Welcome, everybody, to our 2:00 session. I'm Jon Atkin with RBC, and I lead the investment research effort in communications infrastructure. Pleased to have the CFO of Equinix, Olivier Leonetti, joining us. Welcome.
Thank you. Nice to meet all of you or see some of you.
A couple of quarters in the seat. I guess I'm just kind of interested in what are some of your assessment in terms of what you've inherited and the initiatives going forward and the playbook that you're executing against. Any kind of mid-year update as we head into the fourth quarter?
Yeah. We said it before, our CEO has said it, this is our moment. If you look at today, the drive for AI inference, what you need is low latency. You need a global presence. You need sovereignty. You need neutrality. You need a dense network of players in your data centers. You need amazing interconnection solutions, and our company is able to offer all of that, significantly more than what is present in the market. This is our moment. Actually, the team undersold the value proposition of the company. We're very excited. You saw we raised our guide lately, three points more growth on top line and bottom line. We believe the best is to come.
Great. Let's maybe dive into Demand and AI and non-AI demand drivers. What does the pipeline look like today compared to before you started a year ago, let's say? Any trends to call out?
Q2, we announced total sales velocity of about 30%. The demand is strong, and today we could, but it's not a smart thing to do, have a higher order book and discuss that all this morning. We don't want to do that because we want to nurture the best ecosystem for long-term value for the assets we deploy. Demand is strong. We are now pushing back on some level of demands to push a diverse set of customers being present on our platform. Very strong.
Maybe dive into the cloud drivers, cloud connectivity, AI training, AI inference, traditional enterprise. Any kind of quick bullet points on where you're seeing particular strength or expect growth going forward?
The strength is impacting many of our customer set. Enterprise, neocloud, cloud providers, large language model, all of them are interested by joining our network. But the key theme behind this is the need to do agentic AI. I go back to what I discussed earlier. Agentic AI, you need what we offer. That's why people want to join now.
The other question is, on this AI agentic trend, all of you are as informed as I am. We are just at the very start of this. We believe that in the next three years, AI agentic will be 50% of the AI demand. We think that AI agentic in the next three years will be 30%-40% of the data center activity. We're just at the start, and this start is driving the demand we are having at Equinix.
On the Cross Connect side, you put up quite strong numbers in the last reported quarter. To what do you attribute that?
To the same theme, right? Why would you join us? You do not join us to have access to compute. You join us to have access to an ecosystem of participants. You want to be free to move from one AI model to another. You want to be free to move from one cloud provider to another. You join the network for that. You want to have an open infrastructure in the new world.
That is what we offer, and that translates into very strong interconnection business. Again, we think that more is to come. We have announced about a month ago a new offering. It is called Fabric One, which allows our participant to connect to actors in the ecosystem in seconds, when before it used to take about a month to connect. So interconnection deal, and that is why people join our network.
The thing that the business has seen episodically is, as you go from that next quantum of optronics to bigger pipes, and this is for Cross Connects inside the data center. There has been circuit grooming and a little bit of churn. Maybe give us your thoughts on whether you are contemplating more of a usage base. That is usage-based, in fact, and sometimes usage can increase, but with optronics, the physical cost connect count can actually go the other way. If you pivot the pricing model, what are your thoughts on avoiding that through usage-based types of pricing mechanisms?
We haven't discussed how we're going to price. We're too low, and it's a bundled offering. You come to us to have access to space, power, ecosystem, and then interconnection. It's a bundled offering. It's very difficult to dissociate interconnection from the rest of the business. There are a certain number of things we know though, where that will drive interconnection revenue faster than the base. That we know. I will leave it there. We are going through a few models in term of pricing, but we think those offerings will be priced higher than the prior one because we offer more value.
As we think about open source models, open weight models, there's been a lot of developments on that in just the last 6 weeks or so during the summer. Just philosophically, is that a development that you see affecting your business one way or the other, whether it's frontier models or open source and open weight?
The answer is yes, what is the logic behind it, right? If you are closed, you establish a relationship, and this relationship is going to last. That's the closed model. In an open model, you want to be open because you want to have choices. If you want to have choices, you want to interconnect more. If you want to interconnect more, you want to be in a place where you have a rich ecosystem and the ability to connect easily with the participants. Open is better for Equinix, no question. The trend is going to open for all the reasons you know. One large language model is better one week is not better the second week, right? You want to be able to change. You have geopolitical dimensions, right? Geopolitic drive, sovereignty.
You want to be able to manage your load, and that's an open model, and that's a better thing for our company.
Want to hit on power and grid. Your typical new build is 50 MW under the Build Bolder program, and how do you navigate utility queues when competing against half a gig or even gigawatt scale projects for the same transmission infrastructure?
Yeah. I feel bad about answering this question because it could look like we have turned deaf, right? We have been able to deploy data centers at an accelerated speed. When you look at the press in our country, in the U.S. or in Europe, a lot of sentiment against data centers. They are not impacting us that much. Why is this?
One, to your point, data centers which are much smaller. Two, we have those data centers in communities that are known to us and in communities that are agent know, right? NTT, they know us. Demand knows us. The general contractors know us. We are able today, because of this connection we have been able to establish with the community and the lower footprint and the good footprint, we have been able today to deploy capital at an accelerated rate. Let me give you two numbers to finish on the point. This year, we deploy 40% more power than we anticipated at the start of the year. 40%. In 2026, we deliver twice the power we delivered in 2025. Why? Because what I've explained.
I guess I'm interested in kind of a supply chain and challenges that you're encountering. I think you're successfully overcoming many of them and bucking the trend by delivering faster rather than slower under some of these initiatives. But given especially your prior tenure, pre-Equinix, thoughts on construction design and ultimately delivery of capacity and how you see that changing.
Yeah. I spend a bit of inside baseball here with Saidane Hafid . Hafid is our Head Operations. He has been running this part of the business for two decades at Equinix. I asked the question, why is it this way? He said we have long-term, same theme. Long-term relationship with our general contractor and power management players. When we place an order, we have certainty. When we say we start a construction at that site, we have certainty. We have also placed long-term orders with those players. Today, we have been able to navigate this part of the equation very well as well.
Maybe just go across regions because you do report regional metrics.
Yeah
APAC, EMEA, Americas, and any kind of highlights to point out in terms of what's driven your growth year to date, and then opportunities that you see going forward.
This is a good market. A participant in our ecosystem will be present in 60% of the cases in a global network, right? People join us to have access to a global reach. Today, we see the U.S. growing a bit faster than the rest of the world, not because demand is lower, but because also it's managing the P&L, the best economics in the U.S., and also us having more access to space and power here. But the demand is strong in the three regions, with a small benefit to the U.S.
M&A has been somewhat muted, but there is atNorth in the Nordics.
Yeah.
How do you see the pipeline, and then how do you see growth across regions in some of your either recently entered or recently acquired markets?
A bit of numbers. Deploy between now and the end of 2029, 1 GW of power. After that, we will have 2 GW of land and power available to us. What I am saying is that we have access to land and power today to really support the growth. Point number one. Point number two, during to do acquisition of land and power, two on an opportunistic basis, atNorth was one of them. You can rest, where is the other position we are going to invest M&A is in technology. We have not done that much in our history. We want to acquire assets, technology asset, tuck-in, not large, to complement our interconnection offering and more to come on this. We want to spend time on M&A in that space.
If there are questions, by the way, make yourself visible. I think we will have some time. xScale, so off balance sheet, you are using financial partners across multiple regions and what is the growth path going forward and anything around M&A on that side as well?
Yeah. xScale is part of the algorithm. We want to manage our portfolio for maximum return, right? The maximum return today for Equinix is to do what we know how to do best, which is retail, colo, small footprint, rich ecosystem, and very high return. We're going to focus on this. If you look at xScale, it's an interesting part of the portfolio. Now it's starting to be a crowded portfolio. You have many players, the footprint is different. You sell a kilowatt, so it's very economic. xScale is part of the portfolio. Retail is the focus.
I guess I'm just interested back in your core business, how you see the demand pipeline. How resilient is it to economic disruption? Perhaps hyperscale CapEx sees some slowing at some point. I don't see it personally.
Right
What are some of the risk factors to be mindful of or things that you're monitoring to assess the resiliency of the growth in your core business?
Just spending a bit more time on the likelihood of the scenario, right? You said that there was no indication that there is a slowdown, right? CapEx, hyperscaler, neocloud this year is 80% higher than the prior year, and we see no sign of demand slowing down or investment slowing down. Again, I go back, AI inference is at the start, and it's where we play, right? Slowdown, not yet. Likely, I don't think it is. But again, then which part of the market will slow down? Is it large language model training centers slowing down, or is it inference? It's likely to be training versus inference. We have a flexible P&L. We can delay CapEx. Our leverage is relatively low, so the debt burden is low in term of if we were to have adverse market events.
We should be able to react better than other players because of ability to move CapEx and ability to have low debt and ability to manage OpEx as well. But again, no sign of that scenario today.
Finally, maybe just a two-parter. One is on capital allocation and balance sheet and kind of financing tools that you have at your disposal to finance the development pipeline and then the growth algorithm around earnings. You gave kind of a multiyear guide as you referenced. There's a lot of drivers around operating efficiency. There's top-line growth. There's equity issuance, refinancing. So difficult question to answer in two minutes.
I'll try to do that, right.
Yeah.
Again, we increased our growth profile to 2029 by three points, on the top line and bottom line. I want to clarify this. We had interesting conversations earlier this morning. None of that is due to the additional CapEx we said we will deploy. This additional CapEx is going to have an impact on the P&L of the organization as of 2029. If anything, this additional CapEx is a drag because you build a site, takes you two years.
You fill it takes three years. So before we have full power of the economics of the site, you have a drag. So despite the drag, we have been able to generate better economics, right? The demand drivers are strong. That would be one. How are we going to finance the growth? We prefer debt. We have a low debt leverage today, 3.5% on a net basis.
We have room to increase it. We want to be investment grade, of course, and we think we can do that and support the growth. We are not looking at issuing equity. We do not think we need to do that. That is not a great return source of financing. Now, vehicles. Those are looking at innovative vehicle to finance the growth. We do that as well. If we do that and see that for colo retail balance sheet, if we do that.
Let us call for questions. We do have one.
[Your data centers or in]
[inaudible]
Like where does inference take place?
[No. I didn't.]
[Inference.]
[Honest.]
Inference.
Yeah.
Like where does it take place? Is it in your data centers or is it in the cloud?
I don't understand what the question. Could you clarify? Yeah. Go back. Yeah. Yeah, clarify.
Like, you have the inference and then you have the different models, and then the inference has to get to decide which models. I guess I'm wondering like where is the-
The orchestration going on is, yeah.
Where exactly does it sit?
It will happen really more in our data centers. The orchestration would come in our data centers, and you see all our data center today increasing the power density to manage those edge compute offering. You want to do that for economic reasons, latency reasons. Yeah.
Seeing that on-
Today. By the way, you see today in the workload, and we use cabinets as a unit of measure. You see that the power deployed by a cabinet has increased like 60% in the recent past. If you go to some of the data centers you have at Equinix, some of them have the latest technology with very powerful rack to do exactly what you said.
Great. That was the last question. Appreciate your taking the time.
Thank you very much. Thank you. Have a good day.