don't you think?

All I know is that a more efficient data center is definitely an improvement over the alternative. Well, for sure. And so I think suggesting that we might consume efficiency as a result is probably missing. the point in that regard. I will say that it's also important to remember that because of the constraints, we're seeing this kind of innovation, right? And this innovation actually maybe doesn't come around or doesn't come around as fast.

but for the constraints we're seeing. So I think there's a number of good things that are coming from this, but really what we're trying to focus on is how can we deliver a better product. It's much more efficient in terms of the customer need, right? The customer's getting more for the same price. And how can the grid be better off than it was beforehand? I will tell you that having been in control rooms on hard days.

Having assets like these, and that's really what we're thinking about these new data centers as, is these are grid assets, not grid liabilities. Having them at the disposal of control rooms in places like... Carmel, Indiana or Valley Forge, Pennsylvania or Folsom, California would be an extraordinary benefit to running the grid more efficiently and more reliably. It really goes without saying almost.

i want to ask about the grid benefits but just to like make this point before we move on the point i was trying to make is if you're a data center and you want to be running on clean energy as a lot of these companies say and you can't find clean energy, this infrastructure, this power distribution infrastructure that you've created, makes it easier to plug clean generation in directly if you just want to boost.

the amount of clean electrons you're using, this data center can help you do that too. It's suited to that goal. It is suited to that goal quite well. And because of the flexibility we have, we can complement that with... shifting some of the consumption from a grid that might be carbon intense at times to periods that are less carbon intense to the extent that that is a desired outcome of the customer and most of our customers definitely have these goals so right so you could bank like

clean wind power at night in your batteries and then run on that during the day, maybe while the grid is dirtier, et cetera, so that you could even decarbonize a little bit just with the flexibility. Yeah, that's exactly right. And it doesn't even need to be moved that much. Obviously, there's quite a bit of solar energy being injected into the grid during the day. So part of this is also giving the grid operators new tools so that they can manage.

what is an otherwise intermittent fleet, and actually capitalize on the low-cost energy that's coming from wind and solar. I assume you've been talking to utilities in the process of this. I think listeners can understand intuitively why...

data centers as currently constituted are not good grid citizens. You know what I mean? Like they need anything that needs constant, always on extremely high amounts of energy. throughout 24 hours a day is inflexible and difficult for grids. So this gives them some flexibility, gives them tools. One of the things I'm wondering about is, are there ways for grids? or grid operators or grid managers to compensate the data centers for this flexibility? And if so, is that part of your sales pitch?

Here, that in addition to the benefits, you might also get this secondary income stream of sort of flexibility payments. Is that on the radar? At some point, it may be on the radar. I think for now, we're focused on ensuring that customers get the services that they're looking for out of data centers. And that looks like more efficiency, higher volume of compute workload for the same megawatts of interconnect, faster availability of power.

At some point, it may be a world where we're actually working with the utility to develop new forms of compensation for flexibility. But frankly, there are quite a number of flexibility payment structures out there that have. compensated thermal plants in particular for their flexibility for many years. So it's not new in that regard.

I do want to call out one thing that you said, though, and just say, look, there's nothing about a data center that makes it a bad grid citizen any more than you or I using our smartphones to live our daily lives makes us bad grid citizens. I think what we're talking about is turning what's today purely a consumed service by you, me, or a data center on our behalf into something that actually turns into a service back to the grid, right? That we're taking what is...

an obligation to serve and essentially turning it on its head and saying, we're going to serve the utilities as much as they're going to serve us. Well, I mean, maybe bad grid citizen is not right, but difficult grid citizen. But like the size... that makes them difficult. Also, if they can be flexible, makes them extremely powerful assets.

Do you know what I mean? This is not a small amount of flexibility. This is not like moving my EV charging from day to night. This is the ability to move, to ramp down or ramp up. you know, 70 megawatts at a time. And that's like, that is a large, just as it's correspondingly, you know, difficult to supply if it's not flexible, is correspondingly powerful asset if it is flexible. I couldn't agree more.

In other words, it seems like grids should be very excited about this. And they are. And that is why when we were standing up the... demonstration project at the National Renovable Energy Lab. MISO and Xcel were very excited to support the development of the testing regimen such that we could show that what we were doing, what we were creating the capability to deliver,

was actually going to be aimed squarely at meeting the needs of big grid operators. And that's what we showed. And we're excited to actually bring it into commercial operations. And the technological difficulty, is it just...

like shifting like once you've built this power distribution architecture i guess it's easy then to like shift what is powering a rack of servers at a given moment like from the grid to the battery without any interruption in service or glitches or whatever. Is that part of the sort of technological challenge here, how to switch back and forth power sources? It is, but it's actually simpler in some respects than depending on a diesel generator firing up and hopefully switch gear switching over.

Yes, this is true. Batteries are more reliable than diesel generators. This is a good lesson to take out of this. They are. And the kind of switchgear that you implement for our systems, the kind of infrastructure that you implement is actually designed to be used on a regular basis. One of the reasons why, in addition to environmental constraints from emissions, that traditional data centers aren't able to lean on diesel generation to get off the grid at times is not only...

Is it going to violate emissions permits, which theoretically you could relieve if you didn't care about that sort of thing. But moreover, more importantly, at some level, the infrastructure just isn't designed to be switched on and off on a regular basis. It's designed for what...

what its purpose is, which is to be a backup, hopefully never, but maybe tested once a month at most. Right, yeah. I mean, it's true across applications across the world, really. Anywhere there's a diesel generator, it seems like a battery would work better. But let me ask you a question. This is not something that Verus is doing or responsible for, but it's something that I feel like you might have some insight on having worked at Meta and dealt with.

data centers, I'm sort of wondering, this is something I ask all the hyperscaler tech people when I talk to them, how flexible can the compute load itself get? I get a lot of different answers on this. If you ask this publicly, what they'll tell you is, no, it's all crucial. It all needs to be up five nines all the time, all crucial all the time. But if you do sort of like...

ask them in private, they'll say, yeah, we can shift stuff around somewhat. Do you have any insight into that, having worked at Meta? How much... compute loads themselves can be ramped up and down or moved back and forth, or do you know what I mean? Add to the flexibility.

Yeah, I'll just couch my answer in the context of what's out there in the public domain. And that is that it is relatively limited, as you alluded to. And the reason it's relatively limited is that it is somewhat technically complex to me. that people are being served. In some cases, you might call them services that are discretionary, but one person's discretionary service is another person's crucial service.

I'll add that there are certainly plenty of services that are absolutely mission critical. We have crucial government workloads going on in many of the data centers by the hyperscalers. There is some amount of flexibility, but it is... modest. And I don't expect that to change very much. And candidly, to suggest that a tech company's workloads ought to be flexible in that regard is kind of like saying, why doesn't everybody subscribe to an air conditioning controls program at their home?

Some people do, some people don't. I think at some level, it's a question of why do we have a grid? Why do we have power supply? And it's because people are... getting real economic utility out of the electricity supply. And what we want to try and do, I think, is figure out how to be as efficient as we can in using what we've built as we're thinking about what else do we need to build to support.

continued economic growth. In some sense, this is one of those things where when I hear about it, I think, well, duh, why didn't we do this a long time ago? You know what I mean? Like making the power distribution within your data center, more like a microgrid, more sort of controllable and networked. And then adding battery backup just seems...

obvious on some level. I can't believe everybody's not doing it, but like, does this, does the combination of those two things get you as much flexibility as you think you're going to need? Or is there a tech? frontier here. Are there next steps along this line or are these two tools that you've brought to bear enough to get you the flexibility that you're going to need to get all the data centers you need online?

The short answer is we already have the solution fully engineered and designed, and we're now in the process of heading towards starting construction on our first couple of hyperscale sites. Do you have customers yet that you can reveal? I'm not making any announcements today, but we're excited by what the future is holding. I guess I'll leave it with that. But I think it's important to also say that...

that this is not a science experiment. What we're doing is we're taking well-understood technologies and we're integrating them in somewhat novel ways with a little bit of... algorithmic IP sprinkled over the top to actually make it all work together very efficiently and effectively and in particular to be able to respond very quickly to utility and grid operator calls for flexibility, and even to have that

call for flexibility be extended down to our customer servers when they have some of their own flexible compute capabilities that you were suggesting maybe we could have more of. So all of that together turns into a solution that is It's relatively straightforward, but it still required a couple of years of design and engineering work to get to the place where we could demonstrate it last month at scale and now get ready to start building it.

So, yeah, I think that's a good note, actually. None of this is particularly, no piece of this is particularly cutting edge. You know, microgrids are well understood. Battery backup is well understood. Shifting loads back and forth is relatively well understood. I guess the new thing here is just applying all this to these data centers. So obviously you would get benefits if we...

stopped building in flexible data centers and started building flexible ones. And you'd get sort of incremental benefits building up over time. Is there some hard barrier to retrofitting old ones to be more? flexible like does it just not pencil out is that is that a kind of a fixed thing is that ever going to change

You know, it will depend, I think, on individual circumstances. And I think in general, the innovative design we're bringing forward really lends itself to this kind of solution. Whereas I think... legacy designs are fit for the purpose they were built for. And in some ways, it goes back to the analogy you raised earlier, which is to say that We don't actually need every data center to be flexible any more than we need every car on the road to be a rental car that is used 90% of the time.

There are lots of reasons why we need some that are flexible, and then I imagine we will continue to have some that are built for a somewhat different purpose. In any case, we're excited by the... innovative approach. And we, like I said earlier, we've got this two for efficiency gain and gaining dramatic efficiencies inside the data center campus.

be able to deliver huge efficiency gains by connecting to parts of the grid that are otherwise underutilized. Does designing them this way affect in any way the sort of appropriate... scale of the thing? Like, does it enable you maybe to build smaller data centers? This is another dynamic around data centers that I'm sort of curious about. On the one hand, you get obvious efficiencies of scale.

Like, on the one hand, it seems like the trend is bigger and bigger. You know, you got these 100 megawatt, gigawatt data centers being proposed now. But I sort of wondered, like, especially... in a data center like this that is flexible, that has backup, that is sort of like a microgrid, you could see a smaller data center running like that, you know, like integrated into...

towns or cities like you know in downtown Seattle we have a data center where they use the excess heat the waste heat to heat to nearby buildings and I always thought that like how is that not tons better than spending a bunch of money on heat control technology. So I guess my question is like, does this affect how big data centers are going to be at all? Or is it orthogonal to that question?

I think it's more the latter, that the size of the data center is not really driven by this kind of technology. Although I will say that...

As you scale data centers up and up, the need for efficient backup actually increases. And so our solution... probably offer some real benefits at larger scales versus sites that would require... gigawatts of diesel generation backup and so we're excited by the prospect of of delivering all sizes to customers and you know we'll lease space at a building at a time so if somebody wants 70 megawatts they can come to us if somebody wants

multiples of that, they can come to us and they'll be able to get what they need from us and we'll be able to deliver all these benefits to them in any case. All right. Well, interesting stuff. It's not often you get to see sort of like a new industrial category kind of taking shape before your very eyes. It's been very interesting to follow these developments. Lots to come.

So thanks for coming on, Jeff. Any final thoughts? Just really appreciate the chance to talk to you today. I've been a fan of the pod for a long time and many friends have been on this. And so it's nice to finally have a chance to talk to you and share a little bit of what we're working on. Thank you for listening to Volts. It takes a village to make this podcast work. Shout out especially to my super producer, Kyle McDonald, who makes me and my guests sound smart every week.

and it is all supported entirely by listeners like you. So if you value conversations like this, please consider joining our community of paid subscribers at volts.wtf or leaving a nice review. or telling a friend about volts, or all three. Thanks so much and I'll see you next time.