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

Okay. Right. Good afternoon, everybody. I hope the conference is going well. My great pleasure to welcome Keith back from Equinix. Great to have you here, Keith. I think you have a safe harbor statement.

Yes, I do. We might be making some non -- some forward-looking statements. And to the extent that we do, please refer to our risks and uncertainties in our SEC filings.

Great. And for Morgan Stanley, morganstanley.com/researchdisclosures. If you have any questions, reach out to your MS representative. So it's -- the conference theme is like it was last year. I think AI and generative AI and certainly, data centers have been at the center of a lot of those discussions. It feels like you've done a lot since we sat down this time last year. But you had your Investor Day in June. You put out your guidance just a few weeks ago. So perhaps just talk us through what you see in the kind of the 2024 and then beyond that.

Well, yes, so thank you for sort of setting the stage or setting the table. We just gave our guide on February 14 for 2024. Coming off the Analyst Day, continue to feel very good about the business, albeit we talked about some crosscurrents or Charles referred to some crosscurrents in the business. But notwithstanding that, which I'm happy to talk about, the theme for me would be that demand feels good, although there can be troughs. Lines aren't always straight as we know in business. And the supply feels like it's going to get more constrained, and therefore, the pricing environment feels very, very solid. That coupled with where we are in our evolution, we're inflecting on our cost model. And so simultaneous with continuing to grow the business, we can operate the business more efficiently. I foresee that as something that you should continue to expect on a go-forward basis. And hence, the return that we can deliver on a per share basis -- cash flow on a per share basis or, in our case, AFFO per share, I think continue to be very, very healthy. With one thing that's somewhat out of our control but something we certainly manage against very tightly is currency. I think currency, again, on the first guide, you saw the benefit of a small move on the U.S. dollar. And I think as the U.S. dollar continues to trade, I think, more downwards, you're going to see further inflection.

What are you, 60% non-U.S. at this point?

Yes, we're probably about 60% on U.S. And so as I said at Analyst Day, a 5% move can have a dramatic impact on AFFO per share, and to suggest that 5% weakening of the U.S. dollar, 5% relative to some of these currencies, maybe it's the cleanest shirt in the dirty laundry, as they say, but it's still -- it feels like the -- I think with the U.S. eventually going to see rates coming down, I think you're going to see a move in the dollar accordingly, and that's going to benefit us.

Great. Well, maybe we just pick into the guide a little bit. You had record xScale bookings, 90 megawatts, I think. And that spoke to some of the AI demand. But then you still have that elevated churn. So is it very, very different between hyperscale demand and enterprise demand or...

Probably a little bit like the equity markets, right, Magnificent 7 to some degree. Look, the skill demand is very real. And as we said, we sold about 90 megawatts, $1.9 billion of -- $1.8 billion of contract value, 15-year term. The hyperscalers that you know, that you would expect are consuming that. We're about close to 85% utilized of what we've announced and what we've built. If you add to that what we're negotiating right now, we're probably north of 95% sold out. And we have a commitment to put up roughly 800 megawatts, and we're at 330 right now. And so we have a runway -- a nice runway ahead of us.

Concentrated in Europe and Asia. Yes.

I think -- and to some degree, the Americas, so Brazil, Mexico and the like, Canada. So I think there's an opportunity. I think the opportunity can mean much greater. But we're going to continue to show, I think, some level of discipline around how big that -- we want that to be. But I think as Charles alluded to previously, I think that we were historically not going to focus on North America. I think actually North America has applicability now, and so that's going to be an opportunity. And then as it relates more to the core business, yes, we saw a little bit of -- we've seen churn, as you alluded to. But churn to me is more -- I feel like we're a little bit of a lagging indicator. So when you saw a lot of those variable price, variable consume businesses that we're -- they saw a sort of a downturn in Q1, Q2 of last year. For us, I think we're just a bit of a lag. And so with the market continuing to pinch on a number of customers or pinch just generally speaking, you're certainly making -- you have the winners and you have the losers. I mean we saw that in the earnings call. There was a fair number of companies, I shouldn't say customers because they're not all customers, that saw they're disappointed on the guide. And for us, I think that just -- it presented itself in the last sort of 1.5 quarters. And so we're -- we anticipate that's going to continue until we hit the trough. I think you're going to see an inflection in the second half of this year, notwithstanding what maybe some of the macro conditions will do, but it feels like we're in a good spot. We -- I felt it was a good guide knowing what we saw and although disappointing in some respects, 7% to 8% versus our 8% to 10%. But I'd rather be -- go out conservative than go out aggressive because then we're just chasing it. And we knew that we're doing other things to create value for our shareholders, and that was managing the cost model differently. And that's why you saw -- you'll see a 160 basis point improvement on our margin line. And I think that there's plenty of runway ahead of us for that. So it's a combination -- if I just sort of finish off my thought, we want to be as transparent as we think. We don't need to go chase something. I'd rather delight you later than disappoint you. And so our view is always let's do what we need to do for the business. The market knows that it is a tough market and if anybody sees it differently other than those that are investing in the Magnificent 7 and that had a hell of a run. But at some point, you got to step back and say, well, that's not a space that I'm going to play in that much. We're going to play in it, but we have to be careful that we're not putting all our eggs in that basket. And it's akin to what happened with the cloud 10, 12 years ago. It feels like a little bit of a Groundhog Day that everybody wanted us to go chase the big cloud opportunities, but that wasn't where we wanted to be. We wanted to be the place where once the clouds were established and the customers were consuming, they had to bring it back to a place that the data got aggregated and distributed and consumed by the consumer or the enterprise. And that's the major metros. And that will hold true for AI too, that, eventually, whatever happens in the large training models and where they happen to be and these big deals that get done by whomever, that's going to -- we're going to be the beneficiary of all that. But it just takes time to root it.

Right. And on the enterprise behavior, I think there was a comment on the call something to the effect of that IT budgets were getting squeezed a little bit, that CIOs had to make room for AI spend and that they would look at their overall IT spend a little bit more critically because the Board is not going to give them an extra 10% or 20% just to -- for kicks.

Right. One of your peers at another bank told the story at one of our events. He said, when I talk to our internal team and they said that our IT spend is being cut by 10%, the data center spend is going to increase by 20%, and it just gives you a sense that this is -- you got to find ways to create capacity. And whether it's cutting stuff that got you to here but won't get to where you need to go, I think that holds true. And Equinix, too. I mean part of it -- we're stepping back and looking in the mirror and saying, well, what are we doing differently and how should we sort of spend our money. And our vendors and suppliers and partners know full well that the days of enjoying too much profit on the heels of Equinix, those days are over. And that's why we've invested in a procurement and strategic sourcing team, and we have energy teams that focus on sourcing, hedging and renewables. So we're building these teams to create advantages for ourselves and hopefully put us in a better spot relative to anybody else in the market -- in our market. I think -- and that holds true for our sustainability and renewables team under Katrina. These are investments we've been making over the years, and you're now starting to see the fruits of our labor. And it was hard because I know that, certainly, as an investor, you wanted to see more profit, but we also knew we had to invest for our future, and we've done that. Now we're benefiting from it.

Great. And you brought up the margin expansion point earlier, which I think was a positive surprise for the Street, and the people have given up waiting for 50% margins down the road. But can you just unpack the -- what the elements of that sequential improvement are on margins?

Well, there's a number of things, but it's just -- it's still early days, Simon. The -- but the things we did and you heard us talk about it, we needed to look at our human capital differently. We've got different initiatives underway. One of them is Project Home. I've spoken to some of you in the room today. Project Home is really about an initiative to source your human capital in a more cost-efficient market, Toronto, Dallas, Singapore, Manila, Bangalore, Bogota and Warsaw, so 7 markets. And we think we can create 50 to 100 basis points of improvement just on that 1 project alone. Now it takes time. It's not -- it's what we call -- it's not a lift and shift. It's a [ cap ] in a trip. So as people trade out of the business, we're moving their -- the workforce to a different part of the world. So that's Phase 1, going after our corporate real estate. We just don't need as much corporate real estate. We closed 2 offices in the fourth quarter, as you're aware. We're being more judicious on our head count planning. We're going after optimization and our product. And I would tell you that there's a list, we hired a transformation officer in Nicole Collins. Some of you might have met her, and she's awesome, but we're going to go after those 3 or 4 strategic projects. And it's really about driving revenue, customer success in our go-to-market engine. And there's things that we have to keep on doing because we've grown the business quite substantially over the last 5 and 6 years, but the systems and processes have to grow. You can't be a $10 billion or $12 billion or $15 billion business on the back of a $5 billion system. And so that's what you'll see us continue to invest in our CRM and our quote-to-cash opportunities.

And what are you seeing on cost inflation? I mean it will come on to your pricing power. But how is that...

Yes. Look, inflation still exists. The levels are much higher. All you have to do is go to a restaurant or your utility bill, you'll see rates are high. And so we pass that on. And you've got to increase your pricing to reflect that. And we think we can continue to inflate, which is good. We're going to pass our higher costs on. I think we can continue to get a nice return even off the higher invested capital. As a business, inflation isn't necessarily all bad depending on where you are. And so we're working hard at driving down our costs, both our capital and operating costs, and simultaneously working to get the revenues that we need. And I think, going back to my earlier comments, in a more constrained environment and whether -- some say why. You can't have a trough and yet be constrained, and I would argue it's just a matter of timing. I think we are going to get into a constrained market largely because of power, supply chain, civil and social activism, higher cost of capital, higher sort of borrowing costs. So the capital markets are tough. A lot of these big deals that are being announced out there, they still have to figure out their funding. So it's going to take time. And we know it's hard to raise capital in this market, notwithstanding what some can do. But I just think it's the way that the model is evolving and the demand that's going to be put on the industry, I think, is going to be substantial. But I also think that's the opportunity that allows you to get outsized returns relative to many other industries or companies out there.

Great. Can we talk a little bit about the pricing opportunity? Your same-store sales have been pretty strong over the last couple of years here. You did have a large increase of power passthrough a year ago. We were sitting down and it just sort of happened here. But I'm sure there were some learnings from that as well. But your contracts are, what, 3, 4 years? And how are the escalators looking, right? So relative to maybe some of the hyperscale firms, you've got a faster cadence here, which may give you a little bit better levers to move pricing up on a more consistent basis.

Yes. The -- look, I think our power pricing, we did a modest reduction this year, about $68 million in revenues related to power costs.

Is that global? Or is that mostly Europe again?

That's global but mainly Europe, mainly Europe. We have a very strong hedging program. Some will like it. Depends on where we are on the upswing or the downswing. But when you hedge into your positions, you're either -- you can be above or below spot, and you have to look at your forward contracts and your ability to contract looking out. I think we do a good job between our regulated and unregulated markets. But it's not just about how you contract with the service -- the utility provider. It's what are you doing to run the business more efficiently. And our objective is to take roughly 4% to 5% efficiency out of the system or put -- be more efficient and take that cost out of the system. And so you get to take that and redeploy that, if you will, in the infrastructure. I think that's another really important thing. It's not -- you can squeeze the lemon from both ends. You can work -- to obviously work to increase cost or pricing where appropriate. But at the same time, if you can run the place more efficiently, the customer benefits from that. And so it's a combination of those 2 things that I think has made a difference to our model. Looking forward, too early to tell, but we're already putting our hedges in place for '25. And again, we have parameters in which we live with, and we think that it holds really well for the customer. And who knows where the markets are going to go over the next sort of 6 to 12 months. But again, I feel really good about what we're doing.

So as we look forward in your guidance and the long-term guidance, is it going to be more balanced between increased cabinets, billing and MRR because a lot of the last year was primarily MRR?

I would -- look, if we do what I think we can do, you're going to see more MRR and more cabinets go out the door. But I also think if we don't -- if we get less cabinets, you're going to see MRR per cabinet go up because it means that more power is being consumed. So a higher dense environment and the like. But look, the expectation at some point, notwithstanding the higher density, more cabinets should be -- for us to deliver against our expectations, expectations we set for ourselves is we sell more units, more cabinet units.

Good. So just to dive into AI a little bit deeper. I think a year ago, you were saying we're really positioned for the inference side of this. And obviously, you've got xScale getting some benefit right now. But you signed the NVIDIA partnership for sort of private AI. So update us a little bit. And I think even NVIDIA themselves on their earnings call, were saying a bigger percentage of the chips were now inference focused than they were training. But what do you see right now and the next year or so?

Well, I see us executing -- there's going to be a multipronged approach, no surprise. The xScale business will be the beneficiary of those large hyperscalers consuming our capacity, not only present but on a go-forward basis. That will sit out in the xScale. We'll reap the benefits on a nonrecurring basis on the revenues from the sales and marketing fee and the development fee. And then eventually, it will turn into our operating assets, and we'll get asset management fees and operating costs. And then the third piece of recurring will come in through AFFO through our 20% equity interest.

And what's the timing of when that really becomes material?

Well, I don't think it ever really becomes material in the sense that I've always said -- sorry, pardon me, I've always said that, on the revenue side, it will be 1% to 2% of our revenues. So I mean -- so 2% on an $8.8 billion business, so let's just say roughly $180 million, and that will grow over time for sure. But it will never be substantial but the value on an AFFO per share basis of 3% to 5%. And so that adds a lot of value to the business. I think there will be promote fees as our partner or partners choose to monetize their equity share. I think we have the opportunity to earn promotes as well. But that -- those are all prenegotiated, and we'll see how it plays out. But overall, I feel really good about that. But that's sort of the AI or the large hyperscale stuff. And it might not just be AI. They'll be doing other things with that capacity. I also think that there's the Lambdas, the CoreWeaves, the Crusoes and others who are coming in and basically being an aggregation or network. They're putting these nodes inside our facilities, and they're doing what they do, but the benefit is they access our environment to do part of it. But that's not where the value is. I think the value is what you alluded to earlier on in the inference. And the generative work and the inference work, where that falls, I don't know, but this private DGX cloud with NVIDIA and the opportunity set that we would have with those customers, as they think a little bit about how do they use AI as a means to an end inside their environment and keep it private. And the way you can do that is no different than cloud. You can have a hybrid cloud, hybrid multi-cloud environment, but you can also have AI, which is very, very private. And I think doing that inside Equinix, we're best suited for that, largely because we have all the networks. We have the cloud onramps with the cable landing stations. And I think we're going to have an AI ecosystem. And that just feels like it's going to present itself not -- we're seeing the green shoots, the early opportunities. Again, working with our friends over at NVIDIA, hopefully, that will present itself much more meaningfully. But what I would say is I still think it's really early. Everybody wants -- they want something to go and hang their hat on. And there's going to be a tremendous number of great use cases. But they're not presenting themselves just yet. I think you need a few more quarters. And you heard one over the week. I forget the name, [ Klarna ] or something. It was a call center. And they talked about how they can reduce their cost to operate and the value that can bring blah, blah, blah. I get that. But how does it present itself inside an environment like Equinix, I think, is going to be very, very healthy.

And maybe just pull back a little bit on -- for people who are not familiar with your announcement with NVIDIA and just how the economics work, who's buying the chips? Who's buying the servers? How do the economics work?

Yes, the beauty is we don't buy the chips or the servers. It's just the customers. So it does serve -- like NVIDIA has a broad set of customers.

A bundled offer to the customer. Yes.

Well, I think that they're working, obviously, with the large hyperscalers and others, but they're also working with very large enterprises. And those enterprises are, well, how do I take what you're offering and put it into a production -- a productive environment.

In a proprietary dataset, yes.

Where would that be? Equinix. They have the networks. They have the clouds. They have the capacity. That's the right place. And so you create an environment where somebody else is paying for it and we're the recipient of it. So NVIDIA sells what they sell. We do what we do and the customer benefits, but they have to ante up.

So you're leasing cabinets.

Just the -- [indiscernible] system.

That brings up the question of power density and liquid cooling. And how do you think about that? I think you'd had a reference in your K about building new IBXs at 2x the power density of the existing one.

That is true. But that's true just because there's an increase in density across the different industry groups. But there's an element of there's use cases where you need liquid cooling. It's just it's so dense that you either have to do immersive cooling or, I mean, you can do, to some degree, probably up to 20 kW per cabinet you can manage within the current environment. But once you get -- you increase beyond that, you have to do immersive cooling technologies, and you basically -- you're immersing the chips and the chipsets because they're getting so hot. And you probably can manage the business. Again, I'm not a technical person, but 50, 60 kW per cabinet, so meaningfully higher, right? Our very first data centers, we're building to 1.75 kW per cabinet. Now we're talking about 60,000 per cabinet in some cases. But we have the ability in 45 markets or roughly 100 data centers to do liquid cooling and then on a go-forward basis, our technology will -- I'm not sure we'll ever build a data center that's wholly liquid cool, but I think there is a scenario where there could be components or data halls that would be -- would have a different technology to make sure that we can sell at a higher density.

Give you the flexibility and all that.

Have the flexibility.

But you can deploy the NVIDIA solution with the tech that you have today.

Yes.

Yes. Great. So I think you talked about sort of late '25 into '26 has really been when you thought inference would scale for Equinix. Is that the right time frame to really be material?

Yes. Okay. One, scale, I'm not sure material. Obviously, that's very subjective. But I think the significance that if it takes root as we all are anticipating, I'm not sure why you wouldn't start to see some of these use cases in a more meaningful way start to show up in the second half of the year but certainly in '25. And it'll be these one-offs, Simon. I mean -- and then it will be early days. And company X will do something, and whether it's a pharma or a financial institution, they're going to go do things. And then others are going to try and emulate them. And that's how ecosystems get created and -- but I don't know if it's widespread yet. And then how AI gets infused into whether it's an Oracle or a ServiceNow or any of these large software or solution-oriented companies, how they come to market with that offering. Look, I think that presents itself maybe differently on maybe they don't give back some of the space that were going to get retooled for other purposes or returned to us, and they start to think about, well, what else can I do inside that environment? And Charles alluded to it, I think on the call where we're starting to see some customers who are moving their technology and already bought, if you will, technology type 2, and they still are living with technology type 1, and they've got to move that out. But they're saying, well, maybe I'm not -- I'm going to collapse that and move it to technology 2, but I might keep the space because I know how constrained it is, particularly in some of the markets around the world and particularly in the assets that are really critical to their offerings.

Yes, that makes sense. And you talked about the U.S. for xScale. So what's the plan there?

Well, I was basically just saying, I mean, we haven't guided you to anything, but don't exclude the U.S. anymore or, for that matter, North America, Canada as well. I think there's opportunity that will present themselves. And we're already doing xScale in Brazil, in Mexico. We have Silicon Valley. There'll be other markets as well. I'm not sure you'd see us though -- and this is just -- maybe this is just Keith speaking. We'd have to talk to Jon Lin about it, but I'm not sure Ashburn's a place that you see us do xScale, where everybody is built around us and we're that sort of hub that everybody is connected to. I'm just not sure we need to play in that space.

But probably off-balance sheet.

Generally, off-balance sheet. It's hard to imagine that we do much on-balance sheet. I mean, I think there's a scenario for at least a temporary period of time that some stuff might be on-balance sheet. But it can -- as I was talking to some of our investors today in our meeting, [ one ], capital is -- it's hard to go get that capital. And think about constraining oneself just to go do these large deals when there's so much else that's out there. Maybe that's naive on my behalf, but I think we want to continue to take advantage of that opportunity. And if we all of a sudden constrained ourselves because we went and did a big deal, and I mean, as you know, there could be very sizable deals out there, our balance sheet can only tolerate so much. And that discipline that we have shown over the last few years allowed us to continue to grow in scale when others were basically selling off their future.

You have to pick your spots.

Yes, I think you got to pick your spot. And again, I feel very comfortable that we have the wherewithal. We certainly have, I think, the business plan, and we shouldn't veer from it. AI is going to be a great opportunity as whatever the next thing is after AI. And I think just having that foundational element of networks and clouds and subsea cable landing stations and the like and eventually satellite, I think we just keep on doing what we do. And if that sometimes means moderating your growth, because you and I have talked about this over the years, we could fill the darn thing up really fast if you want, but it comes at a price point that I'm not sure is good, and we know that supply is going to get more constrained.

And the customer who has connectivity who needs to be there.

And then eventually, you become less relevant. And so you always have to create environments for people to grow in. And so that sometimes means you step back on your growth rate.

And so let's assume inferences is coming in the next 20 -- 18, 24 months. Remind us of why Equinix is so advantaged in that scenario with your global locations. And how do you stand up against competitors if they're looking to find that low latency connectivity?

I think you just answered it. I mean the fact is we are globally diverse. We'll be in 75, 76 markets. We're looking at more. It wouldn't surprise me if we get to 80 markets over the not-too-distant future. We're going to be in 35 to 36, 37 countries. We have the networks. We have the cloud on-ramps. We have the subsea cable landing stations. We have the fabric that connects everything together. And so when you think about a customer and how they want to consume -- aggregate and consume and disperse, I just think we're the best representation of something that is very fluid.

And you have dominant share on the cloud onramps, right?

We have dominant share on onramp.

It's probably a good proxy for them.

But the reason that we have dominant share of the cloud onramps is because we have dominance on the network side and the cable landing side. That's where everybody wants to aggregate to. And that's why the cloud is -- they sit on top of that. And then the customers sit on top of that. Basically, we've given them the entryway to the Internet and the cloud and therefore, the customers to consume it because it goes both ways. It's those who are distributing and those who are receiving.

So you alluded to the power issue a couple of times already. You've got a license. You've got 20 megawatts, I think, in Singapore. A lot of people didn't, have to go to Malaysia. Dublin is problematic right now and whether it's Ashburn or Silicon Valley or whatever. So I think if you can just take the issue more broadly before the Equinix-specific situation, but is this -- is that going to be a real limiting factor? Or can people move to Manassas or move to Ohio or Mississippi or...

Well, I think we can, yes. Simon, I mean, look, some of our investors asked us that question, well, what does it mean when somebody is going to have to go somewhere else to get access to power [indiscernible]. That has held true for many years, as you know, whether it's the Apples or the Facebooks or the Oracles, Googles, whomever. There are a lot of very remote locations, but the customers haven't moved. And they still have to consume, the consumer, the enterprise. And so there'll be some element of that. Yes, we might have to build elsewhere, and we do that. So when Singapore ran, became full and they put basically a limitation on future growth, we immediately went to Johor, Jakarta and Kuala Lumpur. We're looking in Thailand, Philippines, Vietnam.

Are there properties to buy as tuck-ins? Or is it primarily just buying up some land?

Probably a little bit of both to be honest.

But there's no big platforms, anything [ like that ]?

There are no platforms. I mean, Global Switch, but I don't think that's a platform that we would -- there's Global Switch. I'm not sure that would make sense for them or for us. But then there's Hong Kong that's becoming more relevant again. There's a little bit of thawing of relationship between ourselves and China, and I think that will present itself and so Hong Kong because Singapore is all full up if you will. Our house is full there. And so people -- you can go down to Sydney, but that's too far. And so Hong Kong feels very relevant again. We're building our Hong Kong 6 asset up there. And so those are the kind of things you do, but we also can bifurcate a market. Think about London. We're in Docklands, but we moved the market to Slough. In New York, we moved it to Secaucus. There's other things you can do. And the beauty that we have is we take the assets and we put them all together anyway. They sit on top of our fabric. We have, if you will, at least a name, maybe not an application. We have a platform. And that platform allows customers to be residents somewhere and access all of our assets everywhere in the world. And so that's something that others just can't do. And we're not going to let them do it. They can ride their own fiber if they want to.

And what about other realms of the supply chain? We've been hearing like transformers and backup generators and I mean [ you probably have ] development pipeline.

Yes. And so again, having a good sourcing and procurement team has done worlds for this company. And we're planning. We have like London, Slough. We reached capacity. I saw there was a note today from SEGRO at least out in the marketplace. But we bought up a large piece of land that will take us out the next 10 to 15 years. And so we'll develop on that. We'll prepare the land. We'll get the permits. We'll access the power. We'll do our best to make it as sustainable or renewable as possible. And we'll continue to build on those places. And we're looking at least in the major metros as far out as we can that makes on a reasonable basis or a responsible basis with hanging up too much development cost sitting on our balance sheet, land development. And so we're doing that. We're managing out and thinking about it, and we measure it on a per building basis across each market. And we -- the buildings will be defined on whether it's a network-oriented building or a cloud or an AI or whatever it may be, and we're going to map it all out. And the team does a really good job of figuring out what we need, when and where. And there'll be some markets that are just going to be harder to build it. And so you have to decide whether you build or you will go somewhere else. And we're okay with that. I mean it is what it is, but we know we're not competitively disadvantaged. And we have access to all those networks. So wherever that may be, if we're full up in Frankfurt for whatever reason, I'm using that as an example, then what is the next market that we could go to? Warsaw is very relevant. We talked about we sold a full building into one of the large players in Warsaw recently. Madrid is very important, although it's -- Madrid, I should have said broader Spain, but it's got limitations because of water. So there's a number of things that are going on. And I think as long if -- as you're staying ahead of it -- and we have professional teams that just focus on that, that I think create these opportunities. And I can always talk about our competitive advantage. And some of it -- I referred to it as somewhat proprietary. Nothing is really proprietary per se. But by investing in and getting ahead, whether it's our sustainability team, our renewable team, our power procurement or sourcing or hedging teams, these are all value adds along our supply chain that create opportunity for the future. And I just think that despite what others might be doing, I think we've got our model, right, and we just have to keep on investing and think the next step forward. Singapore, we did a lot to win that. I can't tell you exactly -- you made a reference to a number. I can't sort of confirm or deny that. But what I can tell you is we made a very heavy commitment to the Singapore government. And we also are going to invest heavily in Indonesia and Malaysia and work alongside all the different parties so that we can create capacity for our customers in that part of the world. And those are good markets. I'm excited about what Indonesia has in store for us. And we have a really good partner there in Jardines. So things are going well from that end.

And you always have that great slide in there of your cash-on-cash returns and your stabilized businesses. How are you thinking about the incremental returns on the dollars that you're committing today?

Yes, the beauty is that it's not changing. It's not changing. And largely, it doesn't change because we deliver something that's different than most.

So still high 20s. I forgot the last number.

Yes. See, it ranges depending on the phases and the markets are different, but it's 20 to 30 and you have some stuff that's a little bit low at times because you're first phasing or it's a new emerging market. And then we have stuff in the high 30s and low 40s. But I think playing in that ZIP code of 20 to 30 feels really reasonable, and that's before leverage. Then at the corporate level, as you know, we put leverage on the business. I think we're appropriately levered. And even though we carry a lot of cash, that cash is all spoken for. You got to carry a certain amount of cash in the business, given the investments and the growing dividend and the like. And so a lot of what you hear us talking about now is not about cash for this year. We're already planning for '25 and '26. Okay. Well, what do we need to raise next? What debt do we have to refinance? And we're probably going to be a little bit ahead of what we thought. Quite openly, our cost of capital, at least to refinance our debt, is going to be below what we anticipated. So that's good. It gives a little bit of wind at our back. Euro seems like the natural place for us to raise our future debt. Our teams are -- but our teams are really smart. We raised capital in Japan and Switzerland last year, 2023, because we can. We have operating assets of scale and size that we consume in market and/or we've got swaps or other things we've put in place to fully advantages of that opportunity. When we did that JPY 600 million equivalent in March of last year, we basically transferred the part that we didn't need in Japan back to the U.S. at JPY 115, and the yen is trading at JPY 150 right now because of the swap arrangement we put in place. I think it was JPY 112, to be honest. So really good rates. So when I translate that back into dollar, I get value for it. You can't just go and raise a whole bunch of money in Japan 1% or 2% and just move it anywhere in the world because now you've exposed yourself to the currency or the tax implications or the accounting implication. And as a public company, we just have access to these markets. And I think that's something very, very different than most.

Great. Well, maybe just one last one, tying into that. You've done some green bonds here and maybe just a last word on sustainability and where are you seeing the biggest -- are you moving to on-site generation to -- with all these grid issues? Are there sort of -- are you looking at wind, solar, fuel cells? What's the -- nuclear? What's the hot topic here?

I know we're running out of time here. But absolutely, we're looking at all sources of energy. We have teams that just focus on sourcing, so from renewables to nuclear. And I would just say we do on-site generation. In Dublin, there's no air grid that can't deliver the capacity, and so we're using gas turbines, effectively jet engines inside the plant for a hyperscaler. In other -- in the Silicon Valley and New York, we're going to use Bloom Energy's fuel cells. So I think we're setting ourselves up to do the PPAs. We have a commitment to be 100% renewable by 2030. We just signed our last PPA for 1 of the markets that will take us to the 100%. And so I'm excited about that. So we're in a really good spot from an energy sourcing perspective.

Great. Thank you so much, Keith. Appreciate it.

Thank you, Simon. Thank you all.

Bye.