QTS On the Data Center Outlook

Speaker 1:
Welcome to TD Cowen Insights, a space that brings leading thinkers together to share insights and ideas shaping the world around us. Join us as we converse with the top minds who are influencing our global sectors.
Michael Ellis:
My name is Michael Ellis, the communications infrastructure Analyst here at TD Cowen, and we are at our 11th annual Communications Infrastructure Summit. Today's a very special day. I am joined by Tag Greason, who is the co-CEO of QTS.
Tag Greason:
Awesome.
Michael Ellis:
Thank you very much for being here with us. Really appreciate it.
Tag Greason:
Mike, thanks for having me, and congratulations on the 11th Annual. Quite an event. Very well known, and congratulations.
Michael Ellis:
Thank you very much. Well, this is a special day for me, because the first company that I ever did deep research into was QTS.
Tag Greason:
All right.
Michael Ellis:
So now I've seen-
Tag Greason:
Does that make me nervous or does that make me happy? We'll see, right?
Michael Ellis:
This'll be good. So what I'd like to do is there are many people out here who have followed you from the public markets, just like I have. Walk me through the evolution of the business. How's it changed, how's it scaled? Whatever you think may be relevant for this conversation for folks to-
Tag Greason:
Yeah, Mike. Well, thanks again for having me. It's been a journey. I've been at QTS for 15 years. I was employee number 57, and had the privilege to work alongside the leadership team as we were a private company, then went public, we're eight years in the public market where you covered us, thank you very much, and then taken private again by Blackstone in September of '21. And that journey has brought lots of challenges and lots of opportunities, but I would say on balance, lots of opportunities, and the take private has really allowed us to expand what we thought was already a great expanding, growing business exponentially. When the take private happened, and Blackstone took us private in September of '21, we have grown the company tenfold since then in a four-year period. And I'll tell you, it's been an amazing journey.
Michael Ellis:
Now, let me ask you, if we look at the next five years for QTS, what does the roadmap look like? Is it another tenfold increase? Because it seems like the demand is there.
Tag Greason:
Yeah.
Michael Ellis:
What does the roadmap look like?
Tag Greason:
The demand probably is there. You and I could have a debate about the demand. I've been in the data center business for 25 years. I've never seen anything like we're having right now. So I wouldn't suggest that this hockey stick that we're on will continue forever. But if you put the demand aside, you can look at a number of other factors that will influence the growth of the market, the growth of the industry, and the growth of QTS. If we think about our business, we spend time on land, power, supply chain, and labor. And if we can answer the question of each of those four categories, we believe we will continue to be successful. And so you've got challenges in land around community engagement, and sometimes there's some negative press out there about a data center next to a neighborhood. So we're very, very conscious of entering a community in a very gracious way and spending time with that leadership team.
On the power side, you know the challenges as well as I. We can talk about traditional power, we can talk about innovations in natural gas and combined cycle, we could talk about nuclear, but power is going to be part of that story over the next five years. Supply chain, as lead times continue to extend, we've got to be more thoughtful about supply chain. And then of course, labor. Having the right labor and not having too much density in a specific region, which will dilute the labor pool, is very important to us. So when we think about the next five years, we think about it tactically in answering those questions.
Michael Ellis:
That's fantastic, and we'll dig deeper into each of those points in terms of the power and the supply chain. But I do want to touch on the demand. You said it's the best that you've ever seen, and I'd completely agree. I've never seen anything like this. But if we were to talk about, let's say the complexion of the demand, put aside the number of megawatts, what are you seeing in terms of how the market is evolving and what the customers are looking for? Is it more focused on the major metros? Is it more on the remote side as we think about training data centers? Where are you seeing the demand?
Tag Greason:
It's a great question. Every day we read something either from TD Cowen or otherwise. There are other sources that I go to every once in a while, but I will tell you, you can't listen to the TV, you can't tune into a podcast, you can't read an article without hearing the words AI. But I want to remind people that AI is additive to a cloud services business that was already growing pretty significantly. So when I think of the hyperscalers, I think of 20, 25, 30% growth on cloud services alone, and AI is on top of that, which is really exciting. So as you think about that exponential growth in AI and the complexion of what that workload is, you're still going to have cloud services and they're still going to be tethered to an availability zone.
You've got to be in the right location, you got to have the right power, the right timing to intersect with a hyperscaler, and then you'll be successful in a leasing model. But the AI workload is different. It can be detached from the availability zone. It can actually be in areas that are not traditional data center areas. And that's really exciting for the market. It's really exciting for QTS. And as long as that demand of cloud services and AI continues to grow, I think the next several years in the market will be fantastic.
Michael Ellis:
Now, let's take that and let's disaggregate it a bit, because I think we talk about sometimes AI like it's a monolith. The way I think about it is, all right, you have the training which can, to your point, be disaggregated from the availability zone, and then you have the inference, which I do believe needs to sit proximate to the cloud storage within the availability zone. So to that point, as you think about the training side, which I would say is the precursor, where do you think we are in that journey? If we're going to use a baseball analogy, are we in the third inning? Are we the first inning, the seventh inning? Where do you think we are?
Tag Greason:
It's a great question. It's the question. It's hard to pin down, because training is evolving right before our eyes. When you first think of training, you think of one single monolithic computer system growing together at 300, 400, 500 megawatts, and you're thinking about latency within that monolith. It's this huge single mainframe, which I always find to be fascinating that we're talking about coming back to the beginning of this compute engine. But there are now training instances that are private training, that are unique, that are smaller, that are customized for individuals. And so if I think about training as we understand it today, which is one large monolithic computer all working together, then the location of those and the disaggregation of those from the availability zones will continue.
But as training evolves, you'll start to see it pop up in availability zones, you'll see it pop up in private instances. I think training has a long journey ahead. So training in general, first couple innings, but it'll be interesting to see, Mike, when we talk at the 12th Annual, what will happen with training. And I think we should come back to it, because I think 12 months from now, we may have a different answer.
Michael Ellis:
What's interesting is that as I think about training in moving data centers into, let's call it more and more markets or non-traditional markets as an example, we know what the value of a data center in Northern Virginia is. It's established. Same thing with Dallas. As I think of some of these newer markets, it's a new proposition. How do you think about going into perhaps new markets and pursued, or in support of that opportunity for the hyperscalers? Do you find it compelling? Just curious how you think about it overall.
Tag Greason:
Yeah, we find it very compelling, but we also find it coming with a set of new disciplines that are important for the business. For example, we are very focused on the top hyperscalers, the best credit in the world, those that will be there long before your and my career are ended. We think about long-term leases, 15, 20-year leases. We think about 100% confirmation of that rent stream, no out clauses. So we are approaching this market with excitement, enthusiasm, but also with some discipline to make sure that we're not setting up either QTS or the customer for something that will be relatively devastating. You go too long on a project that doesn't last, that's a problem for everybody involved. And so we mix this enthusiasm with discipline, and we stay front-footed on that opportunity.
Michael Ellis:
Perfect. I want to start to pivot to enterprise, but before we get off the AI thematic, I do want to talk a little bit about inference. When I think about the commercial returns for inference, I do think about it being with enterprises. Those are the ones who are going to be using this and paying for it and driving the returns. From your perspective, are you seeing the hyperscalers roll out any, what I would say is enterprise-grade inference offerings to the extent where they're getting to a position where they can sell to the enterprise and we could start to generate maybe that return, which then could drive demand for data center providers?
Tag Greason:
Yeah, it's a great question, and I think they're all trying to figure that out, because I'm very close with all the hyperscalers. We have this conversation often. I'm interested to see how it, but it feels like it's a little bit of the early days of the internet. There's a little bit of spaghetti being thrown at the wall to see what's going to stick, and I don't blame them. I mean, I understand. Trying to monetize the capital investment is everything, right? The first race was who can get out and spend the most capital, deploy the most infrastructure, put the most dots on the map, deploy the most inferencing. The next race is who can monetize that capital investment.
I will tell you, Mike, I'm not smart enough to tell you which is the winner either by hyperscaler or by product, but I think there's a lot of demand. I'll give you a quick anecdote, very, very short. There's a very good friend of mine who is not technology-focused at all, and when she told me that she was using ChatGPT to plan her family vacation, it blew my mind. It was like, wait a second, this is as mainstream as you can possibly get.
Michael Ellis:
That's right.
Tag Greason:
This person barely looks at email and now they're focused on using AI in their daily work. I mean, it is really going to be a fascinating couple years as the hyperscalers determined the best way to monetize.
Michael Ellis:
Let's talk about the enterprise. When I remember QTS in the public market, I remember you had the breakout. Enterprise was, what, 66%, around 64%. Hyperscale was right below. Seems like things have changed, but I always remember QTS as enterprise was in the DNA. As you have scaled up the organization, I'm curious, how has your focus on the enterprise evolved? And as part of that, what are you seeing in that segments of the market?
Tag Greason:
It has evolved considerably. The good news for us though, as we think about our go-to-market and the diversification of our products, we still have a very vibrant hyperscale market that we just talked about, very vibrant enterprise and a very vibrant and growing federal TS space. When I think about the enterprise, to answer your question, the unit of growth, which is what we think of when we lease something, went from a cabinet to half a cabinet to 200 KW to 500 KW. The enterprise market now is two, four, six, eight, 10 megawatts, which was a hyperscale deal five or six years ago.
Michael Ellis:
That's right.
Tag Greason:
So the unit of growth has grown. We continue to see the demand. Our biggest challenge is keeping up with the inventory. What we end up doing is we lean forward and build a building. I don't like to use the word on spec, but we lean in to have inventory that's available for the enterprise and we have to be very disciplined that a hyperscaler doesn't come in and say, "I'll take the whole thing."
Michael Ellis:
Yes.
Tag Greason:
So maintaining that discipline and keeping inventory available for that growing sector of our business is very important.
Michael Ellis:
And to that point, I think that's one of the dynamics I've seen in the market, which is that because the hyperscalers will lease two years out, by the time the enterprise is ready to look at capacity, there's nothing there.
Tag Greason:
Exactly. Yeah.
Michael Ellis:
So as part of that, would you say that you're focused on building enterprise buildings, kind of standalone, making sure that you have capacity to sell to them? Or is there an opportunity to actually bring them into the same facility in a multi-tenant environment with the hyperscaler who's running those big workloads?
Tag Greason:
Yeah. The technology would allow us to build the same product, which is great, and we standardize on that product. It helps us with our supply chain. It helps us with our operations on ongoing. I will tell you though, I can't find any hyperscalers that are willing to share. So when we build a building, it really is about just our operational discipline to say, "Hey, we're going to kind of ring-fence this 50 megawatt building, this 200 megawatts across the campus, have availability for the hyperscaler, for the enterprise." We want to continue to foster that business, A, diversification, but B, that's where the next hyperscale is coming from. So if we have 1,200 enterprise relationships and they're growing, who is the next one of those that will take a single building? Who's the next ones who's going to take a campus? And we see it. In our pipeline now, we have multiple enterprises that are starting to ask about single-building total leases, which is great. I wouldn't call them hyperscaler yet, but that's still a 30, 40, 50-megawatt lease with a single enterprise.
Michael Ellis:
Crazy how the deal sizes have increased, even on the enterprise side.
Tag Greason:
Yeah, amazing.
Michael Ellis:
As I think about QTS going private, one of the things I think about is the capacity delivery engine of QTS has really accelerated, and as part of that, you have, scaling up like that, it's funny, you'll almost become a hyperscaler in a way, right?
Tag Greason:
Yeah. Yeah.
Michael Ellis:
As you think about delivering the incremental capacity, what would you say are the major bottlenecks in terms of delivering the capacity and bringing that online?
Tag Greason:
As we deliver the capacity, so we've answered the questions of land power, we've got supply chain and we have labor, but supply chain and labor are dramatically impacting their ability to deliver. So what we have seen over the years is supply chain has gotten better and more predictable, but the quality of the supply chain is not at the standard that we would like it. For example, we receive pieces of equipment and we go through an entire process of Q&A, wait a second, we got to change this configuration. We got to make this tweak. We've got to get this component that didn't come and wasn't shipped. I think the supply chain itself is still strained, sending out as much equipment as possible. Many, many times, we see we get a piece of equipment, oops, we didn't count on that two extra weeks to now have to rejigger it before we install it, so we're building in some of those cushions.
Second, we're looking at multiple suppliers for the same piece of equipment. So we have an A and a B. Redundancy is part of our DNA, as you know in this industry. It's redundancy in the supply chain as well. And then the last part of the operational or the delivery part is that labor force. If you're building a hyperscale campus next to a hyperscale campus, you're going to struggle to get electricians. You're going to struggle to get the workforce. And so having the diversity of that workforce and the diversity of location is something that we look at a lot when we think about projects and how we move forward. But thinking and leaning into supply chain and labor allows us to then deliver at the requirements we've made with our customers.
Michael Ellis:
One of the dynamics that I find interesting is on the power side, we talk about power constraints and it's all over the news in terms of, hey, we need more power. But for all of that talk about power, I see a ton of people at these conferences saying, "Hey, I have 200, 300 megawatts. Would you like to buy it?"
Tag Greason:
Yeah.
Michael Ellis:
I'm curious, how do you reconcile the two? Is it that there is power that's available and we just have to find those pockets? Or is it that they may not be as close to the power as we think?
Tag Greason:
Yeah, it's a great question. I went to a conference in Europe, and everybody talks about Europe. I'll come back to the US to answer your question.
Michael Ellis:
Okay.
Tag Greason:
But I was in Europe, and I must have had eight or nine people say to me, "Hi, I have a gigawatt of power. Would you like to engage with me?" And I'm thinking, as I walked away, I was like, well, apparently there's eight gigawatts of power in Europe that is just sitting there waiting for people. What I find though is when you dig a little bit deeper and you go a little bit into this process, you realize they don't really have it secured. There's a lot of people who are, I've submitted an application, I have a request. I'm waiting for feedback. A little bit of poetic license in leaning forward. On the US side, I see a similar dynamic, but the relationships we have with the power companies and our track record allows us to quickly cut through, is this real power or is it not real power?
And when we sign up for a minimum demand charge or we agree to pay for transmission or we agree to build the substation, those are real dollars that solidify those opportunities, and that's how we get comfortable with the opportunities that are in front of us. Mike, there is one other piece of this equation that I think is fascinating. If a hyperscaler comes out to the market and says, "We have a 500-megawatt requirement, they then approach QTS, Vantage, CoreSite, Digital, and all of us go to the power company and say, "I need 500 megawatts."
Michael Ellis:
The amplifier.
Tag Greason:
And a power company says, "I just got applications for two gigawatts," and they start to freak out. But the reality is, it's still all the same 500 megawatts. I don't know how to solve that problem, but it is a concern that they have to, by law, take all of our applications seriously. They can't just say, "Oh, QTS, yeah, you're not for real." They have to take it seriously. But it's over-emphasizing that we really have a power constraint, but I don't know exactly how constrained we are.
Michael Ellis:
Okay. That's a dynamic that I've highlighted to investors. I call it the amplifier effect. One requirement-
Tag Greason:
Yeah. I like that word, yeah.
Michael Ellis:
...turns into five, and then the utility gets overwhelmed. Yeah, to that point on the conversation with the utility, right, how has that involved? And as part of that, yeah, you mentioned the money that you would put up to essentially, I would say, show that you're committed to the upgrade. How is that conversation evolving? Are you seeing more things like clawbacks on the power? Are you seeing more mandatory minimum commit?
Tag Greason:
Yeah.
Michael Ellis:
What are you seeing when you're having the conversation with the utility?
Tag Greason:
Yeah, we're seeing all of those. We are seeing those, and I think it's a challenge for the power companies to try to weed out speculators, people who are buying land and sitting on power who aren't really developing or delivering. They're not the hyperscalers, they're not the well-known third-party leasing companies. They're literally just developers, and maybe even speculators. And so a speculator is not going to put up $250 million of minimum demand charges. They're not going to put up that capital, and so what I see is a tool that the power companies are using, different levers, clawbacks, minimum demand, et cetera, that are helping to weed out those that aren't really going to deliver over the long term.
Michael Ellis:
I just want to shift and talk a little bit about the pricing and the economics side of this. We've seen development yields rise, demand has outstripped supply. I'm just curious from your perspective, with all the demand that's on the market, do you think that there's more room to run in terms of the economics of these projects, or have we stabilized and we're in the place that we're going to be?
Tag Greason:
Yeah. I think that there is going to continue to be pressure, and what I mean by that is we're starting to see new entrants into the market that have never been at scale before, who are legitimately working on projects with OpenAI or Anthropic or CoreWeave or the hyperscalers. So you've got now a market that has long-term traditional players like QTS that are well-established with the right track record, but you also have new entrants who are hungry and are really, really active in the market, and that combination is putting pressure, I think on pricing. I talked to Microsoft or Meta or Google, I'll just say generically the hyperscalers, and they're saying, "Well, Tag, I have another opportunity for 500 megawatts, and their pricing is X. Can you get down to their pricing?" The answer for QTS is stay disciplined, stay focused on what we do well. Stay focused on what our return thresholds are. There's a lot of opportunity out there. We do not need to knee-jerk react to one price or another. We're willing to lose a deal and stay disciplined and focused on what we think is important for our business.
Michael Ellis:
Perfect. Now, just to round it out, you've seen multiple cycles in this industry. This is a very exciting moment right now, I would argue, a generational build-up. Based on the lessons that you've seen in your time in this space, if you had to give advice to the industry, what would you say?
Tag Greason:
Yeah, if I think back to the journey, I think about the dot-com boom. I was in the data center industry when DeliverPetFoodToYourHouse.com was an actual thing, and now it is a thing called Amazon, but ultimately, we went through that cycle. We went through Bitcoin where people were trying to get a hundred megawatts, and Tag, I'll be happy to pay you in Bitcoin if you'd accept that as a payment. We've gone through some upsides and downsides. What I will tell you is, the cycle we're in is a generational cycle. It's super exciting, and we're going to try to lean into this as much as possible, but not lean into the point when the cycle changes that we're exposed. We have to be super disciplined.
My advice to people who are leaning in who don't have the benefit of seeing these cycles should be really careful that when the music stops, they aren't so exposed that they've jeopardized their entire company. That's what we think about at QTS. That's what we talk about with our partners at Blackstone. Jeff Berson, my CFO and I talk about all the time, let's lean in, but let's make sure we understand we're listening carefully, because when the music stops, we want to have a seat.
Michael Ellis:
That makes sense. I could sit here and talk to you all day.
Tag Greason:
That's amazing.
Michael Ellis:
Thank you so much for being here with us. I really appreciate it.
Tag Greason:
Mike, thank you for the time.
Speaker 1:
Thanks for joining us. Stay tuned for the next episode of TD Cowen Insights.