So welcome, everyone. Thank you for joining us this afternoon. I'm Dave Barden. I head up U.S. and Canada Telecom and Communications Infrastructure Research for Bank of America. Thank you for joining us. I'm really, really pleased to have with me Jordan Sadler, head of IR for Digital Realty, and then this guy Andy Power, CEO, also important. And we're gonna talk a little bit about today the data center industry. So thank you guys so much for joining us. Really appreciate it.
Thanks for having us.
Thanks for having us too.
Really appreciate it.
So I guess I'd like to start at a big, big picture level with maybe Andy. You know, there, there's a lot of uncertainty about the political climate. There's a lot of uncertainty about the economic climate. There's a lot of uncertainty about the rates climate. What does all that mean for Digital Realty sitting here at the end of twenty twenty-four, thinking about how to plan for twenty twenty-five, and what should we think? How should we think you think about it?
A few things. One, we're a global company, and we're supporting five thousand customers on six continents, fifty-plus metropolitan areas. We're a massively capital-intensive company, so we're spending billions of dollars on new footprint infrastructure capacity. We are a real REIT, so we cannot retain our capital, so we're a capital sensitive company, which go in the mixing bowls of economics and interest rates can fall in. All those things are obviously risks to the business that we have to operate within and execute.
I'd say the fortunate piece of the equation is the demand side has proven over numerous economic cycles, at numerous vintages or places in the telecommunication technology landscape, from the dawn of the internet to mobile computing, to cloud computing, now to Gen AI, that there's just these secular tailwinds of growth that can outgrow the broader macroeconomic backdrop. And I believe we are yet again seeing that play out with our customer base, many of whom, including some who just were on their earnings calls less than twenty-four hours ago-
Mm.
- talking about, great results, revenue growth, I think quoting the word data center-
Thirty
... 30 times as an integral piece of infrastructure to the technology that they're building out. And that was just one example of many. If you looked at the landscape of the, call it hyperscale, providers over the last several weeks or months.
Mm.
So they're all towards... We are supporting things that are rising above the economic trends, be it digital transformation-
Mm-hmm.
cloud computing, and Gen AI.
Yeah, I, I think that that's a good kind of segmentation, so that there's three big secular forces, digital transformation. I think the question might be, if the economy slows down, could that slow down, or could it accelerate if the economy slows down? Because companies are gonna be looking to save money. Which of those two things you think is more important?
What's also unique about these components of demand is they're unique at different phases of their growth or maturation, but they're also very linked and coupled here, right? A digital transformation project for an enterprise customer, be it like Bank of America, is gonna have some place for data center and private workloads sitting inside of our four walls, probably in numerous parts around the world, moving off of on-prem locations or data centers they built and operated long time ago. It probably definitely has the use of many clouds, so numerous cloud computing growth interwoven in that, and I don't think anyone's thinking about digital transformation or cloud computing, and not also thinking about how Gen AI is gonna be supporting it down the road.
I don't see these things as isolations that you would stop and cut back on. These are all pieces of business for an enterprise as the end customer that are improving their top line and their bottom line in terms of efficiencies at the same time.
Yeah.
You kinda see it in the, call it, CIO or IT survey reports. They rank the prioritization of everything that IT does, from data centers, servers, to mobile devices and laptops.
Mm.
Right? And AI, data center, digital transformation are top of the list of our customers' priorities.
Yeah, I mean, I guess in the way I've said it is, you know, expansionary times, people are looking to gain revenue by going digital, and contractionary times, they're trying to save money by going digital, and they're trying to do it. I've learned, being here at Jeff Spector's Global Real Estate Conference with Sarah Cooper and everybody, that you guys are gonna jump in at some point randomly. So if you guys are gonna do that, just go ahead. This one over here, that guy. But so the second thing I wanted to ask was just about, you know, the next big driver.
It's become weirdly controversial, is the cloud, which is that from the birth of the cloud, the idea was maybe that: "Well, who needs a data center if everything can live in the cloud?" And then we kind of gave birth to this hybrid world where people want to keep a little bit to themselves, put a little bit in the cloud. But then more recently, in the last couple of years, we've had some regurgitation of these bare theories that the cloud is going to eat the data center marketplace. Could you kind of opine a little bit on where we live today?
Excuse me, the terminology cloud, given that we're in a very physical-oriented conference here.
Mm-hmm.
It was probably not a great description 'cause it gives someone the illusion that that means the data just floats out to the ether, and appears on our devices when it is happening in a physical infrastructure, in a data center, many of our 300 data centers. And I think you're right. There's been different vintages or views on cloud. Are we gonna have a one cloud world? Will AWS rule the world, right? And then move to... I think that view has kind of been shelved by even the biggest the proposers of cloud, saying multi-cloud, take the best of all the different cloud providers.
At the same time, private cloud and hybrid IT infrastructure inside customer-owned servers in a facility purpose-built like ours is part of the architecture as well. You've also had, like you mentioned, customers that were born in the cloud, pop out of the cloud due to efficiency and scaling, and some pop back into the cloud for some of their workloads at the same time. So I look at this as we at Digital are supporting 5,000 customers, so corporate enterprises like Bank of America to the hyperscale cloud customers, of which those customers we're supporting in 30, 40, 50, 60 different locations around the world. And we're a physical, trusted infrastructure partner for both those customers, trying to be that one-stop shop for their space, power, and connectivity.
I would layer on there, right? In terms of IT architecture, the argument is it's just not well-structured or thought out relative to what happens across, you know, IT landscape over time, right? So as Andy indicated, people are, you know, moving into the cloud, and some, you know, very significant portions of which actually the large majority of large enterprises today actually expect to pull workloads off of the cloud in the near future, right? Based on the same survey work that he was referencing. So, you know, we see people repatriating, you know, potentially at a bigger rate over time. Certainly, the cloud players would, you know, and these hyperscalers would like everybody moving into the cloud, and they will, and there are some great use cases, especially for different applications.
But when you look at overall architecture, there's lots of use cases that make the case for, "Hey, we should have some of our own compute.
Right. There's a proprietary. There's elements of proprietariness, there's regulatory reasons. There's also the nature of the cloud, which is variable cost for an institution. And so if you're going to be using things maybe in a bursty way, it's very helpful, but, you know, in terms of regular compute load, there might be owners' economics to having your own facilities in a data center environment. So just before we move on from that kind of conversation, you mentioned you have 5,000 customers, but you've got some incredibly large, important customers, and some of these large and important customers are the subject of certain governmental scrutiny, if you will. I'm speaking specifically of TikTok and what could happen if they disappeared. How do we, as a group, think that you think about that?
So we're obviously very confidential on our customers unless they give us their blessing to use them as an advertisement of the great quality service we're giving day in, day out. So I'm not gonna comment on any specific customer name, but you can imagine, as a global company, that's been in business for now north of twenty years, we come into situations like this and have to think about, we're making a substantial investment in each of these markets, that long-lived infrastructure, and the counterparty risk. This is another way of saying that. Will that counterparty disappear for whatever reason? Could be the things you illustrated there, could be credit risk or bankruptcy on the other bookend.
So I, my handicapping of the facts and circumstances, and I have no inside baseball on, call it the geopolitical, or regulatory framework we have here, I would say any customer that is with us in 10, 20, 30, 40, these big customers, hyperscaler you talked about, they didn't start in 2023 or 2024 to be our customer, when we had this massive inflection in the rates for our product. So they very, very likely have contracts in place with us that are, very attractive to them, less attractive of what we would have signed if they would have showed up at our doorstep this year.
Most draconian scenarios you could possibly imagine, a vacancy in our portfolio caused by draconian events would obviously be potentially post the downtime from re-leasing very economic windfalls to our bottom line, given the rates that those type of customers pay on their installed base to what we're signing in the market today. I'm not, I know, I'm definitely not rooting for that for any customer, and I don't think in my wildest dreams has that all kind of showing up at our doorstep, like that. I believe there's a long road of jousting and regulatory things that could happen.
I also would say, unlike our B2B customers, Azure, AWS, Google Cloud, Oracle Cloud type of customer, that has locational sensitivity and sovereign sensitivity, GDPR, they're in certain of our European countries, for European data. B2C type customers, think using their apps as just consumer, do not have that. Many of the B2C type customers put a substantial amount of infrastructure in the United States, but have most of their users outside the United States, right? Banning a B2C customer per se doesn't necessarily mean that infrastructure leaves our shore 100% at the same time. That's a long-winded way of saying I think via our diversification below market rents and I think a pretty low likelihood that that benefit shows up back to us.
I don't think I'll be reporting that out anytime soon.
Let's talk a little bit about how business is doing. First quarter, you know, record leasing, I think you said, and this is the third leg of the kind of secular demand stool. Roughly half was AI related. We've gone 14 minutes, and we haven't mentioned AI yet, so I might as well go ahead and bring it up. Talk to us about, you know, the opportunity that Digital Realty faces, and why Digital Realty is positioned to benefit from it.
So just to recap your accolades, which I appreciate, record first quarter, record first half. First half of this year, double the pace of the prior year. We did call out a 50% contribution for AI in the first quarter, 25% in the second quarter. It's important to understand how we are attacking the opportunity. We are not chasing this opportunity to unproven markets just 'cause a customer asked us to build them a data center. We're sticking to our 50-plus metropolitan areas where we see robust and diverse customer demand that is not just AI training model demand. Numerous cloud computing companies, compute, network, enterprise demand, markets that we truly believe in the locational or latency sensitivity of the applications living in those markets, like in Northern Virginia, Frankfurt, Singapore, et cetera.
And could you just talk a little bit about... You've said something interesting about how demand is actually concentrating as opposed to diversifying?
When it comes to AI, which AI is also a multifaceted piece of demand, we're seeing the preponderance of large capacity blocks from the hyperscale cloud companies wanting three things. One, they want large, contiguous capacity blocks, big and together. Two, they want it right now, like they're running to the bodega to pick up something they forgot.
A megawatt of capacity.
Yeah, exactly. Or 50 or 100 megawatts.
Exactly.
And three, their preference is fungible markets or markets I'm referring to, where if they get their AI demands wrong at that very minute, they can put their cloud compute in that same very data center. The third leg of the stool is a requirement, not a must-have, hence, in many of these core markets, they can't get that third leg because the core markets have had the greatest supply constraints, so it's spilling over to some second-tier markets as well.
To the power?
Yeah, due to power transmission, generation, other trends like that. So we're doing that on the bigger, larger capacity blocks. I think on the first quarter record, we announced a big supporting Oracle on a big GPU cluster for their enterprise AI cloud customers. We've also been supporting on the more enterprise piece of AI. We had a win with the Novo Nordisk Foundation, where we and NVIDIA are building the largest supercomputer in the Nordics. That is not a fifty-megawatt type of deployment, more enterprise-oriented use cases. And I think both of those segments are growing, but the big deals are probably taking up the most spotlight. And I think we're still getting started.
Training, in my opinion, is not done. Two, when we know the next leg of this is the inference, think of the users of the applications with AI, devices, be it people or technology, querying the models, which we understand, one, be a multiple of size of the addressable market that we're already experiencing in a large and fast-moving training market, and could come in capacity blocks that do not need that, have that contiguous requirement, and we hope, we don't know definitively, that that lends itself back to our core market and campuses-
Mm-hmm.
where we have that infrastructure, where we're hosting cloud computing and private AI or private compute for our enterprise customers.
I think that the question then is, first, there's been so much money put into the data center industry that we've all heard about, from the private guys, from others. Our team estimates that the top five, just the top five hyperscale guys, are gonna spend $240 billion in CapEx this year, and $280 billion in CapEx next year. And you're gonna spend how much?
Just a couple billion.
So it's not like, it's not like you need all the market share, right? You know, there, there's a lot of runway that you've built, and it kind of just happens that you were building a runway for digital transformation, then you added cloud to that, and you put land, and you put power, and you put, you know, capabilities and common sense together.
And it so happens that now, when the hyperscale guys hit the panic button, go to the bodega, you're the bodega, right? So-
We were supporting them for their traditional needs in many, many markets for a long time. We were pre-positioning when we saw a few things going back several years. One, we saw this business going global, right? We were a largely U.S. company, call it 10 years ago. Now we're across six continents, right? 30 metropolitan area or 50 metropolitan areas and 30 countries. Two, we saw everything was just starting to get bigger. Scale mattered, right? The data halls, the buildings, the campuses, the runways for growth. We had this word future-proof our customers' growth, and their future just kept getting bigger and brighter, and we need to be pre-positioned and ready for it.
And we also lastly saw that being in the full spectrum, from the enterprise to the hyperscaler, and everything in between, was value add to our platform because we're driving cloud consumption by our enterprise customers, cross-connecting to the public compute, and I imagine the AI workload will follow a similar virtuous cycle as well. So that's how we've thought about this. We've got north of three gigawatts of land on our balance sheet. That was underwritten and mostly acquired before they were talking about GPUs. So this is obviously well-positioned us to capture this well faster than we initially underwrote and pulled forward a lot of that capacity. And that's what we're doing right now.
So I guess I wanna ask my big question, which is that you had a record first half in terms of new leasing in the first half of twenty twenty-four. Can we beat it?
So this is, this I said-
Record a better half in the second half? Or you wanna know about a quarter?
Just give me the numbers. You know what I want.
As I said, unlike my typical disposition on the first quarter call in conjunction with the record, similar question may have been put to you. I was asked about, well, will there be another record? And I'd said, "Listen, you don't usually get. In this size of business we're talking about, you don't usually see records upon records that consecutively or that quickly, right? Because we don't build these. We build these in a modular fashion, right? We don't go speculative all these shelves or suites and right.
And so we're de-risking our capital outflows, and hence, we don't have shelves lined with product to sell out of the store and set records traditionally. That being said, and my commentary was, one, we're in an era where everything's in the triple XL size category is the most popular-
Mm.
Right? The biggest capacity blocks. Two, our shelves were not barren in that category. We've still had some very attractive, and I think that theme carried into the second quarter, where Dallas led the way in terms of our signings contribution. Not a record quarter, but a very respectably high quarter. And we still look back at markets like Northern Virginia and others, where we have large continuous capacity blocks, an opportunity to put up a record. And then I said, "I got three more shots, three more bites at the apple, so I won't rule it out." Now, that has changed. I got two more bites at the apple. I'm probably less convicted on that than I was with three shots at it, but I wouldn't rule it out still.
So-
First, Dave, you mentioned numbers, top five, $2 billion-$3 billion of our CapEx this year, $280 billion next year. Do you have a breakdown of how much of that is directed at data centers? That's the first question, and then-
The question was, of the hyperscale CapEx numbers that our internet guys put together, do we have a breakdown of the $240 billion-$280 billion this year and next year that's specifically related to data centers? We don't. We know that the vast majority of it is gonna be the chips and the servers and all the bits and pieces that go into the data centers, but they're absolutely self-provisioning a portion of this. We could probably double-click into that a little bit more and figure out their numbers.
But I think the message I was trying to communicate was, like, that when you think about Equinix and Digital Realty being the two largest data centers on the planet, you know, their spending is a small part of this huge opportunity, and there's a lot to go take advantage of it. And
I can maybe lend a little bit of a hand there.
Please do.
I think in our experience, we see fit out, so data center fit out by these hyperscale customers, right? You know, by the megawatt, right? We build for, we call it $10 million a megawatt, keep it round numbers. We see them investing in that same space at a rate of about $40 million-$50 million a megawatt in servers, racking, stacking, and cabling.
So-
For it to fit, so yeah, per megawatt.
our telecom and technology-
20% of that spend is on the data center.
Yeah, they're-
To you guys.
You're going in... He's saying the-
With that build cost, they actually pay us, you know, $1 million of rent.
CapEx.
Right. Right.
... absence, you know, the discussion of the record, whether you'll beat the first quarter or not, how would you characterize the nature of the conversations? You know, is the activity higher? Is it I mean, how is that trending, you know, in conversations?
Can you repeat the question?
Yeah, sure. I think Jordan's asking about how to characterize the conversations with the big customers in the current environment. What we're seeing is a continued urgency for these large capacity blocks with the nearest term delivery, with certainty around power, and it's from the same group that were the top buyers of cloud computing, but slightly expanded because some of those were doing more self-build themselves historically, less now. We're doing deals where they did shell deals and built inside the data centers themselves, less now. And there are some new names around the top of hyperscale today, like big buyers, as well. All in a backdrop where supply constraints due to power generation, transmission, substation components, switch gear, sustainability concerns, moratoriums, NIMBYism are all intersecting the market in many fashions.
So our value add being pre-positioned with all those attributes and be able to build, to operate, is extra appreciated right now on that backdrop.
Strong demand and strong pricing.
Correct.
And if I pick up on pricing, and you kind of alluded to it, you've seen this very big jump in hyperscale price, North America particularly. Kinda how do you see that playing across to your traditional colo business?
How do you see pricing across the marketplace?
So the hyperscale business has shown proven to be the most volatile in pricing. Part of that is just it went from earliest innings and matured as an asset class. That happened in a time period when interest rates were only going down, and there it happened broadly in markets that didn't have supply constraints, right? And so rates for a market like in Northern Virginia got probably pushed down to the seventies, and now call it popped up well north of one fifty, and we have the potential to be called printing close to two hundred in terms of rates, as an example. That phenomenon I don't see playing out in our enterprise colocation business as much. Hyperscale is called longer term contracts. The colo contract format is usually a shorter term contract.
There is, I'd say, a greater stickiness or less churn in the colo, and we've just had a more granular, more regular pricing power in terms of escalation, so as contracts. So you have less of dislocation or really massive supply constraints in the colocation market, like hyperscale.
So, we're gonna run out of time here really quick. So, let's kind of rewind a little bit and maybe close out that conversation. You know, a couple of years ago, one of the challenges that Digital Realty faced was negative releasing spreads, and that was because the customers that you had signed a decade earlier became these behemoths, and they came back, and they asked you for a lot better pricing at the tenth-year anniversary, and you had to give it to them. Are we starting a new cycle where it feels good today, but we might have a problem in the future?
I mean, there's always... I mean, the volatility in the business always has the potential to resurrect. I think the way we're pursuing our strategy around hyperscale, when you think about what we're doing today, we are obviously signing at better rates than we had in any prior year of recent times. We're locking in the longest contracts we've ever had, 15 years. We're also pushing on the escalations. I think last quarter, our biggest deal had a 3.5% rent bump. We could potentially do better than that in the coming quarters.
Is that domestic, 3.5%?
Yes. I think what you have is the overall market saturation in 15 years' time, the locational sensitivity. There's going to be fewer places for these customers to put build the cloud out. So I think that combination with also. There's been inflation in build costs, right? There's no question that the per megawatt costs we would quote were certainly single digits for many markets years ago. We've been now throwing around 10 as a more rough swag average today. These campuses are getting bigger. It's gonna take longer to build out, so the impact of inflation in the build costs could continue as well. So I don't see. I also believe the pain points on new capacity. Yes, they may get solved, but I don't see quick solves with permanence.
Yes, the southern line may come in and relieve power needs for-
NOVEC, Western Star. Yeah.
But they're gonna need to do a northern line after that, and that's gonna take a series of years, not months, and other markets are gonna, we're butting up against other colo uses, too.
Mm-hmm.
When Northern Virginia started, it was cornfields, right, and we were welcomed as an asset class. Now, we got to be very good neighbors, and I think Digital Realty stands out in that in terms of where we locate our data centers next to the airport, not next to the battlefield.
Mm-hmm.
and things like that. So, and I think this. I'm picking on one particular market, but of our 50 markets, the preponderance of these types of supply constraints and more thoughtful, long, elongated development timelines. I think these are features of the business that are going to be here for some time.
Yeah, I would just point out, so thank you for that. So I would point out that today, the Telco Tech, Industrials Group put out a report called Who Makes the Data Center? So if you want to know about who builds what that goes in the data center, that's a big part of it. And earlier this year, the Industrials Group, Andrew Obin and his team, put out something that talked about the grid demand, and that data centers are not the only thing that's putting a tax on the grid. It's EV, electric vehicles, and also the onshoring of manufacturing, which is a big deal.
I want to maybe wrap it up, Andy, by talking about something that kind of dovetails from our, the prior question, which was that, you know, a couple of years ago, it was negative releasing spreads. It was dilution from acquisitions. Last year, it was trying to fix the balance sheet in a rising interest rate environment. The FFO per share growth guidance for this year is 0%-1%. There's a target to get to mid-single digits, and I think an aspiration to get better than that. Walk us through how Digital Realty gets from where it is from now on a bottom-line growth, given all these great top-line things that are happening, to better bottom-line growth.
So we basically came out at the beginning of this year and said, "Next year, twenty twenty-five, net of headwinds from the deleveraging that took place over twenty twenty, this year, that mid-single digits is the goal, and thereafter, that is not the goal, that is the floor, and we want to do better than that." And our path to that is through obviously execution on the pricing lever, with our cash mark- to- markets on the install base, leasing up our vacancy, delivering on our capacity coming online, blocking and tackling.
Thereafter, post the headwinds from the deleveraging, we're really. We think there's a path where we're essentially taking this demand, this AI demand and cloud computing demand, and turning it into long 15-year contracts with three, three and a half, maybe even higher escalations, and building record backlogs to have a long runway of growth that we want to drive to the bottom line. And the M&A dilution is behind us, and the only thing that puts headwinds are things that we think are long-term goods, be it contributing to private capital sources, stabilize assets and attract evaluations, or how much development we share with partners. So we're not happy with five, even though that's the goal for next year.
We think there's better ahead, and we want to build a long runway of comp, consistently compounding the per-share bottom growth as the top priority for our company for several years in the future.
I think that's a great place to leave it. Thank you so much, Andy.
Thank you.
Appreciate it.
Good to see you.
Thank you, everybody. Appreciate it. Thank you.