Good morning. I'm Andrew Obin, BofA small-cap industrial analyst, and we have the management of Vertiv with us. We have Giordano Albertazzi, company's CEO, and David Fallon, company's Chief Financial Officer. The format is going to be. We have Lynne Maxeiner in the audience. Thanks so much for making this possible. This is our top idea. It's on the US 1 List. So very excited to have the management team here. We're just going to go right into Q&A, into fireside chat. Yeah. Thanks so much.
Thank you.
Thanks for being here. So, maybe we can start with the big picture. I guess about 1 year ago, ChatGPT flew into the mainstream, and NVIDIA launched its H200 AI chips. Gio, you've been in the data center business over 25 years. What's the closest historical analogy to the current AI investment cycle?
It's slightly different than anything we've seen before. If we go back to the probably two biggest cycles, the early one is the Internet acceleration, kind of a very early 2000 that didn't, you know, end very well with a certainly a bubble at the time. Or we can go back to the cloud business acceleration, the cloud services acceleration. But this is something different, yet. I think we have a much more maturity in terms of the industry as a whole, certainly way more mature than it was when the whole Internet acceleration first happened. But also we have a much more robust capital structure behind that. And another thing that we see is that there are the use cases and the adoption of AI is actually happening.
So, I would say different from what happened before and faster than what we saw happen with cloud computing in general. So if anything, it's building on the experience of cloud computing.
Depending on sort of the definition you use for market share, but we estimate that the top three vendors, Vertiv, Eaton, and Schneider, have about half the market. None of you have done a significant amount of M&A in the data center space. Eaton with Tripp Lite is really the only large deal in the past few years. So I guess the question is why.
Well, if you, if you look back the last couple of years, couple of, two and a half years, as, as Vertiv, we have done two significant acquisitions. One, large-ish, at least, for, for our size at the time, that E&I acquisition that significantly stretched, portfolio. And, and another one recently, important from a technology standpoint, CoolTera, that again, strengthened our, our liquid cooling presence. M&A is part of, of, our capital allocation strategy. We're much more, let's say, structured in the process than we, than we were in the past is a, is a fundamental piece of our, let's say, the way we run the company. So, I can't speak for the others.
Yeah.
But we will continue to do the moves that are necessary to continue to lead the space.
Maybe we've been getting a lot of questions on CoolTera. Just maybe just expand a little bit what it is strategically you've achieved with CoolTera acquisitions. Like what is it IP? Is it capacity? What does it give you?
It's, it's fundamentally IP. We've been cooperating and partnering with CoolTera for quite a few years before we, before we pulled the trigger, so to speak. And just simply because, in, in a very early stage of, of a technology, you have to add your bets, your internal bets, your organic and inorganic bets. And that's what we did. Definitely, acquiring one of the leading technologies in the liquid cooling space and adding to, to our portfolio and our thermal, thermal chain, the in-rack part that simply was not there as a, as a market, as everything was air-cooled and was just, intrinsically built in the, in the rack. So, not so much capacity.
As we explicitly shared with the investor community, we are rapidly making capacity available for the market and for our customers, leveraging our procurement and manufacturing footprint to expand the liquid cooling capacity.
So we'll definitely talk about capacity later, but maybe we can just talk about, sort of AI-related demand. So Vertiv sized the AI benefit at 300-400 bps benefit to growth. What does that imply for AI penetration of the data center market by, let's say, 2028? 10% of megawatts? 20%?
At 2028 is, we'll go fast, but it's still fairly, fairly far away. I would, I would believe probably, more than that. I don't know how much.
More than 20?
More than 20.
Okay.
I don't know how much. But one thing is, from an infrastructural standpoint, it could be it will not be a clear delineation of what is what. One thing that we see is a lot of the data centers that have been designed today are designed for being hybrid and being future-ready. Not necessarily a data center owner today knows what the load in terms of high density, low density, liquid cooling, air cooling, how that profile will change over time. So we tend to talk with our customers about how to make that transition gradually possible with the infrastructure. That has implications in the way you design the cooling part, the way you design the power part. But again, these conversations confirm the fact that there is a fluidity out there.
You know, well, what exactly is AI? Is AI 100% high density, or is it a little bit of a mixed situation there? So a lot's yet to be defined in the future.
Gotcha. As probably everyone in this room knows, orders were up 60% YoY last quarter. And, even our strategist knew that number. So, guidance is for sequential decline in second quarter. And I think part of it is timing of sort of larger orders. But can you just talk about what, what, what, what is a large order for Vertiv at this point? Is it $50 million? $100 million? $200 million?
Well, all the numbers that you mentioned are big orders. So that goes absolutely without saying. I would probably not focus so much necessarily on the size of the individual PO. That could be almost, let's say, deceiving a little bit. Sometimes you're going to have. We like to think in terms of buildouts. You're going to have a giga site or anyway a very large site that is built in chunks. So it's very important to be there with the first step so that then you can roll out because data center owners like to have the same infrastructure across an entire campus. So it's not so much about the individual PO, but it's about what that represents in terms of the opportunity that you have across the buildout in specific.
So, we talk about big campuses. We talk about large accounts. And yes, $50 million, $100 million, $200 million are certainly big, big orders.
But what do you think is the key driver of the cadence of order activity through the year-end? What do you think is going to determine it?
I would say, from here to year-end.
Yeah.
You were saying? Oh, it's just a pipeline and how the pipeline translates into orders. So, again, we've been explicit about, you know, year- to- go, book- to- bill.
Yeah.
Bigger than one. And again, the rest is pipeline translation into orders. What we noticed and we were vocal about that is the pipeline, let's say, speed is accelerating.
Right. You did say last quarter that pipeline for AI project doubled between January and March. Generally, Vertiv would get involved after the land-purchase electricity agreement. Apologies. Early layout is done. So how should we think about the pipeline being ready to convert to orders over the next six to nine months, given that, you know, what has to happen on the ground?
I would say that normally we think in terms of a cycle in the pipeline, all that kind of orders, between 3-9 months.
Yep.
So, yeah, that conversion should be there. Mind you, there is always a win rate attached to that. But all the rest being equal, we are positive about the translation of that into orders going forward.
Gotcha. You target having 25% spare capacity in your own factories. So where did you stand on that metric in the first quarter? Second, 2024 CapEx guidance did go up a bit with first quarter results. Is capacity keeping pace with the backlog?
So there are multiple angles to your question. First of all, that let's call that 25% is a wiggle room, a margin that we like to keep in our capacity as we design capacity for the future. Our capacity clearly is consistent with the trajectory of the business and is consistent with the trajectory of our pipeline. Hence, it's consistent with our backlog. So expect an ongoing capacity growth trend for Vertiv. That 25% is built in the way we think about our infrastructure. So we do not design for 24/7, seven days a week. We typically design for a little bit sub- two shifts. On average, that's not necessarily true always for every line of business in terms of exactly where we are in terms of loading.
Because we know one thing is sure about a forecast: it will never be exact. So you will have peaks or you will have things happening a little bit earlier or a little bit later than you forecast. And that 25% gives us the chance to follow demand and absorb the difference between forecast and the demand. So that is always going to be there.
Just thinking about just capacity additions, because, you know, I think you sort of talked about doubling it. If you look, if I look at your announcements, there really does seem to be a lot of announcements coming out of Vertiv on capacity additions. But it's very regional, right? You have sort of Central Europe, you have India, you have North America. So how should we think about sort of your global supply chain? Should we think about it sort of these regional verticals, or should we think about your supply chain as global? Can you use, right, because different electrical standards globally, do you have access to your global supply chain if you're in North America, or is it really region for region?
It is multidimensional in its very nature. There are two aspects. One, well, when we announce things that are specific when, for example, we announced India, thermal, factory in Chakan, close to Pune, it's because we had that. But when we think about supply our supply chain, we think in terms of, a global supply chain, a global manufacturing footprint. Let me separate supply chain like the stuff that we buy from the stuff, like the input components or material, raw material, and our manufacturing capacity. We increasingly think globally in terms of accessing the capacity and serving, globally. When we do that, though, we are acutely aware of, the importance of being, serving a market in the proximity of the market at the same stage. That is a matter of resilience. That is a matter of, shorter supply chain.
So, we balance the two facets. We have capacity. When we look at capacity, we look at capacity globally and region for region. So, don't be misled by maybe a regional announcement or, for example, when we are communicating about liquid cooling capacity or we were communicating about switchgear and busbar busways capacity, we were talking always in terms of global capacity. Because there is certainly a global manufacturing footprint that is being utilized. But again, a lot of attention to serving, for example, the North American market and having the right capacity in North America to serve the North American market. The same is true for the supply chain like raw material and inputs. Always think global, but also always think in terms of building redundancy within the supply chain. And we're becoming very scientific about that.
Specifically, maybe we can just sort of drill down into North American supply chain. You know, clearly you have made efforts over the last three years around adding suppliers and having alternative components. So where are we now, and what level of supply to commitments are you seeing now versus a year ago?
So the efforts that you mentioned here is certainly what we've done for North America. But this is our approach globally. Already at design stage of new product launch, we make sure that we have redundancy built in the supply chain and multiple suppliers, for you know in terms of raw material and raw components. And at the same time, we have developed tools to make sure that that kind of intrinsic resilience is maintained. As a consequence of that, we relative to a year ago and certainly even more so a year and a half ago, the stability in our supply chain is a totally different level. Is perfect? It's never perfect.
But we are satisfied in terms of the ability to have a very stable output situation. Then, you know, when you have new products, new product lines, there is always the kind of a warming up, the supply chains, et cetera. But absolutely within the realm of business as usual, we are so much more resilient than we were in the past.
Gotcha. So, maybe we can sort of go to liquid cooling, which I know is sort of, you know, maybe a couple of people.
Yeah, people like liquid cooling a lot. We do too.
Yeah. So, you know, when you have a transition like with liquid cooling, that's an opportunity also for new entrants to offer point solutions like, for example, coolant distribution units. So what would be if I were to go against Vertiv, what would I pitch? And, what's your, what's your competitive advantage, right? What keeps these new entrants out, and what keeps the current industry structure? Because the industry structure, to be fair, right, it's, you know, if you include services, it's you, it's Schneider, it's Eaton, it's Siemens, Legrand. And then there is half the market, right, if you take the more expansive.
Yeah. One should probably separate power and thermal to slightly different groups of competitors. But yeah.
You know, but yeah, so how should because we're getting a lot of questions. That's like if there is one question, we get a lot of people call us and like, what about entrant X and what does it do to ecosystem and how does Vertiv keeps these folks out or how does Vertiv work with them? We can probably spend an hour just talking about that.
Yeah, exactly. But so I will talk about what we believe are value proposition and at the same time, competitive advantages. So first of all, you have to have the technology. And the technology cannot just be a CDU. It has to be something that has been developed with a chip manufacturers and proven and approved. And it's very important that you are, and that's true for Vertiv. We are working with the biggest chip manufacturers to really match our product roadmap, liquid cooling roadmaps with their chip roadmaps as the complexity and the heat density is going to change. We see today, finally, the transition going to happen from air to liquid, as for example, Blackwell will come out and that will be de facto liquid cooling only.
So that acceleration is going to happen. So first of all, technology and the technology partnership with the big chip manufacturers is absolutely fundamental for the long-term success and the guarantee for performance. But then I would say that having the right product is just a portion of the total equation. The other element, very important, is the ability to scale capacity. Not all players and certainly the small and kind of new entrants may not have that ability. It is about having the relationship with the customers. So not only the chip manufacturers, but then the de facto end users and have their trust in designing an infrastructure that they can scale. An element that is extraordinarily important, though, is the ability to service. We like to talk in terms of thermal inertia.
The thermal inertia is how forgiving the technology is should something go wrong. In the traditional technology of air cooled, you have a failure in something in the, let's say, cooling system in a data hall. If you plot the temperature in the data room, will increase very gradually. Now think instead about the heat density inside a rack. And think something goes wrong with a CDU, coolant distribution unit. At that stage, you really have a brain and the blood vessels that go through the brain. An interruption of 15 seconds, as we know, same is true for the brain, is 15 seconds too many. So clearly the systems have the ability to shut down, et cetera.
But it can be quite abrupt and certainly can be pretty damaging. So the critical nature of the application, if anything, is one order of magnitude bigger. So in that situation, it's very important to have someone who can have the right remote monitoring, telemetry, condition-based or predictive maintenance capabilities and then the ability to deploy resources to fix the problem. This is a very important competitive advantage. And one that takes years to build. Because you need to have a global presence of service and have the right technology behind that.
Right. And historically, just to sort of, dive into a little bit more into this, Vertiv has not sold products inside the rack. But with liquid cooling, the cold plate is on the server. So how will this change Vertiv's relationship with chip manufacturers and server builders? And what have they been like in the past? And where are they going in the next few years? And maybe specifically, obviously, NVIDIA highlighted you as one of their partners. Maybe as a follow-up, we can talk about that. But just generally, I think just sort of very philosophical questions that goes to, you know, your heart of your business model.
You really refer back to my kind of extensive experience in the industry. And so if I go back to, let's say even 10 years ago, there was no such thing as a relationship with the chip manufacturers. There was something in a box. We knew the characteristic of what came out of the box, a server, and probably the characteristic of what comes out of a rack. And that was it. But again, the things have changed dramatically. If you do not have the relationship with the chip manufacturers, you do not know where the technology is going.
Right.
You do not know well, you can know you may know the technology today. You may know the technology is being deployed tomorrow. But it's the long-term multiple years that will definitely make a difference. So things have changed dramatically in that respect. And it's good to have a strong partnership in place.
What, what should we make? Because NVIDIA put out an announcement, and then the day after, you guys put out an announcement, and they sort of said that you're part of their cooling solution. What does it mean for the next iteration of, you know, Blackwell chip? What is the relationship between you and NVIDIA?
Well, you know, there are press releases. But I'm just trying to understand. It's a multi-faceted relationship. Of course, there are confidential elements to the relationship. But it is, first of all, and very obviously, there is a partnership within their ecosystem. And that's very much a go-to-market. There is a big technology and relationship at an engineering level. So our R&D people and their R&D people work together, talk together. We are involved in common projects, even, let's say, ARPA-funded projects. But, you know, we work with them a lot.
And again, it's aligning their roadmaps and our roadmaps, you know, certainly not the only one, but certainly an important enabler of what they will do in the future.
Gotcha. Can anybody be a preferred supplier to NVIDIA? Or is it a fairly small number of players?
Well, if I were NVIDIA.
I have not seen a lot of press releases.
Yeah. If I were NVIDIA, of course, I would make sure that there is more than one for obvious reasons. So yeah, they need an ecosystem.
But our sense is the ecosystem is based on working with the industry leaders. That the foundation of ecosystem is that they went to the people who are the leaders of the market.
Definitely.
That's who they partner up with going forward.
Definitely. That's an important, I think, prerequisite to be an important partner for them and for others. Absolutely.
Just a little bit to digress a little bit, you know, I think several years ago, there was a lot of talk about immersion cooling. That sort of went away. Why?
Well, it hasn't gone away totally.
Yeah.
I believe that direct chip is more natural in terms of the way you operate a data center. That's why you see the bulk of the market going in that direction. We believe, and we have, immersion cooling technology in our portfolio, that that is there to stay.
Right.
Though more of a niche play than a mainstream direct chip.
It's because direct chip basically allows you to preserve the basic architecture inside the data center. Is that?
Yeah. That's what we see. And we believe also you can better utilization of a data center footprint.
We've been seeing just to finish, and then we'll go to sort of margin. We've been seeing sort of data centers on high density sort of standardized around 60-70 kW. Is that the right number to think about?
For now.
OK.
Very much for now.
Oh, for now?
Very much for now.
OK.
We see that going past 100. It will go up.
That's good to know. Thank you. Like maybe in the remaining time, talk about margin a little bit. Pure EBITDA margin saying the low-to-mid-20s. Vertiv's 2026 target for 20% adjusted. Operating margin suggests 21% EBITDA margin. Is there a reason why data center products should have a lower margin than electrical equipment for residential or commercial construction?
Yeah. The short answer is we don't think so. I mean, we provided a framework in our investor conference as it relates to our expectations for margin growth over the next five years. We were at 15% for full year 2023. We project that to grow to 20%+. And that plus after the 20 is important because we don't see 20% as a ceiling as it relates to our potential. So when you compare us to competitors where they are today versus where we expect to be in the future, I would use that 20%+ as hopefully something that provides some upside potential. So, you know, we in the investor conference, we you know certainly put the different buckets of opportunity out there.
Those buckets of opportunity scale well beyond that 20-20% + that we targeted.
Excellent. That's a good answer. Within Vertiv, the margins for data center sales above the margins for telecom and industrial sales?
They are. Yeah. Yeah. So, of course, the dynamics for the data centers today are fairly strong from a margin perspective. Strong demand and limited supply. That's very helpful from a margin and pricing perspective. You know, telecom has, you know, had some challenges with demand over the last couple of years.
Yep.
You know, C&I, I would say, is an area that we see long-run opportunity and probably long-run margin growth there. But from a, you know, market vertical perspective, data center certainly is ahead of the other two.
Good. So last quarter, you said that AI-specific products like coolant distribution units would have similar margins to existing thermal products. But more broadly, do you see any mixed benefit or drag from a new build AI data center versus a traditional data center?
I would say probably premature. We continue to confirm the linearity in terms of AI, non-AI. What we see, we like going in the right direction, but premature to state anything different.
Maybe in the remaining time, just talk about the industry capacity. I think one of your competitors, right, because if we do basic analysis and, you know, I think everybody has done it, if you take sort of the power generated by CPUs and then you overlay the power generated by GPUs, and then you do content per megawatt, you're coming up with a bigger number. I think one of your competitors pointed out that in an unconstrained environment, that's a great way of thinking about where the revenue is going. But having said that, we are in a constrained environment. So can you talk about what are the industry constraints, right, when you talk to the customers? What are the key constraints? What does it mean for the industry evolution over the next 2-3 years?
Sure. It here is very much aligned with what we said in several occasions. So there is certainly a power availability constraint. Now, the smart operators, be they hyperscalers or colo, and the successful ones, the ones that are able to navigate the situation, permitting, depending on the country and jurisdiction, can be a drag. Let's not forget that building a data center, it's a big construction site. So it doesn't happen wielding a magic wand. So it takes time. But again, when we think about the smart operators and the very successful players in the industry, they run at their pace. But in general, that is true. For example, we're talking about Europe, specifically earlier this morning.
There you see a very kind of a regulated environment, with the usual suspect locations maybe not so eager to grow to allow the industry to grow. So there are a number of constraints that, if you will, I wouldn't say a limit to the hunger for capacity. But it's just very normal. It's very normal in every industry. From a power perspective, we see a lot of things happening. So I don't know if 2, 3 years. But eventually, things will accelerate.
And just maybe on Europe and reminding, you know, it's been pointed out to me that what also is happening in Europe, that they are taking a lot of very energy-intensive industries offline. And also, Europe has a different design for the grid. So ironically, Europe might have more marginal capacity available. How does that play out versus the regulatory environment there? As you said, some people may not be eager to have that kind of industry come in.
We talk about Europe specifically. So North America, we have different dynamics, for example, in the Middle East. I would still say that it's a little bit early for Europe. That AI build-out wave has not really hit the continent. We see some green shoots. We see some pipeline acceleration. But nothing comparable to what we see in North America. So I think it's a little bit early.
I think we're out of time. With that, I'm going to thank you. Thank you for being here.
Andrew, thanks. Thanks a lot for your time. Thanks, everyone.