Hi, we got Gio Albertazzi from Vertiv, the CEO joining us, and I'm very very excited. Gio's been very active in PTC, was one of our great speakers last year, but also sits at one of the great spots in the power side of our business in terms of really looking at one of the key drivers of what's going on in digital infrastructure. For years, none of us really talked about the whole power side of the digital infrastructure and Gio sits right at the crossing points of everything that's going on. So, Gio, maybe just give us an idea of how was 2025 in terms of was it a great year? Was it a booming year? How did it go for you?
Well, first of all, we are, of course, the entire infrastructure - power, thermal, prefabrication, services, the whole thing. And I have to go back to what we were saying to our investors back in October. So certainly a year very convincing for Vertiv, going in the right direction, and especially what we continue to notice is a strength in the industry. The strength in the industry is a constant, and that as I said to investors, as I told the investors in October, was very well reflected in our pipeline and the direction of travel. So the industry is moving very very fast, and our role in the industry is becoming more and more important. So we're really pleased, never satisfied, but certainly pleased with where we are.
Yeah. None of us, I think, really saw this wave coming that AI has brought, right? I mean, it's a, and how is it kind of obviously, the data center industry, you know, you look at it, it was probably steady 15-20% growth every year, kind of steady, yet buildings going out 10-12 years, and now we've got this AI revolution where you're building massive data centers very very quickly, right? I mean, when you think about it, it's probably what, two years to build a 1-gigawatt facility where that used to be when people thought about it was like this 5-year plan, and now you're actually doing these things in 24 months, and that must be putting huge strain on your business.
Yeah, I would say a couple of things on your comment, Bill. One is clearly the AI and the acceleration that it has, let's say positively caused in the industry, certainly something that is unprecedented. The industry has been almost slow in the past in terms of technological changes, in terms of the IT infrastructure. I mean, a lot of things were happening but nowhere near the speed at which we see them happening today. We are in a good position in the industry. We have very strong relationship with both the silicon side and with customers, be them hyperscalers, enterprise, or colo, and we work with most of them, certainly the most important of them, to really define the roadmap of technology two or three years out. So we're not surprised by this acceleration. Certainly there was the change moment for the industry, but we've been really ready to make sure that the industry was ready to accelerate, and that's a role that we continue to have in the industry.
Yeah. No, I think it is fantastic. And it's fun to watch, although it probably puts a... I think I look back and, you know, where you are, there's been some challenges obviously with COVID caused all the supply chain issues that happened coming out of COVID, and now you've got this sort of acceleration, it's almost like you haven't had no time to sleep in the last 5 years, right?
But no, no time to sleep, just the minimum number of hours to be always fresh and full of energy. But no, it's certainly exhilarating in terms of the direction the industry is going, in terms of what we have done in the last few years. And yes, the post-COVID was a challenge for many, but we're pleased with the direction of travel since, very pleased.
What's, what's as you're thinking about going into 2026, what are the big product pushes and areas that you think are going to be new that we will be reading about Vertiv doing new innovations or new ways of approaching some of these massive deployments?
Sure. So if we talk about technology, clearly everything continues to move as the power density, sorry, the rack density continues to move up, up to one megawatt and sometimes even above. Clearly liquid cooling is becoming even more important, but liquid cooling in and of itself is very important, but needs to be looked at in the continuum if you will of the thermal chain. So everything: liquid cooling, air cooling, heat rejection, the system. Then of course we have another big change ahead that is the gradual transition, at least for a part of the loads, to high voltage DC, be it plus/minus 400 or 800 volt DC, all things that will continue to evolve the infrastructure over the next few years. So not only the opportunity for growth, but also the opportunity for a technology shift that clearly is very very interesting for everyone in the industry. And the use of the word 'interesting' is deliberate in the sense that it offers certainly also opportunities.
Yeah. Yeah, I think because you're not only at the place of expansion, but also as power companies start to look at how they're actually approaching these things, a lot of cost is going to be taken out as you go through with the new technologies as you start to bring new technologies to the game, right? I mean, that's very interesting. We were talking to the Nvidia guys earlier today and, you know, when you talk about how much power is used for an AI search versus the amount of power that's used for a typical Google search, right, it's about 10 to 1, but it used to be 100 to 1. And it so the cost structure is actually coming down really really fast, not only from the chip side but also all of the supporting infrastructure that sits behind it.
The efficiency that technology and the evolution of the silicon is delivering is really awesome in a sense that will enable really to continue to grow tokens per kilowatt, per megawatt. So we are very encouraged about that. We are also encouraged by our ability and our role in terms of delivering the infrastructure, be it power, cooling, or the entire data center, to serve those loads. So we do not see the infrastructure becoming simpler. We see the infrastructure becoming more nuanced, based on the type of loads your data center will have to handle in the future. So there will be a lot of designs, a lot of types of data centers depending on what's the strategy, what type of loads, what type of silicon the data center will utilize. We track ourselves in terms of content per megawatt, and we're explicit with our investors in terms of content per megawatt. But you're right, cost per token is going definitely in the right direction.
Yeah. And it is interesting because that may be we may be talking about that next year when we're sitting here in PTC or whatever next time we're having these discussions. Because it's interesting. Everyone's talking about, okay, how much does it cost to build a megawatt of white space, and they all have these different metrics, and the token, I think, will be the next one that we look at. So tell me, when you think about, you know, everyone talks about, and I know this isn't your main area of expertise, but the power in the United States in general, right? I mean, it's, and around the world with AI growing as fast as it's growing. What are we going to do? I mean, it looks like we're going to have a real massive power shortage, doesn't it?
Well, certainly power is central. Power generation and availability of power is central to every conversation in the industry. We've been very very open about the fact that I personally do not see power availability as a ceiling beyond which you cannot go. It's one of the pacing items, pacing the speed at which the industry can grow. But again, there is stranded power that is being harvested if you will. There is a lot of ingenuity and capital being deployed. So if you take North America but also other parts of the world, and not all parts of the world are power constrained, but where power constraints exist, we see a lot of projects with power generation on site, with power independence in mind, or at least power shaving, and battery energy storage systems to make sure that the data center can cope with specific situation of power in the region. So we do not see the growth stop because of power. We see a lot of capital and ingenuity being deployed. That also, of course, changes the architecture of power infrastructure for a data center, having for example to connect a gas turbine to a chip is not straightforward. So the power architecture and the thermal architecture need to be designed accordingly. We have announced a partnership with Caterpillar exactly going in the direction of defining reference designs on the power and the thermal side to be ready to support our customer and the industry in that respect.
Yeah. Do you think, I mean, some people say that one day we're going to be the power generation and the data center may be all co-located in the same spot, right? Everything will be behind the grid. That probably doesn't make a lot of sense if you're going to do things at the edge and if applications are going to be close to customers, but I could understand that for maybe the bigger language models, maybe you get to that. And maybe that's what potentially even Google's thinking about. Google announced they're buying a nuclear power company, right? Do you end up in a place where with these large language models you actually have the power completely integrated into the data center, like essentially going beyond everything behind the grid?
Yeah, I think, and what we see is this industry doesn't move binarily, despite if you will the thought process sometimes is binary. 'Oh, it is all kind of bring your own power' or 'it is all grid'. And I can make many other examples of this binary nature. The fact is that the industry is large, becoming larger, and has a lot of smart people that come up with different solutions, and we will have a mix that goes from pure grid to pure self-sufficiency, and everything in between. And we see cases across the entire spectrum right now. So it will be a good hybrid environment, or let's say a broad range of solutions, and that will be true for the future.
What do you, so you know, just to talk about the thermal side of it, right? So the new GPUs are, you know, they're water cooled, creates a different environment that you've actually got to create inside the data center, right? And I mean, what does the future data center look like? I mean, and how do you deal with, because it's interesting, you can have air cooled data centers, probably I would say probably 90% of the data centers today still are using air cooling technology, and 10% maybe are of the total volume is water cooled. Where do you see it looking?
Sure. So clearly what you referred to is more the current establish. When you look at how new data centers are being built, the majority of the cooling is de facto liquid cooling. I always indicate that pretty much 70% of the loads will be liquid cooled, and 30% of the loads within the same data center will be air cooled. Of course, in all cases, be it liquid cooled and air cooled, then you need all the heat rejection, be them chillers, dry coolers, or cooling towers, probably will be chillers and dry coolers. So again, I don't want to be one-size-fits-all. Different data centers, an inference edge data center, a pure training data center, or data centers that mix the two and maybe sometimes do also traditional cloud loads will have different designs. But clearly there is a trend of more liquid cooling. Now that liquid cooling is today absolutely predominantly one-phase water-based. The future could be more complicated.
