Resuming here with Dave and Gio from Vertiv. Very happy to have these guys here.
Thank you for having us.
I think the most popular, most requested company at the conference.
Wow.
Surprise, surprise. I looked that up in a chatbot . That was-
Really?
Yeah. Yeah.
Oh, wow!
They told me that. Yeah,
Pressure.
So, maybe just talk about just start with the guidance you gave, the 9%-11% for the year, and how you see that breaking down between, you know, hyperscaler enterprise and then also thermal and power. Just kind of break down the different types of businesses that are driving that.
Yeah, well, first of all, thank you for being with us, and, always a great, great pleasure. So, our guidance is pretty much in line with, with a broader and longer-term guidance, and indication we gave, about the market growth that we expect, for, for the industry. We were talking about an 8 to 10 to 10 percent, 7 to 10 percent, in, our Investor Day at the end of, November. So our 2024 guidance is straight aligned for that, with that. When it comes to the various parts of the business, let's talk about the applications.
As many of you know, we operate in the data center space, telecom space, and commercial and industrial, with data center constituting in 2023 more or less 75% of our sales, so certainly a big portion. Half of that 75% is hyperscale and colocation. So cloud, colo, and hyperscale. When it comes to the expected growth of colo, cloud, and hyperscale, we're thinking in terms of 14%-17%, and that's pretty much what we are seeing happen. So it's a part of the business that is growing the fastest. We expect it, and what we see confirms that across the various technologies, be it power, thermal, service, or anyway, the white space part of our business.
So, I would say we gave an indication of market trends three months ago at Investor Day. What we see is pretty much aligned with that.
I guess, how do we reconcile the difference between, you know, the capital spending numbers from your customers that are up 40%, in some instances 50%, and that rate of growth that's more, you know, in the mid-teens from a hyperscale or colo perspective? What’s the… How do we walk that down to that level?
The spend, the capital that you're talking about, Steve, is across the entire infrastructure. A large portion of that infrastructure is the actual IT infrastructure, the servers, the GPUs, the CPUs, and that is, if we take about 100, is the total value of a data center, we think that we are probably around 20% in terms of type of infrastructure. So that acceleration is, we believe, in this moment, are happening more on the IT side of the equation, especially as the new technologies, the AI-related technology is applied in existing data centers, using existing technology.
Now, of course, as the technology evolves, as GPU becomes denser and denser, then you will transition to liquid cooling over time, and that will certainly help accelerating our industry.
So you're effectively saying that the percentage that they spent last year on your equipment is lower this year than the absolute number that they're spending. That's kind of the implication, right? If you were 20% of last year—their last year's budget, and it's up 40%, if you're not growing with that, you're gonna be... That's just, I'm not a math guy, but that's kind of the—
No, that-
Is that the implication?
No. I will say that, in this moment, a lot of the investment is done using existing infrastructure.
Okay.
So that's why that mix, that is true mix for a new build-
Yep
is skewed.
Yes
in this moment.
Got it. Got it. And then when you look at your orders, like for the fourth quarter, very strong, what was the split between kind of like an existing facility and new facilities, kind of a brand-new facility?
The majority of our sales, especially excluding service, of course, a lot of the service is on existing installed base, but the majority of the orders are for new build, for new installations. So, that's what-
Okay.
That's what we see.
Is there any reason why we can't take that order number and just annualize it? I mean, I understand there's seasonality when it comes to construction of the data centers, and when you deliver, there are certain lead times, but, like, orders, is there really seasonality from an orders perspective over time, and what drives that seasonality?
There are cycles, and cycles may be built in the way we see pipeline. So I wouldn't talk necessarily of seasonality, but it's just a cycle of things, how they happen. So we are—we already gave an indication of high teens, Q1, 2023 to Q1, 2024. So we continue to see, we continue to expect orders growth.
That is a sequential, though, step down from 4Q to 1Q, right?
That is correct. That is correct.
Okay. What, what's driving that? Is that, like you said, kind of an ebb and flow of the pipeline?
Just, just the shape of the pipeline. Just the shape of the pipeline-
Okay
at this moment.
You don't view that at all as an indicator of demand softening or anything like that?
We've been vocal at the last earnings call about the fact that we are happy about the pipelines and the dynamics that we see in the pipeline, so happy about that.
Okay. Can you talk geographically, what you're seeing? The order number in EMEA, implied by the backlog, was very strong, a lot stronger than we would have expected. What are you seeing, geographically, US, China-
Yeah
... EMEA?
I wouldn't overanalyze a single quarter as a, as an indicator of a necessarily long-term trajectory. But we see the market in the Americas strong. Effervescent, certainly, where the most relevant part of the AI acceleration is happening. EMEA is following, I think, with a 6, 9, 12 months lag. And so we're happy, of course, about our orders in EMEA. We think that the trajectory is good. Long term, we continue to see a U.S. North America leading from an AI acceleration to the rest of the world, but the rest of the world following.
Yeah, that was something interesting from the keynote yesterday, you know, that this is a real global, almost geopolitical arms race, if you will.
There are various components to that. There is that component. There is also the component that a lot of the AI is latency sensitive in the inference part. So you want to have your compute power close to the user.
Right. That makes a lot of sense. So if you look at the backlog that you guys have and kind of net out the revenue you're shipping out of that backlog this year, it would suggest that, you know, your book and ship is even down this year relative to last year, maybe flat to down. Is there anything in the book and ship that we should be worried about? And any major moving parts in the book and ship business that, you know, face a different kind of cycle? Obviously, there's telecom in there, but anything that you would point out in book and ship.
Yeah, no, we do not see necessarily a lower book and ship in the numbers. We guided to—and, without going into exact math of this, we can do it offline, but, we see the book and ship continuing to be normal to good. The majority of the acceleration is happening, as I was saying, in the colo cloud part of the market, and that's a part of the market that operates typically large data centers on a 9-15-month sales cycle. Actually, I should say, order to revenue cycle. So it's something that hitting now would deliver towards the back end of this year or starting to hit next year.
But there is a lot of the business that is service business, that is the more shorter cycle, the territory business that is on a shorter cycle, as I was saying, and hence contributing to a book and ship that we see consistent.
I think you've said in these order rates that the identifiable AI-related orders, I think you said $tens of millions a quarter ago. It's obviously higher than that now. Can you put that into perspective of, you know, out of the $2.3 billion you did in the fourth quarter, like, how should we think about, you know, the, the-
Sure
related to the new stuff?
And again, I don't want to just point attention to a single quarter, because there could be different dynamics, individual large orders, et cetera, so. But in general, we see that the acceleration is linked to an AI-driven acceleration, AI preparedness acceleration, related acceleration in our pipelines and eventually in our order books. So that's pretty much the dynamic that we see. When it comes to identifying what exactly is for AI and is not for AI, as we vocally said when we were at the Investor Day, but also at the last earnings calls, we caution everyone out there to be kind of a binary. Some of our products are unequivocally destined to AI because, for example, liquid cooling is one-to-one connected to GPUs, et cetera.
So we can make that distinction. But if you think about, UPS, if you think chillers, if you think room cooling, all stuff that can go in any application, and sometimes we do not even necessarily know upfront what type of application it is. And even when we say, "Okay, a new data center is being built, you know, with deliveries last quarter, this year or first or second quarter next year, and it's an AI data center," yeah, that could be the case, but most of the time, those data centers are designed hybrid, with loads that are GPU-based and CPU-based. What portion of that will be utilized for AI and what portion for more traditional loads is something that varies.
So we said, you know, we can tell—we could tell exactly what is unequivocally like liquid cooling, but the world is much broader. The application is more generic. So we like to say we see the AI-related acceleration in our backlog and in our order book, in our pipelines, and that's drive the accelerated growth relative to a 3-5%. That would be the rate of the underlying enterprise data center market.
What are you seeing in enterprise? Is there any pickup there? Are discussions happening? You know, again, yesterday JP Morgan talked about spending a little bit more on their on-prem.
Yeah.
Anything happening there as of now?
We expect, first of all, AI to start to spill over in the enterprise market, and that spillover is a natural consequence of the general AI. To what extent exactly that will happen in the cloud or on-prem, I think it's premature. We expect an acceleration, but we do not know exactly when that acceleration will happen and what shape it will have. So I think it's early stage. I think that some of the enterprise market is starting to flex muscles with GPUs, but there are still types of loads that work okay in the existing data centers with the existing technology.
Clearly, the thought process about what next is happening at enterprise level, but I think it's behind, certainly, the hyperscalers and large cloud.
So you've got this, like, U.S. hyperscale going at it right now, big time. You've got 9-12 months down the road, you know, more of a global picture that's coming on-
Yep.
-and buying that. And then, in some point in between, there's also kind of an enterprise pickup.
Yeah, exactly. Let's see how that enterprise will exactly happen. We do not know yet.
And then just to get telecom out of the way, it's really the only thing that's kind of at all negative here, I guess.
Yeah, it's certainly the least exciting part of our market.
Is there any signs of life there, any catalyst for a pickup there? It's a small part of the business, but-
It is a small part of the business. It is synergistic, very much in terms of technologies. So, but in this moment, we do not see kind of a new technology. 6G is still probably with a question mark use cases, so it will take time. Is there a scenario in which AI will require more bandwidth and maybe very, very close to point of use type of compute? Possibly. Again, too early, too early to say.
Are we at the point where, in the near term, for the next 9-12 months, there's any kind of rationale for an air pocket, where these guys kind of build, they build a lot, and they just say, "We're, you know, okay for now," in the next, you know, 9-12 months?
I'd be surprised.
Yeah.
If but, you know, we all, I think, have been surprised in, in some occasions, but I'd be surprised given, given what, what we see. Clearly there is a limiting factor to the growth of the industry, and the limiting factor is power availability, permitting, so, and the very time it takes, and the complexity of a site work that building a vaster network of data center requires. So all of speed-limiting factors that, if you will, avoid, that industry goes in a tailspin, kind of grows extremely fast and then falls from the height. So I don't think so, but, you know, we are always very circumspect in this in many respects.
We look at the pipeline, pipeline velocity, dynamics obsessively every time to understand, of course, and to align to understand the evolution of demand, to align our capacity to the upcoming demand, but also the very same, let's say, discipline will tell us if something is going a different direction.
What historically has that pipeline been as a multiple of orders in the past, you know, 3, 4 years ago?
Well, look, we,
You know, the funnel, if you will.
The funnel is... I don't want to go into these details. These are a lot of details, so I prefer to give these details for us. But there are dynamics that go in terms of the visibility of the market. We have a good visibility of the market. We always had a good visibility of the market, and I think now it's augmented by the fact that we work more closely with hyperscale and large colos. We like the pipeline that we see, and it's a pipeline that reinforces the growth expectation of the colo cloud and the rest, but specifically in the colo cloud. So we see projecting and evolving in the right direction.
Yeah. It's bigger today than it was four years ago?
It definitely is.
Yeah. Okay. Just getting into the technology and kinda how you guys play here, maybe for people who weren't at the Investor Day, talk about your content, the number per megawatt, type of content numbers and how that breaks down between, you know, some of the various products. Like, what's the biggest part of that?
Oh, yeah.
I think it's a 3, you know, 3, 3.5, something.
Exactly. Exactly. So, without going into the details of the individual products and the sub-technologies, but in general, we think the data center space we think about in terms of $2.5-$3 million per megawatt as the all-inclusive service, thermal, power, white space, everything. And we see that growing to $3-$3.5 million per megawatt going into high-density applications. So, there is some favorability there driven by content per megawatt, but also there is a net increase in megawatts required. And basically, density calls for more content simply because the white space is becoming more complicated. For example, liquid cooling is an addition to our cooling portfolio and to the cooling term.
As far as your capacity and ability to serve, you mentioned the constraints around power,
Power generation.
Power generation. Obviously, there's some T&D that's involved in that as well. You know, some components that have much longer lead times than your components. How are... You know, what are you able to serve if, if they came to you today and said they wanted, you know, 30% more product like that? I mean, what is your constraint in delivering?
Typically, people would not come and ask for 30% more, but because they know that they need to have also somewhere to put that stuff.
No, my point is if they actually had it, how quickly, you know, what type-
Right
... of increase would you be able to serve in a perfect world where-
Uh
power just came from nowhere?
Yeah, very hypothetical. We do not design our capacity on very hypothetical worlds, of course, because that would be a little bit scary for everyone here in the audience. But generally speaking, when we design capacity and when we think about our capacity and our supply chain capacity, we think with some 25-30% leeway, so that we can accelerate if peaks come. And it's nothing particularly special. Typically, you design a factory on two shifts, not on three shifts, 24/7, at least in our industry. And that gives us the ability to manage peaks of a business that is anyway lumpy in many respects, because it's characterized by very large projects. So now, when you have many large projects, of course, statistically, they level out.
But again, it's not necessarily always the case. So, in the short term, there is always an acceleration, and then that acceleration turns into investment, it turns into capacity. When we look at capacity, we typically have between one year and kind of a longer-term capacity, one year and two years, to get the capacity operating. But we can expand in a matter of three, four, five months.
Right. So the point is that you guys are not the constraint here. It's these, the more in the broader infrastructure that is really a constraint to faster growth, in this build-out.
The answer should be more specific, product by product, product categories by product category. But in general, very general terms, I think we have capacity to serve the industry, and to respond to the demand that we're given today.
When it comes to technology, obviously this next generation of GPUs, there's gonna be a greater use of liquid cooling. Maybe could you talk about your view there, how you guys play, and then maybe a little bit of what your take is on how that may evolve from an industry perspective, you know, from the actual user's perspective?
No, absolutely. So as I was saying, well, it's interesting, kind of a year ago, in this very conference, we were talking about liquid cooling, something you know that that was still a lab type of activity. Now it's very, very real, and we believe we have a very strong offering, and certainly, we are preparing capacity to serve the industry abundantly. So, the general description, I was saying the way we manage we manage investment and capacity, it needs to has an exception, and the exception is liquid cooling, where we are, if you will, overcapacitating, because we know that the technology is at its very infancy in many respects, and the adoption is very, very small yet, but accelerating.
We have evidence of this acceleration in the activity we do with the NVIDIAs, Intels of the world, as well as all the hyperscalers and the major customers. And this acceleration is gonna be rapid. We want to make sure that we have all the capacity the industry needs, and that's why some may have seen the slides from our investor day. We are almost multiplying the initial capacity that we exited 2023 with, after the acquisition of CoolTera, almost 40 times, which sounds enormous, but the starting point was not enormous. So just to put things into context. But we want to make sure that we have capacity in place to really serve the industry, to serve the industry very, very well and not to miss opportunity.
So that is a different cycle. That, that part of the technology will-
You said that you think that's going to be 20% of the served market by, like, later in the decade?
Uh, yes.
Like 5 years, 5 years from now?
Yes, five years from now.
Yeah.
How accurate that is, yeah, remains to be seen.
Right.
Again, 2023, the shipments of liquid cooling solutions were very, very modest across the industry. So, the technology is, you were asking technology, I kind of went off the tangent, apologies, but the technology is predominant. What we see predominantly is direct to chip cooling, as opposed to immersion cooling. Immersion cooling, we have, too. Certainly the acquisition of CoolTera strengthened our direct to chip portfolio. What we see and what we discuss with the big players, as I explained, is direct to chip. Now, single phase direct to chip liquid cooling, and eventually, when the power per GPU or chip will probably go past 1.5 kW, probably, let's say, two phase with evaporation taking place. Getting too technical, and I'll stop.
So, do you know... I mean, Dover was here yesterday. They're a very small player. They make connectors for liquid cooling, and they made, they commented that the kind of the next generation of GPU is kind of specced out for them, and that they have an idea that they're going to be, you know, a part of this. Do you have that kind of visibility as of now until, like, when that next generation comes out, like, what they're going to use?
Yeah. Yeah, we do. We do because we work with them.
Got it.
We work with our chip manufacturers.
So you're building this capacity that, you know, some of us would view this as the technology playing field is wide open, and it's a risk because they could go here, they could go there. You guys are kind of building this capacity, knowing already, at least in the intermediate term, what's coming down the line.
Yeah, we believe, and from what we see and the conversation we had, in the single-phase liquid cooling is the technology that will be dominant for specifically in the chip cooling technology, in rack cooling, that will be dominant for a few years.
Are there a lot of competitive products for that?
Uh, the-
Like, what's the landscape look like from-
It's a relatively new technology, as I was saying. So not, not accidentally, we, we went organically and through partnership with CoolTera before the acquisition. And as typically in every case, you have new technologies very early in the maturity curve. There are many small players, if you will, and then there is consolidation. I think we, we made a very good step in consolidation. Let's see what, what happens next in the industry.
Right. I know Trane. I don't think does have this technology. They're kind of a bigger player. I don't know about Stulz, but, like, I guess this is where you have an advantage, where you cut across the data centers. Eaton and Schneider would not have this advantage and have visibility into it. So you really are the player that has a significant potential to even expand your, you know, your content with something like this. We've always viewed liquid cooling as a bit of a risk. This seems like it's actually, in the near term, an opportunity beyond all this CapEx that's coming from it.
We see it an opportunity in two ways. First, as I explained a couple of times in the past, liquid cooling is additive to our TAM. In the past, that part of the cooling was done through a couple of fans on the, let's say, chassis of the server, and just spitting out heat that was then air-cooled. Now we go into the server. We like that. But also, we talk about thermal chain, so all the pieces, heat extraction from the server, all the way to... Maybe you can all the way to coordinate and orchestrating with the cooling, air cooling in the room, and then the heat rejection outside the building.
So there are many, many elements to this chain. The fact that we are on the chip and know what happens, gives us a chance to orchestrate the entire system in ways that not many competitors can do. So we like a lot of the place, and we like the fact that the complexity that you see in the cooling is also, in my respect, the complexity in power. I mean, if you take a rack now is about 10-15 kW. 10-15 kW of power is not dramatically a lot, but when that becomes 150 kW in the same space, there is a lot of current, there is a lot of power going in there.
Being able to orchestrate the cooling and the power and having control on two sides of that equation is very, very important.
When it comes to, just one more question, from me, and then I'll pass it off to the audience. But, you guys had, in the last year, you've had, you know, anywhere from 5%-7% price in orders. You're not, you're not really guiding price anymore, but you're, you're guiding to a $60 million price cost benefit. You've said there's about a $100 million of, of inflation, labor inflation. So just kind of like doing the math on that suggests that you're embedding, like, low single digit price this year. Is there any reason, are you, are you concerned that orders price goes negative at some point this year?
Generally, we do not have that concern. It's a good market environment for price. We saw that in the price we got in orders in the fourth quarter and really throughout 2023. I would say our assumption for pricing in 2024 is prudent. So you mentioned the amount of backlog we have heading into 2024.
Yep.
It's roughly comparable to what we had heading into 2023. We beat the price number last year, but there still is a lot of book and ship. Right? And there's still, you know, a relatively uncertain market environment. It is a competitive environment. We hope, and we're not gonna disclose the specific number. We did that because of the inflationary environment, but we hope we are able to beat what we've implied. But at the end of the day, there still is a lot of book and ship and uncertainty, I would say, more in the back half of the year.
Okay. Any questions in the audience here? Okay. I thought there would be maybe one or two - given how popular you guys are. I guess,
I can answer some tax questions, if that...
Are there-
No, no.
Are there tax questions?
No, there's no tax.
I'm not assuming there are tax questions. You guys had, you know, your free cash number was good. Down payments were pretty strong. That's gonna be down this year, but it's a, you know, it's a flow item, so that-
Yeah
... still means your, your contract liability is, or deferred revenue is growing. Is there, that's based on what you guys are assuming for orders, I assume, so there's upside to orders. The cash would come in better because you'd get, you know, there'd be more down payments?
Yeah. I guess just general on free cash flow, it's a critical part of our journey. I wouldn't call it a transformation, but really a re-energizing of the operations. So, you know, sales mean nothing if you don't generate the profit. Profit means nothing unless you generate the cash. So we actually burned cash in 2022, I think $260 million. With the operational turnaround and focusing on the processes, we were able to generate almost $270 million in cash. That's 115% of adjusted net income. We're guiding to, I think, just south of 95% in 2024. And we anticipate in the long run to be in that 95%-100% range.
So there's still a ton of opportunity. What really drove that, you know, number to be north of 100% in 2023, as you mentioned, was the acceleration of the upfront advanced payments from customers, in particular in the Americas. So that will continue, but we definitely got a one-time upfront benefit of that in 2023.
And then just on the buyback, you know, the stock... Maybe you were hoping the stock went lower. It did go down, but now it's kind of back up. I mean, is this you gonna be opportunistic on this? Is this gonna be programmatic? How are you guys thinking about buyback?
We'll be opportunistic. You know, we look at that as one lever to deploy capital, M&A, share repurchases, and the dividend. So we'll evaluate it as a portfolio and make, you know, decisions as we go forward. I think, you know, there are M&A opportunities out there. There is a pipeline, and I think in the long run, that would be our preference, but we'll continue to evaluate all three as we go forward.
And then just one last one for me. Gio, how, how do you see this playing out from a... I'm sure you look down the chain, and you monitor and evaluate who's actually using this, you know, the technology down at, like, our level. When, when do you think we see, you know, real powerful and scaled, kind of validated business models?
For-
Out of AI.
For AI. Well, I think we-
We already have some, but, you know, obviously not to the scale that would justify $1 trillion of investment, you know, that's getting thrown at this thing.
We heard this in this room from Jamie, probably 45 minutes ago, and it was around the fact that AI is in everything, and it's in everything already today, and it's gonna grow further. So it's not even about the use cases. The use cases are already there and are operating. So and it's just the beginning. So I think that that has started already. Now, the speed at which we'll grow is probably the speed at which the big infrastructure will be able to make AI available or a mix of the two. I think you know, we'll do our best to support, and we'll support to make sure that the world has this infrastructure.
We are very confident in the use case and the long-term nature of this demand, actually, world-changing nature of AI.
Well, what a difference a year or two makes. Thank you so much for coming, and
Thank you.
Best of luck.