Good afternoon, everyone. My name is Brendan Lynch. I cover communication infrastructure here at Barclays. I'm joined today by Equinix's Chief Accounting Officer, Simon Miller. Simon, welcome.
Thank you. Appreciate it.
Forward-looking statements.
Yeah, absolutely. So, we'll be making some statements today, so please refer to our SEC filings for Risk Factors affecting our business.
Great. Maybe as a place to start, Equinix has 20 years now been enabling companies to execute on their digital transformation. Maybe you could just give us an update on where the average enterprise is in that process, maybe any changes in the state of evolution.
Yeah. Really good question. And it, as you can imagine, I'd say it varies across the board. You've got your kind of folks operating on the edge that lean in early. They're probably a good portion of the way through it. I like to consider Equinix one of those companies. We're going through our own ERP transformation right now, converting from Oracle on-premises to Oracle Cloud. ERPs are critical operational set of tools to operate the company. And what we found when we go through that type of transformation is that there's a host of ancillary applications back that need to be converted as well, because they don't tie in from an on-prem basis to a cloud-based really well, and you don't get all of the features.
So a company like us, we're only about halfway through that. I've got at least another year on that transformation just for the ERP, and then we're going to start to get to work on some of these other ancillary applications with that. There's probably some other real quick early adopters, smaller companies who are going to be able to do it a lot quicker. But when you're talking about that broad swath of enterprises, it's, it's probably a, you know, five-to-eight-year kind of transformation that's going to take place for companies over the next several years. You know, the big companies like SAP, Oracle, Microsoft, they're, they're all going exclusively cloud in the future. So if, if you're on-prem today, you're, you're going to need to acquire some sort of, you know, cloud architecture solution in the future.
And you're going to need to do it at a place like Equinix, quite honestly, where all of the other tools that support that transformation, all of those providers are there within Equinix, as well as the carriers that provide connectivity into those providers.
A lot of progress, but still some room to run.
I still feel like it's very early. I mean, for us, you know, we had a few delays. There's just a lot of learning that goes into it. You're really moving from an architecture that has, you know, for, you know, the better part of 15 years. Ripping that a cloud solution is different, is what I would say. You know, you don't customize anywhere near as much, but you have a lot more configuration capability. And I think going through the evolution, really looking at what do you do inside of your four walls, and how do you have to do it differently, that was part of the process that I think we didn't fully appreciate going in, and took us a little bit of extra time as we move through the development phase.
Interesting that, to hear that Equinix has to go through.
I know. I actually wish our CFO would talk about it a little bit more often, because we live it internally.
Yeah.
But yeah, no, we're drinking our own Kool-Aid internally, for sure, and utilizing our own resources.
In terms of the next wave of digital transformation, I would think AI is getting an enormous amount of attention. A lot of companies are talking about how, when they're asked about AI, they say: Oh, we've been doing AI for years. I think there's certainly some truth to that. Maybe you could talk a little bit about legacy AI applications, for lack of a better term, versus some of the AI demand that might be stemming from ChatGPT or a type-
For sure.
demand over the last 12 months.
Yeah, and then we have—like, we've had AI inside of our four walls for several years now. It's, it's—they're more niche applications. When you think about ChatGPT, that is a consumer-facing, large, massive thing. But you've got various use cases for AI, where it's real-time AI analytics, doing something along the lines of for an internal use application suite and enterprise that is very focused. That type of stuff has actually been around for, you know, the better part of four or five years, and certainly increasing in use. By the way, that's the type of stuff that is perfect inside of an Equinix data center, because you're trying to crunch a ton of data, like all of these works historically, and process it all and give real-time answers.
You're using a smaller data set that is very easy to provide compute right next to connectivity. That's the type of stuff that is getting launched and has been launched now. I think it's going to increase in size and volume as, as some of the bigger, I'd say, tool and SaaS providers push AI ultimately into the enterprise in a more, I'd say, generalized fashion to be able, not to necessarily replace, but enhance a set of tools that they have out there. That'll be another layer of complexity. And then there's, of course, the consumer-facing, kind of a big, transformation point in generative AI applications.
Are you seeing any of that generative AI, large language model demand yet, or is it, that's still probably a few years out?
We probably won't see that on the retail side of our business, but we're definitely getting a decent amount of inbound requests on the scale side of our business. So, what you're describing, the large learning models or language models are much more compute focused. They need a high amount of density in terms of power. Generally, the deployers of that, if you're thinking about a hyperscaler, which is probably where they want to play mostly, they're looking for areas that they don't need to be super close to network nodes. So, they're looking for locations where they have a ton of access to real estate expansion and power, because they want to build facilities that start with 200 MW, get as big as 500 MW, maybe as big as a gigawatt.
They want access to the power. They want access to the power cheaply. They don't need to be around large city centers. It's a high compute environment. Where that stuff eventually gets pushed into the inference model, is where I think the, pardon me, the opportunity for Equinix on the retail side of our business, will start to expand. And we're starting to see a little bit of that, but again, it looks a little bit more like the niche, homegrown AI model that people are serving up for themselves.
Okay. Yeah. I want to come back to xScale in a minute, but maybe just a few more on AI. You had some wins recently with CoreWeave and Lambda Labs, Crusoe, I think a few other AI startups as well. Can you just characterize the nature of that leasing? Is it a large footprint in a single facility? Is it distributed throughout the U.S., throughout the world?
Yeah, it's a couple locations and really about leveraging our network nodes. It's that piece of delivering the inference out on the edge of the footprint, as opposed to all of the core compute that you've done on the back end.
As you're building out into the AI realm, I think that the positive would be that you're establishing a new ecosystem, and kind of tapping into it as you did with the cloud for the past, 15 years or so. Maybe on the, the negative side, it's an unproven business model. Talk about how you consider some of these risks and opportunities.
Yeah, it's definitely unproven, but I would say, boy, it does feel like cloud 2.0 all over again. You know, back in the day, there were a lot of questions around our business model, whether the cloud was friend or foe. You know, at Equinix, we have just some of the best team of thinkers on the technology side. We really sat down and developed how do we become the home of the cloud, and blend that with our network connectivity to offer a solution to customers. AI feels like the same. It feels very similar.
Once the inference engines get put on, you know, to the enterprise or to the data center, the way we're looking at it is it's just another stack in that suite of applications that I was talking about as a business leader, that I need to implement and have supported. We're thinking a lot about it as it being just another stack that companies are going to need to have highly interconnected, talking to other sources of data in their architecture. If you think about it, so I have a bit of AI in my world.
It's mostly, wouldn't call it fully generative, but those AI bots need to be proximate to where my data is, because it's grabbing data out of my ERP, it's modifying it in some way and pushing it over here for use by either another human or another bot, right? In my case, that's very sensitive. And so I would expect, and the way that Equinix is looking at this, is another technology stack that ties in really well with corporate digital transformation in general, especially when companies are using a host of cloud-based architectures, Infrastructure as a Service or Software as a Service.
Great. As you're getting more of these AI-oriented workloads or your power density is increasing in general, talk about the metrics that you're presenting to investors now versus maybe what you will be in the future to kind of capture how things are evolving?
Yeah, sure. Right, right now, probably our most important non-financial metric is cabs billing. Actually, this last quarter, we had a lot of internal dialogue because we saw a flat cabs billing, and it was a bit of a surprise to us, and we went back and started digging into it. One of the things that we are seeing is, as we're churning out older deployments, customer deployments in our data center, I'd call it 4 k VA per cabinet. We're replacing that now with densities at 5.7 kVA per cabinet.
We think a lot of that has to do with the underlying economics for cabinet, but also the things that they're deploying overall are more power intensive and need to interact more with networking gear over here, cloud providers over here, cloud providers over there. So I would expect to see that density, that metric, increase. I would expect us as a company to go back and dig through that a little bit and try and figure out if there are other KPIs that we can use to enhance cabs billing.
Right now, we feel like it's by far the best metric to look at the business because it captures really not just the power that we charge for a cabinet, but it includes interconnection, which is a huge part of our managed services, and then any digital services that layer on, so that we can kind of project out what the yield potential is on a per cabinet basis for the investor base. But we're starting to realize that that's-it's a little more complex. There's more to the story than that, and there's timing differences. So I would say there will be more to come on that. We're doing some internal... trying to figure out what suite of KPIs and metrics we could add to enhance that won't confuse or [inaudible] so that we won't balance.
It totally is.
Well, you know, you don't want to rush to it, right? Really, there was a bit of intrigue internally, where, like, do we start disclosing that right as part of this earnings? Thankfully, folks like Katie Morgan on our team are like, "Let's slow down until we can test this and see the trends and make sure it's operating the way we think it is or should.
Sure.
Yeah.
You mentioned xScale. You have your first xScale facility here in the U.S. now in Santa Clara. Maybe talk about what you see as the opportunity going forward, to expand that, here in the U.S.
Yeah. Traditionally, our xScale strategy is focused on retail proximate facilities, so that hyperscalers, in general, can leverage the connectivity that they already have in our data centers. Certainly the last 12 months, I think the uptick in AI, although we had it on the roadmap, but I think it's just moved forward at a pace that, with ChatGPT's, you know, launch and subsequent announcements, is sort of quickening the pace a bit. So we'll continue to invest in that strategy, one type of deployment, because we think, you know, where there is some level of compute that needs to be close to our data center, we build. They're more built to, and the economics are a little bit better, will help our customers deploy there.
But I, I think in the U.S., you're going to see us focus, a little bit more on those AI, AI-driven type of opportunities, which, which are, are bigger in scale. Our traditional xScale deployments are about 30 MW-40 MW, and we require that they're close to our retail data center so that we can leverage the connectivity. I think when you look at us, expanding on the xScale side to take advantage of this AI opportunity, you're gonna see us get further away in remote locations, away from the high density network connected data centers, and closer to places with expansion opportunities on real estate.
There's a lot of dry powder out there-
Yeah.
In the infrastructure world looking to be deployed. What are you looking for in terms of a partner in the U.S. for the xScale application?
Yeah, I think somebody that shares that vision and has the right connections that can help, you know, grease, if we want to get to business. But more importantly, somebody that's happy to let us lead the charge, because you know, we're certainly a leader in this category. We've got the brand, we've got the customer connections. We build and operate and support data centers, we think better than anyone in the world. So somebody that's happy to take that lead and partner with us. You know, the size and scale of the investments that are gonna be necessary are pretty big, to your point, Brendan, though, there's money waiting to find a home. And so, I mean, not much of a shortage of people lining up to talk to.
It's really going through that diligence process and finding that cultural fit, that operational fit. Sometimes, you know, purely financial investors, when you explain operational challenges, it's a bit often. So you want somebody that has at least an intuition around what it takes to operate. Things rarely line up the way you put them on a piece of paper, the way that you model things. And so just working through that, finding a partner where the cultural fit, where the strategic intent, the time horizons for exit and additional funding requirements will match up.
So it sounds like from your previous xScale JVs, there were more, you had a financial partner on all of those.
Mm-hmm.
Sounds like in the U.S., you might have more partner that has some, at least operating expertise.
I think we'll still lean more towards financial, but making sure we get that fit of somebody that understands the operational nature.
Sure. In terms of the going into the business, building these type of scale facilities, I'd argue that there's very strong demand right now, but there's arguably lower barriers to entry, which compared to your interconnection hubs, we see relatively modest cabinet additions over the past few quarters.
Mm-hmm.
But I would argue much higher barriers to entry with anything but those risks and opportunities.
Yeah, while I look at it and monitor it on the retail side, I don't, I don't really worry about it quite honestly, because, I mean, the, we have such a defensible, you know, market position, like you said, that the, the, the cost is just like... You cannot build a data center with 1,800 carriers in it overnight. Attracting those carriers and then getting the people to come in, connect to those carriers and consume services, you know, it takes 25 years. That's-
Sure.
It really does. And then minimally in a single market, it probably takes you at least 10 years. And so being able to sustain that type of, you know, ramp up phase is very challenging for folks. While there is a lot of money on the sidelines, people, I think, still wanna utilize debt to get there, and I think the interest rate situation we find ourselves in, and probably will here for a bit more, it's a slight barrier to entry. I think the way we think about it is, look, we're the best builder and operator of data centers on the planet. We've got a very strong brand.
Once we nail down with our customer where they want to go, how they want to go, how they want to get there, procure the land, you know, make sure that we've got the power allocated, we'll be ready to rock. I mean, I think the thing that we have is that brand, the thing that we have is our balance sheet to leverage. In the meantime, I mean, we still have a couple of turns of debt before we make the rating agencies uncomfortable. We've got a healthy revolver. So we have a ton of flexibility there, and I think our brand just gives us an immense lead in that category.
The brand and the ecosystem, I think, are-
Totally.
Very-
It's a little bit different on the AI side, but if you become the place where AI gets done on the learning side and have a solid track record on inference, even though they might be small deployments, it feels like Equinix will be the place where AI is done in the future.
In terms of the demand that you're going to see, you alluded to it earlier, cabinet additions have been relatively modest here today. I think they're only up about 0.5% after growing some 25% over the past couple of years. I've heard Charles Meyers describe 2023 as a year of optimization. Where are we in that process? Do you think that could continue into 2024?
Yeah, that's a great question. I tell you, that is probably the biggest question we've been asking ourselves internally as we're going through our planning season right now. And I'll say, I think our bet is that, you know, things will maybe clear up a little bit compared to 2023, but there will still probably be a bit of volatility. I think Q1 for us will be a [inaudible] really quarter, seeing how we perform against our pipeline will be really, really interesting to me. I'm an accountant, so I always kinda think in terms of planning cycles within the business and just knowing that most enterprises out there are nailing down their budgets right now.
Once those budgets get pushed down to the functional leaders of the company, they will have their authority to go spend starting in Q1 of next year. You'll get a really good sense of how those enterprises are planning internally when you see the activity. I would say we're looking at things a bit cautiously for what's going to happen. Really interesting, I think, shifts towards AI. It may slow down a little bit. I still believe enterprises are in full pursuit of digital transformation. It just might be a little bit lumpy around that.
Is it generally macro uncertainty that-
Yeah, I think so.
-the slowdown?
It's people dealing with inflationary challenges out there. In our space, specifically, we had to absorb a really big inflationary hit last year on power. Don't expect anything close to that magnitude, this year, which is definitely going to be helpful for us. It was a bit of a challenge to talk to people, you know, when you're pushing out, you know, several hundred million in power price increases, to your installed base, that, "Hey, can we sell you a new implementation as well?" I think that will clear up a little bit here in 2024. But there's still long lead times for customers out there and supply chains. Inflation globally is still pretty high, especially outside of the US.
And there's a host of companies that maybe, you know, the announced actions that they were going to take to cut costs within 2023 are probably gonna-some of that's gonna extend into 2024, the action taking in support of those cuts.
On the power front, maybe you could give us an update on where you are in terms of hedging going into 2024, and also how you will be handling the power increases that you have passed on already in an environment where energy costs are generally coming down?
Yeah. So, we'll give an update on how we're heading into 2024, so I'll just refer you back to the Q3 on what we said there. But generally, you know, where we can be hedged, we're fully hedged. I think that's a benefit to our customers, and last year, when we rolled out a ton of power increases to their bill, we had to walk them through that, how we actually saved them money by being hedged coming into the year. The way that we do it is we kinda look out a couple of years, and we put about four years of layering in there so that we've always... We're really just, you know, cost-balancing the overall portfolio.
Sometimes it's in our favor, like last year, sometimes it doesn't, definitely in a time of increasing prices. So we're going into next year at least feeling like the volatility around that is better. In some markets, we're expecting to see power increases in terms of prices. In others, it's probably going to be... I think it's appropriately pretty consistent, both to the investor base and definitely with our customers, that we'll push that back down to them, that we're seeing material increases in pricing, and they're contractual in nature for us, or at least we feel like they're going to be, we'll be able to sustain them for a period of time. So some volatility, but a little bit less. And where we can be hedged, we'll be hedged, or close to fully hedged there.
That's for sure.
When we look at your pricing year to date, it's up about 9%. Maybe, what component of that is power-related versus just price increases on to the customer base? And any difficulty you might see doing that going forward, given some of those budget constraints and some of that macro hesitancy that you're seeing, contributing to the relatively modest cabs this year to date?
Yeah. My gosh, I'll have to get back to you on that, on the specific breakdown. It's a good, it's a good portion of it. Katie, do you know offhand, by any chance?
Yeah, if you look at the MR [inaudible] quarter report, and [inaudible] roughly about 42 or 43 was the underlying pricing adjustment-
That's right.
Including the power pricing.
About a third of that is related to power. And the good news about how our contracts work is that we generally have a price escalator built in, but they're not automatic. Historically, we've been very strategic about pushing those down. Where we're seeing real inflationary costs, we'll continue to maximize whatever leverage we have against the agreement, push that down to our customer. We will probably be a little bit more strategic depending on the reaction there, because we have that flexibility. At the end of the day, like, 90% of every dollar that we book each quarter comes from an existing customer. We just have to be cautious about doing that. But we have a ton of flexibility within our contracts to do it, anywhere from 3%-5%.
These days, when we're locking in new agreements, we're putting, we're maximizing that language as much as possible, and we're stepping people up to the, what I would call, new, post-inflationary rates, list prices that have been increased by, you know, anywhere from, you know, like, 8% to maybe 12%, depending on the market. So we're gonna. New deals are getting uplifted when we go through renewals. When we go through our annual PIs, we'll be very subjective. But we're gonna be very sensitive to what we're seeing out there with the customer base.
Again, I feel like this year will be an easier time to have that conversation, because last year you're hitting them with a contractual PI at the same time that you're doing this big power PI, and so that just felt like a double punch. This year will feel-
Like a single punch.
Yeah.
Yeah, I can imagine last year was a difficult conversation.
It was.
Some pretty assertive price increases.
I do, but I, I give our team full marks because we started towards the end of the middle of the previous year and having conversations with customers. We sent them three notifications telling them, "Here's the date that we're gonna get back to you with the range of your price increase. Here's the date that we're gonna come back to you with a definite increase." Trying to hit it at a time where they, they would be able to influence whatever budgetary decisions they were making back at their company.
Maybe one accounting question. Maybe walk us through the variety of maintenance CapEx costs that you have, and which ones you consider operating expenses versus recurring CapEx. I suppose the genesis of the question is that PP&E has grown at about 11% over the past six years, while recurring CapEx has only grown about 5%. I suppose there's an element of you have some newer facilities that are probably just operating more efficiently.
Yes. Yeah.
But maybe there's a consideration that there's some catch-up in maintenance CapEx coming in.
Yeah. No, we've got some facilities that are, you know, well over 20 years old. This is actually something that's been on our mind for a few years now. Generally, when we go in and do a big maintenance project, we put a lot of pressure on our design and construction teams to release capacity. Where we are able to release capacity and live up to the standard of how we define AFFO in our earnings release, if we create capacity for incremental revenue, not the main revenue, we push that into expansion CapEx. And, you know, that takes the shape of replacing CRAC units so that you can provide more airflow to the white space, and therefore distribute more power, so you can take cabs off of engineering holds, right?
So I may be, after release, after one maintenance project, release 300 cabs in the data center. Well, that's 300 cabs of incremental revenue, and I'm actually gonna get a return on it, so we'll put that into expansion. Where we don't have that opportunity, it's just traditional recurring CapEx. We've got a s you mentioned, we've got some big ones coming up, putting a ton of pressure on the team to just rethink how we might, how we might engage in replacing some of the bigger parts of that infrastructure, definitely around air movers, rack units, in some cases, power distribution.
The great news is we've got one of the best design and construction teams, I think, in the business, and they're coming up with very creative ways to not only, you know, give us assurance that a critical data center is gonna stay up and live and relevant, and meet our customers' demands, but actually improve on the efficiency side, drive down PUE, and get us additional capacity. So for us, that's affinity and what we're always trying to achieve.
Small order, but-
It's a big one, but honestly, it's a great question because we've been looking at this for a couple of years now. This is not something that, like... We've been sort of seeing this coming for a while. Started having, you know, a couple of just high-level conversations with the team three years ago. But when it came to putting the actual plans, you know, on paper and putting execution and the forms around it, it happens, right? If you talk about it enough and really explain to people why it's so important, and show the value that you can create and why, if you're an IBX operator, your project has a higher likelihood of getting approved, but not just, you know, being planned, then everybody gets on board with it. We get some amazing solutions out of it.
Stuff, quite honestly, we're good stuff right now that I just never—I really didn't dream two years ago, when we were kind of looking at this probably five, seven, year period of, I'd call it, we call it a refresh, but maintenance is probably a good word for it.
Great. Well, I look forward to seeing some of these developments-
Yeah.
in 2020, 2024.
Absolutely.
Thank you, Simon.
Absolutely. Thank you, friend. Appreciate it, bud.