In The Arena by TechArena - Submer's Gabriel Lazar on Heat Reuse and AI Sustainability

Episode Date: June 29, 2026

In this Data Insights episode, recorded live at OCP EMEA Summit in Barcelona, host Allyson Klein and Solidigm co-host Ace Stryker sit down with Gabriel Lazar, Head of Sustainability at Submer, for a thoughtful discussion on what real sustainability looks like inside companies building data center infrastructure at scale.

Welcome to Tech Arena, featuring authentic discussions between tech's leading innovators and our host, Alison Klein. Now, let's step into the arena. Welcome in the arena. My name is Allison Klein. We're coming to you from the OCP Summit in Barcelona. And it's a Data Insights episode, which means that I'm here with Ace. Welcome, Ace. Thank you, Allison. Great to be here with you as always. So, Ace, who did you bring with you today? And what is our topic? Very excited about our guest today. We have Gabriel Lazar.

He is Head of Sustainability for Submer, which is based here in Barcelona. I welcome Gabriel. Thanks so much. So Gabriel, welcome. You've called sustainability your angle on the infrastructure problem. What does that actually mean when you're sitting inside the company that's building data centers at scale if you're talking to one of your customers? Yeah, that's a great question. Sustainability is a broad topic, as we all know, and maybe a bit abused as a common.

So it is important to get more specific. And I think it's around first and foremost this concept of how can data centers be an energy node rather than a net consumer of energy. So it focuses on energy, it focuses on capacity, and it's also a social aspect of how do we get back to the community. So there are a lot of underlying topics without that. We can maybe break that down. But overall, it's about how do data centers not just generate compute, but also get a lot of give back to the wider grid and the community. Nice.

So I sat through a few sessions this morning here at the OCP summit. And as is true of every good presentation about AI these days, they all started with charts that are up and to the right. And there's a whole lot of accumulation of forces going on right now as AI becomes more capable, more complex, more resource-hungry, right? So at Solidime, we think a lot about where the data is going to live and this sort of quote-unquote memory apocalypse that we find ourselves in as an industry this year where nobody can get enough DRAM or RAM to scale.

But power, land, compute, geopolitical risk are all seeing sort of similar accumulations. How do you think about that pressure from where you sit in summer and what it means for how infrastructure actually needs to be built? Yeah, and that's a great one too. I think, again, like with everything, we can't be centered on our field. So sustainability, sometimes professionals get a little captured by things that are sustainability might be relatively broad compared to other topics, but it's still a little short-sighted. Something like geopolitical risk wouldn't necessarily fall into that category.

But I think we need to think more around systems thinking. So capturing everything that's going on, this is extremely complex. I think if we were to pick a field that is best at doing this right now is probably risk more generally. And more specifically, if you look at insurance, insurance companies. They're amazing at capturing this because they have so much data. They're able to crunch it so we can learn a lot from that. And it's all about having a multi-discipline. So people from all kinds of backgrounds, collaborating, communicating, understanding everything that's going on and how it comes together. Because I think that's where we see the Black Swan type events. The Iran War is a

great example. No one understood all the effects of that. And still, I think it's hard to understand all the effects of that in what else is to come. You know, right now it's memories, it's CPUs. Like, the vast array of scarcity right now in this industry is insane. And one thing that I just recently read is less than half of data center operators actually have sustainability plans in place. As a head of sustainability for one of the leading cooling companies in the industry, what would you say to that?

Yeah, we need to get way better as an industry, not just with the setting of targets, but also they're trying to achieve them because a lot of companies are going in the opposite direction right now. The resource intensity has to reduce in terms of what we're looking at today. There are ways of doing this. I generally believe there's a path forward to data centers and even the digital infrastructure being relatively lower impact than say steel making, car making, whatever you might call it. And what we need to see here is a genuine push beyond just performance and getting to AGM. or whatever that objective may be, to make this durable.

In the end, that's what sustainability means in its core, right? It's about how do we make an industry that, yes, growth is important, and I generally think there's a way of making growth in a sustainable way, but we need a balance for that. And I think that starts with targets, of course, especially for companies that are set in stone and they've been operating for years at a time. They know where they're going more or less.

I understand the growth is pretty impressive right now, but we need to start seeing that draft divide, with impact going down and the growth going up. One of the things we're continually hearing about in the U.S. market specifically is the limitations of the power grid and the challenges that exist in this kind of disconnect between, call it hyperscaler and enterprise ambitions and what's available. And there's lots of work being done to increase power budget and new power plants in the works and this kind of thing.

But I'm curious across the markets that Submer serves, how serious a concern are the power constraints? And are you seeing meaningful disconnects between operator ambitions and what folks are actually going to have to live with? Yeah. For our size, we're not a hyperscalor size. Let's be very clear here. So I understand they're looking at different scale of a problem. For a suburb itself, we still think that we can find a lot of gems, like hidden gems, excess capacity or old industrial sites, like flexibility. I think there's a lot of different strategies you can look out from the energy market space. And then find that capacity very quickly, not have that 10-year wait time that we're seeing

a lot of grid regions in Europe and in the U.S. It's just about being more energy market savvy, and that might be bringing your own energy. It might be a little bit of flexibility in the sense that, okay, maybe you have to ramp down 20% at certain peaks of the day. I think we need to think outside the box a little bit. Be prepared to tamper with those SLAs just a tad. Again, if we're talking about 20% per week for one hour of the day, it's relatively small. But that would already be outside most SLAs that operators have with their clients.

So how do we come together and collaborate just enough to say, look, if I do that, I'm cutting the grid line by five years. That's insane, right? The time to power right now is number one. So what if we do it in a way that isn't just about, okay, using the most of energy during peak times of the grid? We got to think about the data center industry as not alone in the space. We're working with so many other demanding industries that use energy at the same time. We can't just ignore that.

So I think there's a way to find that balance. So one topic I wanted to talk to you about was heat reuse. It's been getting a lot of accolades, but you have come out and said maybe it's not applicable everywhere. What is your perspective on that? Yeah. Once again, I try to be pragmatic here. I lead up one of the heat reuse work streams for OCP. And so I'm a big fan of heat reuse.

I think logically it makes sense, right? You're turning the compute, all the energy that's going in. I'm saying like 99.99% of it turns to heat. So why not use that, especially if we use higher temperatures, why not use that for other purposes? Unfortunately, it gets very complex very quickly. We talk about tons and tons of stakeholders, especially when we don't have a district heating network,

like many European countries have. This isn't set in place. There's no, who takes care of the infrastructure that goes on public lands? The government isn't used to running that. What happens when the off-taker is five miles down the road instead of two? Immediately the economics start to make no sense at all.

And so we need to be very targeted with this. I do think it's something that could go into siting processes for operators and should. But we need to find the most, let's say, low-hanging fruit applications. That makes sense economically for everyone from a CAP-X, OPEX, and even from selling of the heat perspective. And then we double down on that.

It might just be like 20% of data centers worldwide that are applicable. But I think that makes more sense than trying to hit it all and not getting anything. Don't want to force it. Definitely don't enforce it. Yeah. The Germans, the German lawmakers might be against me here,

but I just think they're asking for big amounts of energy to be reused. Still might work in a country like Germany that does have district heating. But yeah, like here in Spain, we don't have those kind of systems. And we don't need heat half as much as the Germans need it. So fitting those applications, I think there are other applications that we haven't really looked at enough. One interesting one we're looking at is desalination and wire treatment, which would be 24-7, of course. You need water all the time.

but again, it would be only relevant for data centers near the coast or near water treatment facilities. So again, that limits the amount quite a bit, but if you're building a data center near the coast anyway, why not? That's an interesting question. And then it kind of frames up a larger conversation. If we zoom out from heat reuse, potential desalination, there's this question of the relationship between data centers and the communities that they live in, right?

and what is their right to be there? And how do you see that relationship? How do you think about it from within Submar? And is that changing over time? Yeah, yeah. That I think is something that where Hebrew use really brings a lot of benefits. It's so much more tangible for the communities compared to, say, grid flexibility or ancillary services.

It gets hard to understand very quickly. And since the grid is so much more at the national or at least regional level, it's harder to see, but with heat reuse, you get that aspect, that tangela aspect a lot quicker. So I think what we can do there is bring it in as a way of engaging with the community from the get-go. So, okay, we have this data center project coming up in this area, but we want to help out with, of course, the cost of electricity and the local area. You can help reduce that, again, with these flexibility programs, bring your own energy programs. And then on top of that, we're going to get back a lot of that energy in the form of heat, which is going to be useful for, say, this pulp and paper manufacturer who is struggling to get

cost efficient because of the cost of heat or a local brewery that we can open up and you can do visits to the day of center, visits to the brewery. These kind of ideas, I know it's not always practical. It's not always feasible, especially with the need for speed right now. But I think it goes a long way for that permitting process, that regulatory process. and the buy-in, the social license to operate, I like to call it. Now, you guys are operating in Europe, in the Middle East, in the U.S. When you look at that global landscape, does sustainability represent the same challenge

in those geographies or something different? Yeah, very different. I think it's important to have that approach. It's not the same, and it shouldn't be treated the same. A lot of what we like to talk about is, first off, our data centers, liquid-cooled data centers in general, are a lot better placed to work in high-temperature or extreme. locations, places like the Middle East, places like India with high humidity. And again, the problems we run into are very different. Energy seems to be kind of across the board. Water, for example,

is a lot more complex. So in some water stress areas, we want to avoid using water. And that's possible, right? It's very possible. We have some deployments in the Middle East that have waterless and are running on a PUE of like 1.2. So it is possible to do. It is. beneficial to do. Of course, if you use water on top of that, you'd probably be getting a PUE of 1.15, right? So there's a use to water. It's just we need to avoid using it in water stress areas. If it's abundant and climate risk modeling shows that it's going to continue to be abundant, that's an important one, then maybe a little bit of water to reduce energy consumption might be a good thing. So it's not about yes, water, no water. I think it's important to say in the right places it might be

useful. And again, for the lawmakers, that's important to know because if they establish, you can't be a W-U-E of over 0.4 by X-state, but then, you know, in a place with abundant water, you're using more water at the energy generation source because, you know, fossil fuels, nuclear use a lot of water. You're just like replacing one bad with another. So we need to be smart about this. Those are just two examples. Of course, carbon emissions are important everywhere. But again, some green. grids are more dirty than others. So it's about figuring that out and really personalizing it to the region. There's an interesting analog here, which is, I'll take you on a bit of a tangent, but I got out of the hotel yesterday, rode a bike up to Alea, toured a vineyard up there. Oh, great place.

Yeah, it was beautiful. And what I learned was that this particular vineyard, about four years ago, they stopped watering the vines altogether, which I didn't know is possible. I thought you have to have water to grow these vines, but the climate is such, and they figured out a way that It's completely waterless approach. This is the first I've heard of that being done in data centers in the Middle East. That's fascinating. Okay, last question for you, Gabriel, which is if you could change one assumption that the industry is still operating on, something that will look back on 10 years from now and say, I can't believe we got that so wrong.

What would it be? I think it has to be around, and I know this is a tough one because it's hard to measure, but we need to not just look at PUE and look at IT utilization. efficiency and utilization is one that it's super secret. No one's going to tell you how much they're using their servers, right? Right. But that kind of information, if we were more open about it, we wouldn't be having these shortages, literally. Because if we had this more collaborative approach within the industry, here, maybe I'm being too much of an idealist, but this thing of running your servers at 30% or even idle for days out of the year, that's consuming about 20 to 25% of your energy.

It's doing nothing. And this is happening in AI. This is happening in just traditional compute. It's not going to go away. And we need to get better at not just measuring it, but then saying, look, at least internally, we're not doing something right here. We need to change these numbers.

We need to improve this output. And then the IT efficiency piece is, how can we do the same with less? And you know, right now it's just about getting denser and denser. And we're talking about one megawatt racks here. but at some point it's going to be like, who is using this and for what? At the same time, you have the Chinese using their models with technology from five years ago. So they're doing something differently and they're still coming up with a pretty similar result.

So we need to start thinking a little more efficient, a little more resource efficient. And I think it'll go a long way for innovation and also, of course, for sustainability. Gabriel, thank you so much for being on the show today. It was a real pleasure. One final question for me, where can folks engage with you who've been listening along and what to learn more? Yeah, platforms, yeah, like the classics, reach out to us on our website, summer.com. I'm on LinkedIn and all the fun classic ones. And then, yeah, if we're talking topics, energy, of course, getting that power and then using it efficiently, I'm really looking into that these days with partnerships across a board,

how we can create these solutions that come with a lot of different actors from different levels of the value chain and coming with something that is very different and very exciting. So please reach out and we'll talk more. Thank you so much for being on today. And as always, it's a pleasure. As always. Thank you, Gabriel. That wraps another episode in the arena. Thank you. Thanks for joining Tech Arena. Subscribe and engage at our website, Techorena.ai. All content is Copyright by TechEyne.