Equinix, Inc. (EQIX) Earnings Call Transcript & Summary

I mean it goes back to what you just said, which is we build the building, we build the shell, we provide the interconnection and the power. And therefore, the depreciation structure and the CapEx structure are very different than other property types, but also different than if we own the actual hardware that was in the building.

I mean I think like some of our equipment is like 40-year life as long as you're maintaining it well, which we do, it can last for quite a long time. And where we have to update it, we update it.

Can we skip to power?

I just see more -- 2 more. Sorry, I -- Look, I think one of the ways that you can see this coming out in like real-world data is not only in our maintenance CapEx, CapEx, which we disclosed. But look at our performance. I mean, Five9, you can't operate to that level of uptime with shotty equipment. You just can't do it. So you can look at that data point as well to see, okay, how well is Equinix doing on a performance basis operating the chips they're operating?

No, no, please. So I mentioned really briefly, but -- so BofA forecast about 100 gigawatt shortfall and power deliver over the next 4 or 5 years relative to power demand in North America alone, right, talk to customers about this as well. So Stu, how are you derisking that? How is Equinix derisking that? And how much contracted power does Equinix have relative to the $5 billion to $7 billion development spend?

Yes. So what we disclosed is we've got -- we own and control 3 gigawatts of design and power today, roughly 600 megawatts of that is under current development, some sort of development, whether that's earthwork all the way to vertical construction and finishing it, and we should have about 1 gigawatt under production by the beginning of next year. And so we're not -- you guys may have heard the term [indiscernible] before. We're not in the business of announcing projects before we have control.

You're in queue to be in queue to get power by...

[indiscernible].

The balance, like, let's say, beginning of '27, we have a gigawatt under production. The remaining 2, we feel very, very good about, right? We have very good line of sight into the power permits for that capacity.

Okay. Can we shift, I guess, more away from maybe some of the risks or hurdles that are being talked about to the opportunities for growth? And I think shift from model building AI inference, something we all talk about a lot. And that kind of goes back to the risk of Anthropic and OpenAI and what their credit rating is, but you're not playing there. You're not building data centers in frontier markets, retail colocation for the most part. So the idea is that you would benefit more from AI inference as enterprises begin to deploy that. And we forecast inference going from, say, 25% demand to 45% in the next 2 years. Are you seeing proof points and having conversations that support that forecast? And can you talk about how much of the demand that you have been reporting on bookings or other metrics has been AI interference related versus more traditional.

Yes. I'll start and pass it to you. If you had asked me this question a year ago, it would be a different answer. We were talking -- we've talked about this in a couple of other meetings. A little over a year ago, we had a client advisory board, and I posed the question to the customers, and we're talking big financial customers and Fortune 500. And I asked them where they are on that journey in their AI journey. And 95% of them in the room said, we haven't started. We thought about it. There was one who was a tech company that was on that journey. And we met with them again a year later. I asked the same question and every single one of them is -- has started. They're in different places. I would say one comment that was made was training wheels, but they absolutely have been forced into the accelerated implementation of an AI strategy. And an AI strategy is it's not just the infrastructure. It's an entirely -- an entire shift in a business operating model in some cases for a company. So it's a multiyear change that's happening. So now you ask me that question, and I don't know that I have an opinion on the exact number, but we are absolutely seeing an increase from our customers in terms of what they're buying. We had seen the majority of our business still from a perspective is non-AI related. Customers have been on that multi-cloud journey. They are building out digital infrastructure. Now they're building out that digital infrastructure to accommodate AI. And we're starting to see them utilize AI much more, the inference component of it.

And we're still doing business with those AI participants as well, right? I mean I think Neoclouds, I think everyone is concerned about creditworthiness of Neoclouds. Look, Neoclouds are a very important part of the AI ecosystem. They tend to be pretty magnetic for our other customers, right? They want to be close to them and get that inferencing out to the eyeballs, which is best done at Equinix, highly interconnected data centers. And so it's still -- it's an important vertical for us. But that said, it's only like 1.5% of our revenues, I think, today. So it's not like we've overlevered or over rotated to that vertical.

And are you projecting greater cross-connect density over time as well if you have more AI inference? Because I mean, in theory, that require a lot more cross-connects.

Yes, no doubt. I mean companies have their data in many different places.

So presumably returns go higher than even the 20% plus over time if you get higher cross-connect growth.

That's an historic metric, that 20% [indiscernible] Yes, that's what we're currently underwriting to right now.

And then...

I think the great way to think about it is that there is some upside in interconnection.

I think we also think we are very bullish on our interconnection product and feel that not only because we have these incredibly dense ecosystems where you have so many customers that live within it, along with their -- the many -- the providers, the clouds, the network service providers, et cetera, but we have over 522,000 interconnections globally. And we start to see with -- to your point, on inference, that creates a much greater demand for interconnection as you have data residing in different places that ultimately needs to get placed where the user is.

It needs to talk to one...

It needs to talk to one another to have any sort of value.

[indiscernible].

Exactly.

So customers can access their data.

Yes, exactly. And so we think that we are in this very unique position based on our scope and the customers that we have to see significant growth come out of it kind of as Adaire would say, our CEO would say, we were built for this moment. So we see that continue to increase. And we look at that as an incredibly important potential driver more so than space and power in the future if you think about it.

Much higher margin. And Al, you're in the front line of customer discussions, negotiations. And what I hear from large hyperscalers in the model is that negotiating power has really shifted to the developer, right, given the tight supply market, things like a sign end of lease and absorbing more of the cost increase, not your model. But same dynamic, are you seeing a shift in negotiation, whether it allows you to put different provisions in the contract you weren't able to before, change in duration, escalators? Is there any change in those negotiations or contracts given the environment?

I think, no, we're not -- we haven't started to see that shift yet. Again, I think it also depends on where you are in the customer segmentation, right? So if you are a very key service provider at the high end of the market, we're going to have a different conversation with you, and we're going to have different contracting terms with you than we are with a midsized customer. I don't know if you're seeing anything [indiscernible].

We've focused on a lot of volume. You've seen our volume increases, as Ryan mentioned in some of our conversations earlier. And we're looking at at the exact level what we're going to do about pricing -- but for sure, there are market dynamics that would be in our favor from a pricing perspective.

It's kind of same line of thought. I mean, I think like 30% of your cabinets were presold quarter, right? So you were preselling a lot. Does the supply-demand dynamic shift your strategy to preselling, meaning an inflationary environment, maybe you resell less because you have a higher rate in the future, they shift your approach to that?

One, it helps us figure out what that demand looks like. Like Arquelle mentioned, we're not just -- we are building data centers, but we're not just doing it without any customer data. We're doing it with a lot of customer data based on, again, 11,000-plus customers, and we've been in these markets for a very long time. But preselling allows us to get even a better window into what that demand shaping should look like. We're not going to presell an entire data center out to 2 or 3 customers or even 10 customers because we do want to do a mix of that to make sure that we're optimizing the cash-on-cash return for our shareholders.

When we build a data center, we actually -- as the shovels go on the ground, talk about, we've actually built into the business case which customers we're going to put in there and not necessarily customer by name, but a small, medium and large foot we call a large customer that's like a mega plus customer. And that's built so that we can get that rate of return on that asset. So when we then build it and then fill it, we are very careful. We look at asset management very closely to make sure that, one, we're addressing that we've got the right mix of customers to give us that rate of return. Two, the ecosystem is so critically important that we're going to actually -- that we've had conversations with customers about we're going to place you somewhere else that's going to meet your needs, but not there because in this ecosystem, we need this set of customers because that's the value that our other customers are asking for. So we manage all of that through asset management. And it becomes really important in terms of when you're looking at how much you're going to presell, that's also understood because we first of all, we're not going to, as you said, sell out an entire data center to one large footprint or a couple of large footprints because it would rule in the model that we have to have the rate of return.

The gravitational pool of having those core anchor tenants.

Exactly. But we also look at -- we want to save some of that because there's opportunity associated with that as well. So it goes into the model.

And Stu, kind of back kind of talking the maintenance CapEx a little bit, different direction, though. The power density with AI, presumably the same thing with AI inference. Can you talk about how your data centers are future-proof if they are right? The power density you can support and what is required to upgrade the data center to support AI?

Yes. Look, we have a very large portfolio of assets anywhere from pretty low density assets that are highly interconnected that are basically network, just network nodes, right, that don't require a ton of power density to some of our newest assets, Arquelle is reminding me like.

Those are more of a [indiscernible].

We really are sort of the core sticky pieces of the public Internet in a lot of ways, right? And that does not need a ton of density. And then you get all the way up to some of these big AI workloads that might require a lot. Arquelle is reminding me of one of our latest builds in the U.S. is up to 18 kW per cabinet, right? And that's the average, right? So we're pre plumbing all of our data centers for...

That would have been maybe 5 or 10...

I mean when I joined the company, it was like 3 and before, we were like, oh, it's crazy. [indiscernible].

And it accelerated very quickly in a short time.

Yes, the delta between 5, 6 [indiscernible] in the last 6 or 7 years, and now we're up to 18 and growing. But a lot of times, we are also like leaving some generator spots as well, so we can even increase the density above that, making sure that we've contracted enough power that we can stay with that growth, again, pre-plumbing the data center for liquid cooling to [indiscernible], all of that stuff. Look, we -- again, it's not easy for us to like just say, okay, it's going to be 40 kW rack for the single customer because that's the spec and that's the chips that they're bringing in to deploy. We have to have our ops team [indiscernible] in sometimes hundreds of customers into these very large data centers. And so it's a real technical challenge for them, but one that we do very, very well. Our stabilization rates, I think, across the portfolio are like 88%, something like that stabilized. So we do a really good job to win those spaces up.

Okay. That's great color. And, I want to come back to just sales and support in general. And I think perception from the outside with Equinix has been superior for so long that maybe focus as much on sales and support as competitors because customers felt they had to be an Equinix facility. might be wrong, that perception. And our perception is the priority of Adaire has been to rework and improve sales and support maybe to address that perception that sales and support maybe didn't match some of the competitors. I guess the real question is, what have you done in the last 12 months to change sales and support and to improve that function?

Okay. Well, I think what -- so I think that we are -- have been on a journey for a number of years in terms of refining our go-to-market and making sure that we've got the right teams aligned around the right segments, which was really the focus over probably the last 4 or 5 years. to assure that we've got the right team. We've got channel partners who we work in conjunction with and making sure that we're selling the assets that we want to. So there's been a huge effort on not only selling our space and power and interconnection, but also in building out the ecosystem and creating that. So that doesn't just happen on its own either. We have a team that actually focused our business development team that's focused on making sure that we're bringing the right partners into those ecosystems to serve our customers. Where it's gone now is as we continue to refine that go-to-market strategy, it's moving into a more, I think, tight verticalization model. So we continue the segmentation, continue to -- and the reason segmentation is important is, one, how you serve the customer. So you're at the high end, a significant customer at the high end of that, what we call a pyramid. They have a higher touch from us. They have -- there's a greater set of resources around them. They're more complex. They're typically global in nature. So you're going to resource them differently. As you go down market, you don't need that level of touch points with them. And you're going to have different ways of serving them that might be serviced through channel. It might be serviced through a team that has -- because the sale is a shorter sales cycle, it moves faster. It's less complex. The refinement that's happening now is to actually take us into a verticalization. So we have these ecosystems -- we've had a network service provider vertical for many years, same thing with financial services. But to really refine that and say which verticals are driving the most revenue for us are the biggest impact to our business that are going to be the biggest users of AI, for example, in inference and to create verticalization, which enables people, someone to use financial services as an example. When you've got a team of people, whether that's your frontline salesperson, your technical resource that's doing your design work, your marketing organization, your product team, and they understand the nuances of financial services and what's happening, they're going to be able to better serve you versus an account team that has financial services, manufacturing, gaming and public sector and really isn't -- doesn't have deep knowledge to be able to become a trusted adviser to that customer. And that's the next piece of what's happening from a go-market perspective. Also from a sales perspective -- or servicing perspective, our servicing model was one that was focused more on responsiveness to the customer, responding when the customer needed us. And so the service model is changing so that we're providing services based on what customers are asking for. Not every customer needs certain levels of service. And also what's the -- there are certain sectors where you might actually charge for some of the additional services that customers are asking for from a support perspective. You're also looking at there is -- there are a team of people that are having a relationship with a customer on an ongoing basis for the life cycle of a contract, there's an incredible amount of opportunity there for them to work with that customer to upsell, to renew contracts to reduce churn. So really utilizing all of the individuals who touch a customer in a more proactive way with a clear role definition that supports increased customer service, but also revenue acceleration.

That was a great answer and a very quick 35 minutes, guys. Stu, Arquelle, thank you so much. Thank you all for coming out.

Thank you.

Thank you.