Episode 132: The Data Center’s Climate Redemption Arc with Lucend

Molly Wood Voice-Over: Welcome to Everybody in the Pool, the podcast where we dive deep into the innovative solutions and the brilliant minds who are tackling the climate crisis head on. I'm Molly Wood.

This week, we keep talking about data centers, how they are contributing to low demand, how future data centers might connect to the grid or not. But this week, let's talk about the data centers we already have and how to make them a whole lot more efficient than they are today with, as you might guess, data and AI. It's actually all way harder and more complicated than you would expect.

And then of course, we're going to talk about the bigger picture. What happens as data centers start bringing their own power, adding storage, and potentially becoming a flexible resource for the grid. Because at this point, if you aren't a grid asset, are we even friends? Let's go.

Jasper de Vries: Uh, so my name is Jasper de Vries. I am, uh, the co-founder and CEO of a company called Lucend. Uh, we've been around for about seven years, and, uh, in those years we've developed a, uh, data center optimization platform. So it uses actually machine learning, a lot of sensor data, 280 billion sensor readings from data centers globally, uh, from anywhere in their kind connect infrastructure. And yeah, we are able to make those data centers much more sustainable, much more reliable, uh, much more capital efficient.

Molly Wood: What, um, I actually wanna ask, uh, about your origin story too. You've spent something like two decades in tech across a lot of industries. What made you turn your focus to data centers and say, this is a, a thing that I can really tackle.

Jasper de Vries: I started off with this experience. I was approached by one of the bigger data center companies in, uh, Europe at the time, uh, with a simple question, what can AI do for us? It was sort of a consultancy assignment. I was running my second company at the time. I stepped into the, I was introduced into the wonderful world of data centers, are wonderful in the sense of that it's so complex.

And then I thought, wow, this is, this is, this is so cool. You know, this is so complex. And at the same time, the impact is just huge that you can make with something that data centers already have, and that's their sensor data.

Molly Wood: Okay. So yeah, walk me, that's a, a great segue into what Lucend actually does.

Walk me through, you know, if I'm a data center and I sign up, is it all software? Like you said, they have their own sensors.

Jasper de Vries: Yes. Um, typically all of these data centers, whether it's um, you know, uh, a slightly older, uh, data center, like we work with data centers that are 15, 16 years old. We also work with data centers that have just been commissioned as it's called, right.

So it just good life. But all those data centers have sensor data. Some more, some less. So we have a data center in Amsterdam that has about 235,000 sensor, uh, or data points as we call them. Uh, one of the, our facilities in Paris has only about 5,000 data points.

Molly Wood: Mm-hmm.

Jasper de Vries: But all of them have have data that we can use in some form to optimize their operations.

So what we typically optimize for is we started off with energy. Uh, so there's a lot of energy efficiency to be gained by just making your data center smarter. That's basically what, what we're doing. We're using all of that sensor data. We use very specific machine learning models, physics and for global networks, for example, to make data centers and data center operations, uh, smarter.

On average, uh, we save data centers now about 25% of, uh, energy. Uh, but on, with water, we're now on the average of about 30%, uh, water savings, but we also make them more reliable. We make the local operations teams much more effective, uh, because they know exactly now what they should be doing to run their facilities, uh, efficiency, uh, efficient than with less risk.

Molly Wood: Hmm.

Jasper de Vries: Um, yeah, so there's, there's just a lot of, a lot of potential.

Molly Wood: Why, I mean, I, I am delighted to hear that a data center or customer came to you and said, we would like to do this. It sort of feels like if, if what you're saying is that they're leaving money, energy, and water on the table, why weren't they doing it already?

Like they have the data, they have the technology. YThey have the incentive, one assumes.

Jasper de Vries: Yeah, it's, you know, it's… For one, data centers are very complex. So what typically happens is that there's this amazing team of, uh, engineers that design a data center in the first place. And there's a lot of theory, there's a lot of physics behind trying to come up with the most efficient and the, the best performing data center designs and those mechanical engineers, you know, it's, it's, they are mechanical engineers.

They've been, uh, engineering their way into a solution using data for it.

Molly Wood: Hmm.

Jasper de Vries: Yeah. That's relatively new. Um, that's all I can say. I think I've great people in the team that actually come from a mechanical engineering background and started working with data and then also started to see the potential.

Wow, if, if I combine my mechanical engineering knowledge about all of these, you know, the about physics, and then combine it with data to understand real world conditions. Wow. You know, there's, we open up a lot of opportunities. And those people, yeah, those are the people that are, that are working for us. They were, they ran into it.

And to be honest, all of our users get the same excitement whenever they see the possibilities and whenever they see how easy, relatively easy it is for us to open up the black box. Because each data center for them is still some sort of a black box…

Molly Wood: Mm-hmm.

Jasper de Vries: Uh, system because it's so complex. And being able to use all of that center data being, uh, allows you to understand your facility in totally novel way. So once they see what the potential is, they're hooked.

Molly Wood: Right.

Jasper de Vries: But originally, you know. They come from this other background.

Molly Wood: Right. I interviewed someone who, um, had worked on designing regenerative data centers as a concept with Microsoft, and she said, you know, that, that we're not designing data centers with sustainability as the outcome, but if we did, everything about the design process would change.

Jasper de Vries: Yes.

Molly Wood: And so you're sort of bolting on sustainability as an outcome after, which makes sense.

Jasper de Vries: And, yeah. And we're talking to, to data center companies that do everything, uh, including design. And what they are most excited about is the prospect of having a feedback loop about, you know, first they designed this data center, then they built this data center, and now they have this data center live. And to have the feedback about how this data center actually operates…

Molly Wood: Mm-hmm.

Jasper de Vries: Compared to what they thought it would operate like. Right, so it's, it's, um, and, and it's, it's really, uh, you know, it feels like a blessing to be able to work with these people that are so open-minded and are so curious, so to say.

I think by nature, engineers are curious, uh, but also, you know, doing something that maybe is initially something out of their comfort zone. It's like, Hey, we have this thing called data. We have this thing called AI. But once they see the outcomes, yeah, it's really, uh, it's really addicting, I think, for a lot of people.

Molly Wood: Yeah. I mean, it, it, it is of course wonderful to be working with engineers who can understand potential and get excited about outcomes. But it sounds like you're also making a really good business case. Like, I wonder how you, you know, do you think of this yourself as primarily a sustainability solution, but it's an easy sell because it saves data centers a lot of money.

Jasper de Vries: Uh, yes. And you know what, so one of the stories. I like customer stories, Molly, best to be honest.

Molly Wood: Great.

Jasper de Vries: Because, uh, I can, I can, I can hear, I can, I can see how enthusiastic they are. And one of these, the stories, was from, uh, Declan, one of our users, he runs one of the data centers we optimize in, uh, Ireland.

And he once, um, uh, we had a conversation and he once said, Hey, I use your platform as my daily newspaper. Right. So, uh, what do you mean? Well, you know, I, I come in every morning I look at your platform and it tells me exactly what has happened yesterday, but also what I need to be doing today, right? It's, we generate all of these optimization recommendations. We quantify the potential impact for him, and so he understands immediately, uh, what, where to focus on.

And by working with our platform this way, uh, in just one year, he saved $4.3 million and increased his energy efficiency by, uh, 40%. And that's just insane numbers, right.

Molly Wood: Yeah.

Jasper de Vries: And he's now using, he now starts to use our platform in, in, in very novel ways that we could never, uh, have anticipated, so to say that we're supporting by building all of these integrations to these other systems, everything, to make sure that all of these recommendations can be implemented as fast and as smooth as possible in such a way that it really helps him run his facility more sustainable, more reliable, uh, more efficient in terms of capital, for example, as well.

Molly Wood: Right? And talk about that decision to, ultimately the tech does, you know, you give recommendations, you give humans and operators the final call on what action to take. How did you, how did you decide on that versus maybe automation?

Jasper de Vries: So I'm going to be a little bit, you know, I'm going to exaggerate a little bit. The, but, uh, how to optimize in real time, you know, on a system level, that's, that's not new technology. That technology is there. It's from control systems. It's, you know, it's, it's, it's there.

The novelty is the intelligence that you add to the equation. But in this case, um, data centers are mission-critical facilities, right? So, uh, the one thing that they value most is uptime. Uptime requires a decrease of operational risks.

And, uh, so we have this, we had this beautiful example once that we ran into an optimization of a, um, brand new data center. State of the art. Was commissioned, and we got the data six months later. In those six months, they wasted in their own words, 1.6 gigawatt hours of energy.

One of the recommendations was, hey, you should decrease the pressure on your chilled watering from one to 0.8. If I would sit behind the building management system, one of the systems already in place that collects the data that we take, but is also able to change some of the settings in a data center, but it's, you know, really dumb. There's no, there's no intelligence behind it. But if I would sit behind it, behind it, I could look at the screen. I see exactly where to change this number from one to 0.8. It would literally take me three seconds to implement this.

Molly Wood: Mm-hmm.

Jasper de Vries: In practice, it took a 97 page MOP document, manual of procedure document outlining in excruciating detail, a 10 week implementation project for this change to happen.

Molly Wood: Hmm.

Jasper de Vries: Because they were very risk-averse. So to say that if they change something on the roof, what are the downstream consequences in this complex system that a, that a data center is.

So this is how we learn to appreciate the risk averseness, so to say. And today we have, say concept called modular automations. So it's really up to the customer, uh, if, how, when, uh, they want to implement all of these recommendations.

So suppose we have like 50 types of different recommendations. They've been working with our platform manually, for example, for a while, and they say, Hey, I've, I've implemented this recommendation so many times, I want to automate it. Sure. And we can, then we can automate this single type of recommendation out of the 50.

Molly Wood: Mm-hmm.

Jasper de Vries: We can also tell them, Hey, listen. We can also make it a hybrid one, right? So it needs an approval first before you implement. We can, we can also build this if you want these automations to run only between nine and five.

For example, you, you just schedule them between nine and five, right? So there's a lot of, uh, we, we've built in a lot of features that really make sure that operator, who knows its data center best, remains in control yet is able to benefit from all of the potential of applying AI into their daily operations.

Molly Wood Voice-Over: Time for a quick break. When we come back, we'll talk about how it's way harder than you might think to make even a small change in how data centers operate and how the future of data centers, as in all things energy-related right now, lies in flexibility.

Welcome back to Everybody in the Pool. We're talking with Jasper de Vries of Lucend.

Molly Wood: In terms of that de-risking you, you, you also, it seems like one of the things that might increase that confidence is that you call your technology transparent AI, where the customers can see the data trails, right? They can sort of, they can see the data trail behind any recommendation, know that it is specific to that data center as opposed to like, we have aggregated information from a bunch of different locations and made this recommendation. Like that seems like a pretty big deal.

Jasper de Vries: Yeah. And you know, and again, it always comes down to the risk averseness, right? So the most important thing here is to build trust. And there's a lot of, so coming from more of the AI background, let's, whatever people call it today, uh, the, these days, uh, it's right when you seem…

Molly Wood: You seem like a guy who might've called it machine learning back in the day.

Jasper de Vries: I, I still do, but…

Molly Wood: You still do, I can tell.

Jasper de Vries: Marketing sometimes requires me to use AI. But anyway, so, um, right. When we talk about AI, a lot of people immediately assume that it's, you know, the large language models, it's generative AI, it's it the reasoning models, but AI is so much more than that.

And it is very important, I think, to open up being able to open up the black box and build trust first and to show the people that the AI is actually a very different type of AI than you would use to, than you would use when you use the ChatGPT, for example.

And when you look at the backend, how we build that up is we have a very simple three-step process that, you know, is open at all times for anybody to verify. Uh, right?

So the first step is that we look around, identify all of these symptoms, and a symptom can be, for example, a, there's a chiller running. Well, that symptom in itself is not, is harmless, right? But then we get to the second step and then we, that's the di, the diagnostics phase. What we then do is we combine all of these symptoms and then with all of these symptoms combined, we generate the diagnostics.

So the chillers running. The outside air temperature, it's pretty cold. But then it becomes interesting because why is this chiller running then? Because a chiller is, is not that energy efficient compared to for, for example, a dry cooler that you might have also on that, on the data center. So what's going on then?

So we look continuously for all of these other symptoms that are able to explain or create a diagnostics. Why is this chiller running? And then the third step is that we create a recommendation, including a quantification. Like, okay, this is, this is what's going on. This is what you should be doing, and this is the impact if you would act on it.

Molly Wood: Hmm.

Jasper de Vries: And we, we present it, people can click on it, see the full context, uh, behind it, what exactly is going on.

And only when they trust this, they can decide what to do. Do I want to share it with my vendor, for example? Uh, do I want to implement it right now myself? Um, you know, do I want to automate it? And then, and when, when it's implemented, because all of the data is being made transparent. People can immediately validate the impact.

Molly Wood: Right.

Jasper de Vries: Right. So getting back for example, what I just brought up. This data center that, you know, actually had a three second change, it took them 10 weeks. Originally it was planned to take them 10 weeks, but during the, in the first seven weeks, we, we immediately provided feedback, system feedback to them saying, Hey, this is how safe it actually is.

So they, they, they decided in the end to skip the last three weeks and do the last change all in once, because again, we built trust with them. So, this is a typical example how we want our platform to work.

Molly Wood: When you're making these recommendations and quantifying impact, what is the, the impact that you're presenting?

Like are you also, are you talking about emissions at all or are you just sort of saying this is an efficiency impact, or do they have a full dashboard that's like, you can have this, this, this and scope three emissions, yay.

Jasper de Vries: Yeah. Scope three emissions are, that, that's, that's on our, that's on a roadmap.

Molly Wood: Okay, so scope one and two. I’m sorry.

Jasper de Vries: Yeah, no, no, no, no, no. Scope three misses is definitely on our roadmap because we actually, so what's really interesting when you look at data centers, when the first signs started popping up, that you know, all of these tech companies that pledge to be climate neutral by 2030.

Molly Wood: Mm-hmm.

Jasper de Vries: Um, and suddenly they said, yeah, sorry. We have to let that let go of the promise. When you looked at their annual reports, the reason was the scope three emissions of all of the hardware that went into those data centers.

So for, for, for us, it's very important to make sure that, that the scope three impact is quantified. We are, we are doing this.

To give you an example, we always work on optimizing, uh, fans, for example, right? So if we reduce fan speed, we reduce energy consumption of that fan. Uh, we also reduce the, um, the need to replace that fan because the remaining useful life is extended potentially, right? So if we are able to calculate how much that remaining useful life extends, then we are able to calculate what the impact on the scope three emissions is.

And so, so yes, the scope three emissions is definitely something that we influence. We cannot. We've not released yet to our customers, the quantification of it because of, of, of course, we want to be absolutely sure about it.

But we quantify impact on the megawatt hours. We quantify impact on water. Uh, we quantify impact on scope two emissions. Um, we're also working on quantifying our impact on CapEx and, uh, the operational expenses or maintenance. We tick actually all of the boxes that a operations department cares about, we tick all of the boxes. Right? So it's not just su, sustainability, it's not just reliability, but it's also the financial, uh, part here and the impact.

Yeah, that's, that's the beauty of it. Everybody talks about the, how much power data sets consume. That also means that if we are able to shave off just a few percent, the savings in terms of megawatt hours, for example.

Molly Wood: Yeah.

Jasper de Vries: It's just huge.

We just introduced a bunch of new recommendations, uh, water related. Where, for example, we are able to make a much more intelligent trade off between consuming water or consuming more, more power. So there's a lot of discussions about datasets consuming water. You know, or any, every ChatGPT prompt that you do consumes so much, so many liters of water. To be honest, not all datasets consume water.

So part of this story is not true.

Molly Wood: Right.

Jasper de Vries: But data centers is use water to decrease the, the amount of power that they consume, right? So, this is also somethingthat you don't hear.

Molly Wood: Tricky.

Jasper de Vries: If you, if you pre-cool the air using embedded cooling, for example, there's less need to use power to then cool the air for the surface.

So, however, this trade off, it's not e, not that easy. So we put much more intelligence to it to make sure that this trade off not only takes into account water energy, but also takes into account maintenance. Also takes into account the service levels that data center operators need to guarantee their customers now and, and a whole bunch of things.

And so we just deployed a few new ones. Every single recommendation and, and we can find dozens of recommendations in a week. Every single recommendation, uh, saves as much water, uh, as 113 households consume in a year. So, um, the, and, and, and, and again, there's, there's, there might be, I dunno, 50, uh, 80 of those recommendations every week.

Right? So this is just one type of, type of recommendation.

Molly Wood: I mean, you must feel great about this, right? Like this is because this is such a specific place to have impact. It is only growing. You know, time will tell how much it grows because there's a lot of pushback toward data centers, but this also feels like, it feels like something that you can really quantify as having massive impact. But that could even be an enabling factor for the, the growth of data centers globally.

Jasper de Vries: Yeah. What I think, um, so there's something, something interesting happening now with data centers. The definition of data centers is changing, meaning the definition of a data center originally is a building with power cooling, uh, and IT equipment in there, right?

IT equipment being the servers and eventually the chip sets.

Molly Wood: Mm-hmm.

Jasper de Vries: But, uh, the fact that power is not available makes that data center start to generate their own power more and more, right? If they generate their own power, they also need better energy storage systems, right? To store that power, to save that power maybe for later.

And what's then very interesting is that if you have power generation capabilities, if you have power storage capabilities, why can't data centers become a flexible, uh, utility to the grid?

Molly Wood: Yep.

Jasper de Vries: And it's, and it's really interesting actually, um, there's, now there's a new, new regulation that sometimes even requires data centers to be able to redistribute some of their power back to the grid whenever the, the, the grid needs it.

Or there's a specific, um, potentially federal regulation in the US that if you are able, uh, to, as, as a data center to stop consuming power from the grid for at least four hours, you will be able to speed up the, uh, the connection that you need for a grid or, uh, the approval process for the…

Molly Wood: interconnection. Yeah.

Jasper de Vries: Exactly right. So, and combining, and that's interesting from our sustainability perspective, um, there was a, a good article, I think mid-December in Wall Street Journal where they refer to a study from the HSBC, this, this, uh, big, big investment bank. And that said, what's really interesting when you look at the data center power need and you look at the technology that's available for data centers to provide them the power.

It is not gonna come from, for example, existing hardware like gas turbines because there's huge supply chain, uh, challenges, right? If you want to order gas turbine, good luck. By 2031, you will be the first.

There's promising technology, like small modular reactors. It's not that mature that data centers need today. It'll take at least 10 years before that technology is as scalable as data centers need today.

Molly Wood: Mm-hmm.

Jasper de Vries: So what's interesting, what that, what that report said for these data centers, what is available to them, that's renewables. That's solar, that's wind. Sometimes maybe even water, right? So if you combine that, the data centers are becoming their power generation using renewables, being able to store it, release it to the grid in times when it needs, when a grid needs it as well.

Molly Wood: Yeah.

Jasper de Vries: To me, that's a really interesting concept to really change the narrative, so to say, and really make data centers part of the solution instead of the problem.

Are we there? Now there’s a lot of hurdles, but the whole concept, I, I mean, we're moving in that direction.

Molly Wood: Mm-hmm.

Jasper de Vries: So I, I, I think it's a interesting, uh, discussion to have.

Molly Wood: Well, and I wonder is, is harnessing some of these savings, these energy efficiency savings and, and you know, using that to introduce flexibility part of your roadmap as a company? 

Like, uh, you know, I, I recently interviewed somebody who has installed a bunch of energy efficient, who is effectively acting as a utility for a lot of businesses across America franchises. And they are a 7,000 building virtual power plant.

Jasper de Vries: Yeah.

Molly Wood: So it sort of feels like you're in that business too.

Jasper de Vries: No, I, so…

Molly Wood: You're like, Nope, absolutely not. [Laughs] No thank you.

Jasper de Vries: So the whole interconnection, for example, is a topic in itself. Yeah. Is that our, our core business? No. What is our core business? What do we do? Is we optimize systems. We optimize complex systems.

And when you look at the, in the definition, again, getting back to the definition of a data center, as the definition grows, there's more and more complexity in that system. Right. So for example, if you want to optimize, uh, your battery storage, you will probably end up optimizing your discharge, uh, and charging, right?

Because that has an impact on the remaining useful life of this battery. However, maybe what you end up with is that, maybe you want to charge, discharge the battery, but maybe for financial reasons, you don't want to store the power that you're generating, but you want to redirect it back to the grid or you need to redirect it to your facility.

Right. So my point is the com, more complex, the system, the more complex as we call it, the combinatorial optimizations problems are, and that's where we fit. Right. Provide the intelligence to make all of the trade offs. The simple trade off that I just made about, hey, should we use water? Should we use more, more power, right?

So that's a simple, relatively simple trade off. But as soon as this definition of data center grows, the complexity grows and more complexity needs more intelligence in order to be able to run it in the most efficient manner.

Molly Wood: Right? Um. Then there's, there's the question of, of sort of your AI itself. Like, I think, you know, this is something that people tend to grapple with, that, that they're, they've gotten into a mindset that, that AI, as we think of it now, is bad for the environment. You know, it's, it's energy intensive and bad for the environment.

And then there are people on the other side saying, these tools in this data give the ability, give us the ability to abate hard to abate sectors maybe for the first time. How do you, how do you manage that tension? Like, are you, you know, designing the most energy-efficient software possible?

Jasper de Vries: Well, it, it is something that we want to take, uh, into account. Yes. Um, uh, because I, to be honest, I think the big part of our team is, is, um, you know, driven by the impact that we make on, uh, making digital infrastructure more sustainable.

Molly Wood: Mm-hmm.

Jasper de Vries: Right? So, uh, yes, it is something, uh, one of our team members, she's, she's now doing her masters in green, green computing.

Green computing is about, yeah, making, uh, any, anything software related as energy efficient from by or by design, right? So that's just an example. Um, so yes, we do take it into account.

And I'm, I'm a, I'm a more of a nuanced version. I'm not a black and white thinker, so I, you know, both sides. Both sides are right. The current AI or the current digital infrastructure that runs the AI was not, was never designed with energy efficiency in mind.

Molly Wood: Right.

Jasper de Vries: And so I do think there's a lot to be gained. And there's a lot of things happening right now to, you know, not just our solution. Our solution is not just the, the silver bullet that solves everything, right, but it is, I think, an important piece of the puzzle to make the digital infrastructure, and by doing so, AI, more energy efficient.

And this is, this is, this is what I like about what we're doing? Because we use AI machine learning to have a big impact. And, and I think that's a, that's a great story to be able to tell and to share with others that there's, yes, this is one of these examples that AI has a, as a good impact.

Molly Wood: Jasper de Vries, the company is Lucend. Thank you so much for the time today.

Jasper de Vries: Yeah, thank you.

Molly Wood Voice-Over: That's it for this episode of Everybody in the Pool. Thank you so much for listening, and hey, if you would like to talk to people just like you and me, join our Discord server. It's not just for gamers anymore. We are growing a community of climate tech, early adopters who are sharing tips and tricks and buying advice and new innovations and story ideas, and just kind of having fun. It's a hundred percent free, by the way. Think of it as like a bar full of people interested in all the same stuff you're into.

If you'd like to join, please click the link in the description right here in your podcast app of choice, or you can find it in the weekly newsletter, which you can subscribe to at everybodyinthepool.com.

And if you'd like to support this show directly, you can become a paying subscriber to the podcast or the newsletter, or both. To subscribe to the podcast for an ad-free version of the show there is a link in the description right here in your podcast app of choice. Thank you so much to everybody who has already done this, and together we can get this done.

See you next week.