Thank you all for joining us. My name is John Hudlick. I'm the Communications Infrastructure Analyst for UBS, and I'm joined today by Andy Power, the President and CEO of Digital Realty Trust. Andy, thanks for being here.
Thanks for having me.
We've got 30 minutes for Q&A, and I have a number of questions I'm going to go through here. We are also going to leave time at the end to take questions from the audience. If you have any questions, please raise your hand, and we'll get to you. Andy, starting sort of big picture, can you give us a quick overview of Digital Realty and why investors should invest in the company today?
Sure. Digital Realty supports 5,000 customers across 50 metropolitan areas on six continents, with their data center and connectivity infrastructure. We are essentially supporting three secular tailwinds of demand: digital transformation, cloud computing, and now artificial intelligence. We are the largest global provider, with north of 300 data centers, operating close to 3 GW of capacity, with another incremental close to 4 GW of growth or underdevelopment capacity under our control or on our balance sheet. We have been a 20-year public company, so we have been to this conference a few times. We were at the data center game well before folks were talking about GPUs and AI, and we are about building and operating the infrastructure to support our customers' future and growth.
Kind of I think it's a great place to start on the demand side. You're coming off a strong year for new leasing. What's driven the recent performance, and how sustainable is the current demand environment?
These are long-term secular tailwinds of demand that we are essentially supporting, that are foundational technology for our customers. Last year, we had a record of $1 billion of bookings. Underneath that, there were records in our 0-1 MW enterprise co-lo and interconnection category, multiple records in consecutive quarters, as well as records in our greater than 1 MW hyperscale, numerous megawatt category for our customers. We are still supporting customers moving from on-prem data center solutions into purpose-built infrastructure today, and that wave is still running for many, many years to come. We are still seeing the build-out of cloud computing, the globalization of cloud, incremental services being offered via the cloud. You look at our top customers, many of those are the cloud hyperscalers with us in 30, 40, 50, 60 different locations around the world.
We are just getting started at what AI infrastructure will mean in terms of incremental growth. I said for a long time, I think these trends are all up and to the right. They are not a perfect linear line, so there will be some volatility along the way. The way we pursue our strategy, how we go about it, the customers we support and pursue, and the platform we offer them, I think, is tremendous durability to monetize that growth.
Jensen Huang last week, the CEO of NVIDIA, last week on their earnings call, suggested that AI workloads are transitioning from training to inference. Where's the demand for inference sort of materializing within the data centers? Are you seeing that demand today?
I think that's a very relevant statement, but I think that statement somewhat overstates where we are in this build-out of infrastructure. I think it's very relevant because it is really the next stage of AI, which is going to be the mass utilization and the B2B enterprise use cases. Our portfolio is not in every city. We're not in every NFL city. We focus where there are workloads or applications that are locationally or latency-sensitive, and they have clustered in a phenomenon of data gravity that has brought numerous customers of diverse industries, as well as all the cloud providers in these 50 metropolitan areas. I think the transition to inference as being like a rapid shift, it may have started to change quickly, but it has to be in its nascency.
I see very little usage of the AI products out there today, such as Copilot by Enterprise. I do not see a fraction of the agents that could be helping each and every one of us in our day-to-day work. I have not seen any of the potential of use of this technology in the robotics for manufacturing, the consumer experience in our daily lives. I think this is a long tail of incremental demand that we at Digital believe will accrue to these where the locational sense of workloads and the cloud live today inside our four walls.
Do you think over time it sort of starts to really be a driver in that 0-1 segment of yours?
We've had great success really executing in that 0- 1 enterprise segment. I would say our success is on the backs of many years of investing and building this platform and stretching it globally and innovating our product and changing and evolving our go-to-market. Quite frankly, more recently, just very focused and consistent execution, which you saw really in a string of three quarters to date. I would say two things are not there yet. One, the AI portion of that uplift is still yet to be seen. We had, I think, a second highest quarter in that category, and I think 10% of our signings were from AI use cases. I think 10% may be an exaggeration because those were certainly skewing towards larger deal bands.
If you counted the number of transactions we did that, it has to be way less than 10% in that category. Part of this is the solutions are just being built. The products are not even on the shelves when it comes to enterprise use of AI in its full extent. I do think this can be a more meaningful contributor to the growth in that category. We're not sitting here waiting for that to happen because we're executing on those customers, those hundreds of new logos that landed with us this past quarter, or the 600 that landed with us last year, who are adopting new clouds and embracing digital transformation for their infrastructure, and more importantly, becoming AI-ready for that infrastructure and the inference and those applications to come.
That makes sense. Maybe sticking with AI on the training side, how should we sort of think of the demand curve there? Obviously, there has been a lot of noise in the last, I would say, 2 months about the demand curve there. How much runway do you think we still have left on the training side, and how does that impact how you guys sort of deploy resources from a geographic standpoint?
If you look at the estimations of the chips being sold and ultimately shipped today, whether it's training or inference, you'd see a large continuation of growth for large capacity blocks for the hyperscale cloud customers. I would say it's been, while it's a global data center industry and we are supporting global customer bases, the lion's share of AI we've experienced is happening in the U.S. today. I believe you haven't even seen the full extent of globalization for the use cases. You've seen a lot of announcements, but if you're announcing a data center right now, it's going to be years until that's coming online, especially a large-scale one. We have had a tweak to our strategy over the last few years.
One, it was pivoting this company to putting the enterprise connectivity solutions and co-lo first in our pursuit, and we wanted to gain market share, and we've been doing that and will continue to do that. When we took our push to hyperscale, which is germane to where data center demand in large capacity blocks was landing today, we said, you know what, we need to find places where we can add the most value to our customers and not be all things to all customers and be everywhere. Fortunately, of our 50+ metropolitan areas, more than half are places we have a tremendous value add for those hyperscale customers.
As you saw from our major signings last year and going into the first quarter of this year, places where we already had those customers' infrastructure in those key markets, places where we had a distinguished longest runway for our customers to grow on our campuses, places that are more challenging to do business, we could really help our customers. That is where we have intersected demand. I'm sure some of that demand landed to be training, but these contracts are 15 years long. That training may evolve to inference. It may be a mashup of inference and cloud. We are essentially building infrastructure for the evolution of their infrastructure over time.
That makes sense. Again, sticking with AI, Jensen also spoke about sovereign AI initiatives around the world becoming a meaningful driver. That was one of the sort of other big sort of outtakes from his earnings call. You talked a little bit about globalization and most of the activity being here in the U.S. Is it starting to permeate these other markets? Because you're one of the few data center companies we talk to that has a real, truly global footprint?
Literally two Mondays ago, I spent my morning in Paris meeting with one of two sovereign clouds for the country of France about their growth and their infrastructure and their sovereign cloud, which is growing adjacent to multinational public cloud intentionally. I spent my afternoon with my second visit with the French government about bringing AI infrastructure more rapidly and scaling to Europe and certainly France. The initiative is there. The focus is there. If you look at the history of cloud and data sovereignty, you would see a similar proliferation of U.S.-focused build-out, globalization with Europe and Asia following. I believe over time, you're going to see a similar phenomenon. Things move at different paces in different countries, obviously. I think there's a pretty much unified theme is that almost all these countries want to be part of this AI and technological arms race.
No one wants to be left behind. No one wants the infrastructure all to be homed in the U.S. I think you're going to see a continuation just like the data sovereignty cloud with the globalization for AI on multiple parts of the world.
I mean, first of all, was that the first sort of AI sovereignty meeting you've had? And do you think it permeates to the level where you're doing like there needs to be an AI cloud in the U.K., in the major markets in France, in Europe, major markets in Asia, India? I mean, is that.
That's not the first AI or excuse me, that's not the first sovereign cloud conversations we've had, nor the first sovereign clouds we've landed here at Digital. If I look at the umpteen countries across Europe, I've only seen this in a handful to date come to full fruition of partnering with a public cloud provider, bringing together the expertise and system integrators that need to be facilitating this, setting the security standards for what really needs to land in a sovereign cloud versus a public cloud, getting the funding to come together for projects like this. For this all, that string of activity, I still think it is incredibly nascent to where the potential could be.
Got you. Where we are, and it used to be today's news, now it seems with AI, yesterday's news. Where would you say we are in terms of the move to cloud in general? Which, again, before ChatGPT and AI sort of burst on the scene, that was the main topic of conversation. I mean, what inning are we at in terms of cloud adoption with major enterprises and globally?
It still feels like we're early innings and the true potential. I mean, just the amount of on-prem workloads, the amount of mainframe uses today still boggles the mind. I think that the earliest days of cloud, you obviously had one particular leader. Then you had the multi-cloud. You had the acceptance of hybrid cloud. Different clouds and providers are getting excelling in different categories. I think AI is going to further reinvent that wheel in terms of called best of breeds in different applications or use cases. I think this is going to be an evolving technology landscape, but I think more is better, right? More is better for the end user, the individual. More is better for the enterprise business. More is better for the data center landscape, and more is definitely better for Digital Realty.
One of the big topics of conversation in the data center space is power procurement and power. Finding power for these massive workloads, both on the cloud side and the AI side. We saw the announcement this morning from Meta and Constellation. How is the line of sight from a power procurement standpoint that Digital Realty Trust is seeing? Are things getting better, or what is the situation today?
Things are getting better every day because time is passing and activity is proceeding and people are focused and resources are being marshaled. There was no easy button to fix these problems. I was in Ashburn just last week, and it was great to see the high transmission poles for the Mars substation erected landing on our Digital Dallas campus. Now, there were not any power lines running through them yet, but the poles are up. We are stepping inch and inch closer to pain point reliefs. This is a complicated problem we have here, right? There has been an underinvestment in the critical power infrastructure in the United States for many, many years, right? Industries have been less, less power when the need for power has just skyrocketed for technological advancements. This cuts through federal, state, municipal, environmental, numerous issues we are talking about here.
I think you're going to not see one quick fix here, and it requires ourselves and the energy industry to innovate, which we've been doing at Digital. When the shortage in that very market I mentioned came on the scene, we went to our own infrastructure and said, where can we move around electrons and find idle capacity? That was labeled us to pull forward and use more electricity by looking at what we had and was not efficiently deployed. Longer term, in certain places like South Africa, we're investing directly in solar, which will be wheeled to our data centers through the grid, and one project will likely be behind the meter. In between those timelines of the longer term and yesterday, we're looking at other stopgaps for long-term bridges for power until the utility infrastructure can catch up.
That was going to be my follow up question. Just the rest of the world, I mean, how is the situation that we have in the U.S., is that similar outside the U.S., or do you foresee a situation where they're going to have similar problems that we're having now in meeting this demand?
You've seen similar problems. The same problems that happened in the U.S. were happening outside the U.S. first, came to the U.S., then spread back the rest of the world, whether it was moratoriums in certain countries like Singapore or the grid in Amsterdam saying no more data centers. Just the same problems are repeating themselves. A lot of it is the power infrastructure. The power infrastructure goes from generation. We've been moving from, call it, towards a greening of our generation. I mentioned the transmission. You have the substation components, which have supply chain elements associated with it. You also have the major markets that are butting against some NIMBYism around the data center. People are losing focus about what we're offering in our data centers. It's critical infrastructure.
There are technologies that are running the hospitals and the ambulances, solving diseases, innovation that's changing the world and needs to be near GDP populations, critical infrastructure, network connectivity. All those things are creating supply bottlenecks along the way.
Right. What is the sort of the difficulties in achieving power and the supply constraints doing to overall sort of IRRs in your business? Is it changing the return equation at all? I mean, are you investing more?
You could say it initially diluted the IRR because we were pushing out the expectations. I would say it also kind of increased it or created the IRR because the demand inflection has happened in a time of supply constraints. You have seen a pretty strong uptick in rates that translated into much higher returns. I think the net benefit has turned this into a more healthy economic equation.
Right. That's a great segue to pricing. Can you talk about sort of what you've seen recently and maybe give us a little history in terms of the pricing environment, I'd say both in the sort of 0- 1 MW and the sort of 1 MW+ sort of markets?
In the 0-1 MW category, you've seen a more consistent, less volatile, positive and increasingly positive pricing dynamic where, in excess of inflation, it's been able to pass through on prices on renewals of existing contracts without increases to churn, as well as rates on new offerings. Those contracts typically have a shorter duration to begin with, so we have more bites at the apple to keep the repricing. We are delivering, I would say, even higher value add to those customers. They're buying a platform offer here at Digital, multiple markets, multiple products with us. They are, by and large, most customers that are never going to build their own data center. It is a long tailwind of demand. I would say the data center is turning from a security feature to a cost optimizer and a revenue generator, right?
It's becoming the center of your infrastructure when it was never that before. Its relevance is even greater. On the hyperscale side, similar trends, but has been more exposed to a positive inflection as the rush for the large capacity blocks have happened when the large capacity blocks have been even more greatly constrained and likely to continue to be more greatly constrained. Along the way, you've seen the buyer base also widen out a bit. Each incremental buyer, even if it's one, two, three, or four extra buyers buying in the 50 MW, 100 MW tranches, makes it more competitive and puts more pricing power towards the incumbent providers. We've had, of the five last quarters, four were near records or just very sizable quarters. The largest signing in each of those quarters was from a different top customer.
None of those top customers of those four is our current top customer to see a broadening of the hyperscale buyer base.
Makes a lot of sense. We have 10 more minutes left in the session. I have a few more here on my list, but if anybody has a question, please, you can ask it at one of the standing microphones. There is a question from the audience right there. Maybe if you could use the microphone over there, that would be great.
Hi. Thank you for your time. One of the things I was wondering is, do you find that at least moving forward, powering data centers will rely more on front-of-the-meter or behind-the-meter solutions?
I think you're going to see a behind-the-meter solution become more prevalent, but I don't think you're anywhere going to see a world where the majority of data centers are being powered behind the meter. The grid is a valuable asset of resiliency, sharing of infrastructure. I think you're going to see the concept of more long-term bridges, i.e., as we wait for load studies to get done, as we wait for infrastructure to get built up, as we wait for substations to get delivered, that a behind-the-meter solution will become more relevant. I don't think you're going to see a total divorcing from the grid. Quite honestly, it's preference by the hyperscalers to rely on it.
Just as a quick follow up question, do you have any comments regarding some of the review going on in the PJM region and other RTOs and ISOs, particularly with what's happening regarding the feasibility of front-of-the-meter solutions, especially because now behind-the-meter just seems that it's far past the timeline and capital costs many of these hyperscalers are willing to commit to?
I'm not going to comment on multiple regulatory bodies or utilities. We're all incented. We're all in the same boat here. By and large, we want this country to build better technology. We want the world to be building better technology that improves all of our lives here. We're all together, whether you're building the molecule of the electron or the transmission of the electron or the distribution or the data center or the consumption inside the server, we're all in the same boat to get this right.
Thank you.
There's a question over here?
I've seen things have moved quite quickly in terms of the technology and infrastructure in data centers. Can you just speak to what's the lesson in terms of impact and effects around specific data centers?
I'll repeat the question. I think in a nutshell, the question was, can you speak to, given things are moving quickly here, can you speak to obsolescence risk inside the data center?
Related CapEx.
Related CapEx. Thank you, Jordan. By and large, the infrastructure we are building, owning, and operating, if you take a step back, is not tremendous rocket scientists. It is a highly improved piece of commercial infrastructure with heavy floor loads, redundancy of power feeds, fiber optic connectivity, and cooling features. The biggest evolution is right now playing out on the cooling because the power densities are starting to ratchet up for use cases. The misnomer is that all power densities for all data centers and every single bit and byte is going to be at science experiment type power densities to work. That is not very likely, in my opinion, going to happen. The main solution for cooling is almost going back to the future, which is using water as the cooling agent.
We're just bringing the water closer to the heat for dissipation of that heat. Our experience has been, first off, we've done liquid cooling for many, many years ago with numerous customers. They weren't the majority by any means, but we've done this for many years. This isn't the first time we've done this. Two, we still see customers in our capacity refreshing the existing infrastructure with the latest CPUs and not going to the most outlandish power densities you can go to. We have new customers signing with us in 100 MW blocks that we give them the option, do you want to go 80% liquid or 50/50? They're picking 50/50 for 15 years of a contract term.
The cost to retrofit, which we have done, which is often borne by the customer because sometimes they're in the contract or they even come to the end of a contract and they do not want to move, has not been a massive upgrade to the infrastructure. I do not think that there's tremendous innovation happening inside the server, the CPU, and the GPU, and the networking gear. I'm not sure the physical infrastructure to dissipate heat is as earth-shattering as you might think.
Yeah. Any questions from the audience? I've got a couple of follow ups, Andy. So the company's made a lot of progress in recent years to leveraging the balance sheet and diversifying your sources of capital. How should investors think about the level of development spending going forward and the funding sources?
We've made great progress at essentially burning the candle at both ends here, bringing the leverage down from seven times to close to five times, building liquidity up to $6 billion today, and adding to our stable of funding, first with development or joint venture partners and now with our first fund. That's a little bit different than what we've done with venture partners. This is numerous, highly sophisticated LPs doing their work on us, diligent in digital, placing their trust in us to invest on their behalf for what a vehicle now we announced. We're just north of $2 billion of commitments, and we just closed on $900 million or so of proceeds on the first phase of this transaction. We think we'll continue to grow the fund.
That's a foundational stepping stone, allowing us to scale a hyperscale private capital management business where we think we have a clear path to be a leader in that segment, giving investors the opportunity to not have to invest in a co-mingled fund, to not have to invest amongst other infrastructure investments or other telecommunications, and less invest alongside Digital in a pure play data center investment for hyperscale, which we've been doing exclusively and only for 20+ years as a business, time tested, cycle tested when it comes to the experience of the company. It allows us, in response to demand, to pull forward as fast as possible of the call, 4 GW of demand. Just using rough numbers, 4 GW could be $40 billion, $50 billion or more billion dollars of potential investment over time across those 50 metropolitan areas.
We are adding to that growth as we speak. The land that we have been procuring have been strategic additions to our connectivity footprints or hyperscale lands in those same type of markets I described, locationally latency sensitive cloud zonal markets that have time envelopes that are where the customer is most focused on today.
Got it. The JVs are sort of a new investment vehicle for you guys, or source of funding vehicle. How big can the management fees that you guys generate from these projects become in this sort of overall sort of P&L?
I just quoted we're north of two on the way to a target of two and half , probably max just over three. That's equity. That could be close to $10 billion of investment. We announced a joint venture with Blackstone that, I mean, that by itself could be another $15 billion, $20 billion of investment. I look at today, and there's another REIT in the halls of Nareit in the industrial space that has a $90 billion called private capital management platform in the industrial space. The last time I checked, your typical warehouse was a lot less expensive or valuable than your average hyperscale data center. We're in a backdrop, if you listen to numerous industry sources or NVIDIA's and others, where the runway of growth for this infrastructure is quite stupendous.
Right. Okay. Lastly, for the last question, just putting it all together, both the sort of demand environment, what we're seeing in pricing, the deleveraging, the new sources of capital, just how should investors think of the opportunity for FFO per share growth in 2025 and the sort of growth curve beyond?
John, we should have started there. That is the guiding light. That is what passes not prologue here for Digital in that manner. We had to do a lot of things over the last several years in terms of changing the strategy and the execution of the company, the funding we talked about. Last, maybe most important, is making all this growth, pricing power, organic growth, and development flow to the bottom line efficiently. I think even before last year, this time, we basically said we're on a new trajectory and threw out a number of about 5%. We're now out with guidance now for this year, about 6% constant currency. We had a first quarter in that guidance that is called confirming that path. We're looking for a year next year that's going to be better.
We're looking for that runway of consistent compounding of the FFO per share growth to extend as long as possible in that territory.
That's fantastic. I think that's all the time we have. Andy, thanks for joining us today.
Thank you.