Cool Vector Video-Podcast TRANSCRIPT: "Cool Vector Hot Takes: Microsoft Contracts, Building in Johor, RE Allocations, Blackstone’s Utility"
Cool Vector Video-Podcast: Charting the rise of data centers and the digital infrastructure asset class.
Episode Title: Cool Vector Hot Takes: Microsoft Contracts, Building in Johor, RE Allocations, Blackstone’s Utility
Episode Duration: 26:31
Originally Posted: February 28, 2025
Listen to the Full Episode here.
Speakers: Obinna Isiadinso, International Finance Corp; Phillip Koblence, Cool Vector editorial advisor; Nabeel Mahmood, Cool Vector editorial advisor; MODERATOR: David Snow, Cool Vector
Episode Overview:
TRANSCRIPT
David Snow: You're watching Cool Vector Hot Takes. I'm David Snow, your host, and today we are joined by a panel of experts. Obinna Isiadinso is sector lead for data centers at the International Finance Corp, which is a division of the World Bank. Phillip Koblence is a Cool Vector editorial advisor and a longtime data center operator. And Nabeel Mahmood is also a Cool Vector editorial advisor and a longtime technology executive. Very happy to have everyone here. We're gonna have a great conversation. We're gonna cover four topics:
Has Microsoft been oversupplied data centers? Inside the plan for Southeast Asia's largest data center. Real estate investor love for data centers keeps rising. And finally, Blackstone's low carbon power move in Virginia's data center alley.
Before we begin, Obinna, would love to just get a quick introduction about yourself and the IFC.
Obinna Isiadinso: Thanks so much for having me here and joining with Phil and Nabeel. I lead the data center investments for the IFC in emerging markets globally, both debt and equity financings. I work very closely with our regional colleagues in South America, Africa, Asia, Eastern Europe, and the Middle East. I also work closely with some of our global clients, which include hyperscalers and the global multi-tenant data center operators.
The IFC is the private sector arm of the World Bank Group, so our focus really is investing and lending in private sector companies and infrastructure projects in emerging markets globally. We have about a hundred billion dollar balance sheet. About 70% of what we do is essentially lending to private sector companies and investing in infrastructure projects. And about 30% of what we do involves essentially providing private equity.
Prior to joining the IFC, I spent about 12 years investing in private companies across the Africa region. One of my investments was Eaton Towers, the third largest towers company across Africa, and it was acquired by American Towers in 2019.
David Snow: Let's start with our first topic.
Has Microsoft been oversupplied data centers? Why do we ask? Well, Microsoft, according to a recent report from TD Cowen, has canceled data center leases in the US worth a couple of hundred megawatts. Microsoft has also pulled back on converting SOQs into leases. SOQs are statements of qualification, and they are reallocating international data center investment back to the States. So, Phil, on a spectrum of “so what” over here to “the world is ending,” where do you sit?
Phillip Koblence: I usually sit somewhere between those two places no matter what the topic is. I am recently back from Metro Connect down in Fort Lauderdale, which started Monday, the 24th of February, which is pretty much when the Cowen report kind of landed. You could see everybody reading through it on the airplane on the way down there.
Michael Elias, who drafted that report, was actually at Metro Connect and, as you can imagine, was the talk of town.
It is not necessarily surprising that these companies that are taking very large swings at infrastructure deployments are doing essentially real-time analysis so they can determine where the actual deployments of compute workloads are gonna end up going. That is a kind of shifting pendulum. It used to not make news when Microsoft was deploying one or two or three megs at a time and they would determine that they were gonna aggregate in a certain area or grow cloud infrastructure in one area versus the other.
We just happen to be hyper-focused on these huge swings that they're taking at multi-hundred megawatt facilities being deployed, and some of them reaching certain levels of readiness sooner than others. And as a consequence of that, they are writing into a lot of these contracts statements of qualifications, etc., flexibility that allows them to shift gears.
With respect to hundreds of megawatts at a time, it's something that we have to start being comfortable with digesting as an industry, to not say the sky is falling.
And, you know, AI workloads are a fad and we're not going to grow in digital infrastructure at all. There's a lot of nuance in terms of when availability is gonna come online and what the bottlenecks tend to be. And sometimes those bottlenecks are not obvious until you start down the construction process and then you see where you're actually able to deploy and not, while at the same time the actual technology is evolving so quickly that you're making a determination of where you're gonna need to deploy.
I'm not necessarily so sure that a lot of the folks that are looking at it and saying it has something to do with the quantum computing chip that Microsoft recently announced—I don't think those two things necessarily have anything to do with each other. I just think there's a lot of moving parts, and we're seeing it play out in real time.
David Snow: So it sounds like a bit more of a recalibration than a retreat. Nabeel, what is your take?
Nabeel Mahmood: Well, I was hoping for Phil to say the end of the world is here and Microsoft has figured out the time for it, but that's obviously not the case here. I would concur with Phil's assessment.
It's really more about realignment, recalibration, and Microsoft's still committed to the $80 billion investment for this year. So it's not that they're pulling back, it's just realigning what their strategy is. Part of the challenge that we've had in the sector is this over-design and under-provisioning and under-utilization of resources.
As a matter of fact, the entire lifespan of the technology sector is that we have always over-designed and under-provisioned. There seems to be that continuation in artificial intelligence and the hype thereof that we've created, not knowing exactly what the future holds.
I would not necessarily worry about it again. I mean, this is not the first time this has happened. This is not the last time it's going to happen. This is history. It's repeating itself, like we've seen in other verticals and other sectors. And even with other companies. I mean, Meta for instance is actually notorious about this as well.
I would not necessarily blame Microsoft or make any assumptions that the world is coming to an end. Just recalibration, realignment of resources and capital.
Obinna Isiadinso: I think there was some initial concern around demand. Like what does this mean for demand, especially given what we saw with the reaction to DeepSeek. As Nabeel said, this CapEx plan—$80 billion CapEx plan for 2025—is still on track. CEO Satya Nadella came out and he also confirmed that.
It's important to remember, just from a scale perspective, a couple hundred megawatts—let's say 200 megawatts or so—represents around 2% of the $80 billion of CapEx that the company is projecting to spend.
It's not a massive amount for them to put the brakes on, right? So given the dynamics between Microsoft and OpenAI and the recent announcements of Stargate, it seems that OpenAI is essentially diversifying its infrastructure providers, right? And it’s sort of, to a certain degree, shifting away from Microsoft potentially to other sources, whether it's Stargate or SoftBank.
The hyperscalers have been prioritizing North America and the developed markets from an AI deployment perspective, because that's really where they're seeing the most activity. That's where they're seeing the most demand. That's where they're going to deploy most of this CapEx. The large regional hub markets in emerging markets—those are the markets that have been receiving some CapEx deployment from an AI perspective, and there's only a few of them.
Think about Brazil in South America. Think about India. Think about Malaysia in the Asia region. Those are the only markets where there's a combination of hyperscale, cloud, and AI infrastructure deployments.
And I should note that these projects that were canceled—they're all sort of greenfield, early-stage projects, right? Hyperscaler clients may walk away from sites that are still in the early stages, potentially even pre-development phase of these projects.
This isn't necessarily a red flag. We still see that there's good strong demand behind the AI growth curve.
Phillip Koblence: That being said, I have to imagine that a lot of folks that take an absolute cash win when Microsoft signs off on one of these statement of qualifications, that underwrote against those things, probably wasn't the best news that their projects were canceled. When you're throwing all of these projects in the air at the same time, and you're taking these huge swings, no matter who the hyperscaler is—whether it's Microsoft or Meta, or Google or Oracle or any of these companies—they're growing so fast in so many different regions, there's an unbelievable amount of power that you're just shifting. At this point, the scale and the perspective has changed so greatly that these huge hundred million dollar developments can change overnight.
That being said, I have to imagine that the elements that made them attractive to Microsoft from the beginning will likely eventually make them attractive for some type of digital infrastructure development in the future. It's just that the timing likely didn't align with what Microsoft was looking to develop.
Obinna Isiadinso: We're hearing that this shift is a shift towards self-build. They're walking away, potentially at least internationally, from co-location opportunities to more self-build, which is more profitable. They have more control. Those sites are more attractive. There's power to those sites. So that makes a lot of sense. Right? And then I think on the flip side, sites that they potentially could be walking away from, that could present opportunities for other players.
Nabeel Mahmood: Of course, besides real estate and power, I think it's also the efficiency ratios of utilizing your asset to the maximum that you can. Right? So what we saw with DeepSeek and how they built, how they provisioned, and their efficiency ratios—I would say that's kind of like a representation of Microsoft’s strategy as well. To say, okay, well, we are going to increase our utilization metrics to a higher number, which of course, with Satya's long-term strategy on its carbon neutral world, fits in quite well.
Phillip Koblence: I learned for the first time that Nabeel and Satya are on a first-name basis. It's incredible. Let's get him on next. Welcome to America.
David Snow: Let's do it.
Phillip Koblence: Exactly.
David Snow: Phil, you made a good point that a pencil scratch for Microsoft is not a big deal in the broader scheme of things, but if it's your project that's getting scratched out, then it's existential.
All right, let's move on.
CLIP 2: Inside the plan for Southeast Asia's largest data center
We're gonna go all the way to Johor, Malaysia. Southeast Asia's largest data center has been funded to the tune of $900 million. The project is being led by Yondr Group. Among those providing finance are the IFC, Obinna's organization. So Obinna, if you could give us the quick executive tour of this data center in Johor, what can we learn about what's going on in Southeast Asia and more broadly in the emerging markets through this $900 million project?
Obinna Isiadinso: This was a greenfield hyperscale data center project where Yondr Group was expanding into Malaysia. IFC, we were able to support Yondr by providing pre-contract debt financing to enable Yondr to complete the construction of a 48 megawatt core-and-shell and a foundation for a 98 megawatt campus site. That enabled Yondr Group to not only finish that construction but also in parallel continue their discussions with potential customers for that site.
Actually, two months after we were able to disburse the first $50 million in senior debt financing to Yondr, they were able to secure a contract. This is one of the first instances for IFC to provide pre-contract debt financing in the data center sector.
One of the things that we have seen in the data center sector now globally is there's been an increase in these types of requests from data center operators. Given the competition to lock down these sites, the demands from hyperscale customers to really deploy as quickly as possible, data center operators, developers, investors are really seeking to run their pre-development process in parallel. That can really front-load a lot of the work and help them meet their aggressive timelines.
All of the major US hyperscalers are in Johor: Google, Microsoft, Amazon. There's been about $8 billion of investment announced for Johor. There's good connectivity coming into that market. There are government incentives to attract investment into that market as well. For a number of reasons, we were very comfortable with that market.
Once the contract was signed two months later, we were able to then quickly start the process on unlocking another $900 million of debt financing along with the other lenders in the syndicate for this project.
David Snow: You would support a data center development pre-contract if you had great confidence that they were gonna definitely sign a contract. So I guess that speaks to the measurable demand in Southeast Asia for this kind of compute capacity, right?
Obinna Isiadinso: It does. Southeast Asia is a 600 million population across several markets. The market in aggregate is about 1.5 gigawatts. By 2030, the market is projected to be at least a five to six gigawatt market. Today, Singapore has just reached capacity because data centers represent 13% of electricity demand in that country.
There's significant demand, and the hyperscalers are now actually expanding into that region—Indonesia, Singapore, Malaysia, Thailand, the Philippines. That whole axis is now really starting to see more activity from the hyperscalers.
Phillip Koblence: Like the digital infrastructure equivalent of a spec house. Just on a per capita basis—let's just take the US, a population of what, 330 million people? How many kilowatts or megawatts, gigawatts of actual digital infrastructure capacity do we have in the US in the pipeline? Forget about what's there. And you try to take that corollary across the African subcontinent and Southeast Asia, and all those places—that just kind of shows you what the potential demand profile is. The absorption of technology and innovation in these markets that are essentially greenfield, in the emerging markets, is actually at a higher rate once you start construction, because you don't have the burden of legacy infrastructure.
David Snow: I think there are metrics that you can point to that suggest we're only at the beginning of this process as opposed to the middle or the end.
Nabeel Mahmood: I mean, that's where the future is, really. Singapore is actually at a screeching halt. Like you said, Malaysia, Indonesia, the Philippines are going to be the next areas of focus for a lot of the hyperscalers and large enterprise to start moving over there. And considering the fact that the cost to build out there is a lot less, and you don't have to go through the bureaucratic layers that we've got in primary markets like the U.S., and you've got that interest—whereby people are willing to learn and put the hard time and effort into being in the sector.
David Snow: Let's move on to the third topic. Phil, it seems that you have an advisor joining.
Phillip Koblence: Look, this is the way our lives are now. You're seeing it in real time.
David Snow: This is the future. The future is literally here. Digital infrastructure is enabling this wonderful family time that we're all here enjoying.
Phillip Koblence: I've never been a more engaged father than I am right now, trying to ignore the little girl that's tugging on my shirt right now saying, "Daddy!"
David Snow: Okay, moving on. How do we know? Well, KPMG came out with a survey recently that found that of real estate investors surveyed, 40% of them now consider data centers to be the most attractive asset class within real estate. This is up from 27% believing that data centers were the most attractive asset class just a year ago.
Obviously, what this suggests is that within the bucket of allocations that real estate investors have, they are weighing the benefits of allocating capital to each—whether it's residential, logistics, hospitality.
And so here comes data centers, which, if you define data centers as real estate, they could be pretty attractive. But of course, it's possible to not define data centers as real estate and consider them something else—like infrastructure. In fact, there are investors out there who believe that a data center, when it is up and running as an operational business, should be considered more of a private equity play because you go in and you can create operational efficiencies.
What do you think this means about the appetite of real estate investors for data centers and their understanding of what the opportunity is?
Obinna Isiadinso: What's interesting about data centers is there's always this question around: where should we allocate this? Is this real estate or is it infrastructure?
What we have recognized is that it's a little bit of both. It definitely has some infrastructure and digital infrastructure elements, and it has some real estate elements to it.
On the infrastructure side, the long-term contracts, creditworthy tenants, triple-A rated global hyperscalers. And then on the real estate side, you've got development yields that are sort of in that 6% to 12% range. Typically that's what we see in emerging markets globally.
These assets are trading at attractive multiples. Typically in emerging markets, they're trading anywhere from 12 to 20 times EBITDA. In the private markets, these assets could be trading from 20 to 25 times EBITDA.
Most investors are under-allocated to digital infrastructure. Institutional investors have around a 5% allocation to infrastructure. But we have to remember that digital infrastructure is just one small piece of infrastructure overall.
Our digital infrastructure business is going to be the largest piece of our global infrastructure business for the first time ever. We have increased our digital infrastructure business 10x over the past five years. Our data center business has grown 10x over the past four years, so I suspect that we're going to see similar types of growth in digital infrastructure investments globally.
The funding gap for data centers globally is around a trillion dollars over the next five to seven years. So there will be a need for more capital, and there will be a need for more investors to provide both equity and debt financing. But today it appears that even though there is increased demand, most institutional investors are still under-allocated to the asset class.
David Snow: Phil, are you a real estate guy because you work in data centers?
Phillip Koblence: I mean, look, the way I describe what I do to someone when I meet them is, I'm a landlord and my tenants are computers. So to a certain extent, there's no doubt there's a real estate component.
You have a lot of real estate owners, particularly those who have exposure to a very disrupted commercial real estate market over the course of the last five years, that see the types of returns you can get in data centers.
You have all of the major brokerage shops now with data center advisory practices, and real estate owners, landlords, wanting to get on the gravy train—really without the fundamental understanding of what it takes to operate a piece of digital infrastructure.
Yes, the returns are there, particularly if you have a large piece of industrial real estate that has access to a significant amount of power.
Phillip Koblence: Or perhaps a commercial office property that sits at the crossroads of a significant amount of fiber infrastructure. And you could have some network-dense assets that might be a site for a potential carrier hotel in that particular market. But there's just a lot involved that is kind of outside the scope of traditional real estate when it comes to operating—enabling the CapEx associated with the types of infrastructure—which tend to provide some element of sticker shock to landlords that are used to providing tenant improvement dollars in the area of $100 to $300 a foot, depending on the class of facility in a major city, to the kind of $10 to $15 million a megawatt that you see in terms of what it takes to develop a piece of greenfield digital infrastructure.
Real estate people need to go in eyes wide open.
Nabeel Mahmood: It makes a lot of sense. You've eliminated the human element of it. There aren’t people coming and breaking your toilets. There are machines that are sitting out there and doing what they're supposed to be doing. You can repurpose that property very quickly, very easily.
You compare this asset class of data centers with other asset classes within the real estate portfolio—let's just say warehousing—those customers can pick and shift and move anywhere in the world on the go. You can't just lift and shift a data center client from a major metropolitan city to a remote area. There is a lot that goes on behind the scenes from a technical perspective.
So you are going to have the longevity of a customer. They're going to pay their bills on time. If they're not, you're going to know where they're at. It's sort of like the McDonald's model in a way. I believe McDonald's is not in the business of making burgers—they're in the business of owning real estate.
If it's not already the top-tiered class, it will be very shortly. And this is just the beginning of time, and that means there's going to be more and more data centers. So more and more of these sites are continuing to be built.
David Snow: Excellent. Let's move on to our final topic.
Blackstone's low-carbon power move in Virginia's Data Center Alley
This has already been announced, but this item got our attention because just a couple of days ago, Blackstone Energy Transition Partners IV, which is a fund run by Blackstone, closed at $5.6 billion. Among its investments is a $1 billion investment in a Virginia power utility called Potomac Energy Center—a 774 megawatt natural gas and hydrogen-ready power plant that is going to serve largely Northern Virginia, which has, I think, roughly a fourth of America's compute power.
Nabeel Mahmood: This utility is described as hydrogen-ready. Of course, hydrogen is a low-carbon energy source and so seems like a good deal. What's your take, guys?
Well, whether it be hydrogen, nuclear—I mean, it's irrelevant. Or we can say it's fusion. We have to make these big leaps in finding sources that are scalable, that can address the demand that we've got. Otherwise, we're going to be stuck with what we've got right now.
In theory, it makes a lot of sense. It addresses the broader industry challenge of not having enough power capacities in primary markets like Northern Virginia. I believe that it sets the right tone at this point in time—that we are getting ready for the big wave that's in front of us.
Phillip Koblence: It can't be surprising that a company like Blackstone, that has large-scale digital infrastructure investments through QTS and other vehicles they've had around the world, also wants to control the biggest component of enabling that digital infrastructure at this point, which is the power. And then it becomes almost an arms race of controlling all of the elements of the pipeline that it takes to go from greenfield to powered data center running compute.
Obinna Isiadinso: Blackstone is definitely being strategic in terms of looking at the value chain and identifying ways to unlock growth within the data center segment of the value chain.
KKR and ECP announced a $50 billion joint venture. BlackRock and Microsoft and MGX announced another $50 to $100 billion AI joint venture.
We're going to see a number of these power-related joint ventures with the goal of providing capital to invest in power assets—essentially bringing energy to data centers. Historically, data centers have been power consumers, but over time what we're likely to see is that data centers will be power producers.
As they scale and as they become larger, there's essentially a larger power component within the data center project itself. When we're talking about a couple hundred megawatt projects or a gigawatt project, there's going to be a significant power element.
One aspect of the power element in the data center project would be finding ways to potentially supply power to the grid if there's excess power being produced by that power asset that's embedded in the data center project.
We're going to see this combination of expertise on the data center side and the power side coming together. I will also say that Blackstone also invested in a major construction firm that constructs data centers in Europe. They have been very strategic about looking at the value chain and identifying ways that they can address constraints to unlock growth for the data center sector.
Phillip Koblence: Will there be the same level of scrutiny and regulation that you would expect from a utility that is really more of a consumer-focused, safety-focused asset class than, say, a data center?
With a company like Blackstone acquiring a utility, is there a concern that they then get to pick the winners? If they own a utility and they own QTS, as an example—and I'm not saying this is the case, I'm just saying—will they make power available at the same rate if they own a utility to non-QTS data centers as they would to QTS data centers? And the same with the construction companies in those areas?
When you're talking about digital infrastructure that, as Obinna says, has this value chain around it, where one is completely reliant on the other—whether it's the utility or the construction or the supplying of the various infrastructure that goes into these projects—and the difference between success and failure is access to that inventory in a timely and cost-effective fashion, is there a potential risk that this type of hyper-capitalism could not be the best thing for the evolution of our industry?
David Snow: What this suggests, Phil, is that we need to get Steve Schwarzman on here to ask him those very reasonable questions.
Phillip Koblence: Or as Sile calls him—
David Snow: Stevie.
Obinna Isiadinso: Exactly.
David Snow: Excellent. So Obinna Isiadinso, Phillip Koblence, and Nabeel Mahmood—thank you so much for joining Cool Vector Hot Takes. And of course, we want to hear your hot takes next time we convene.
Phillip Koblence: Thanks, David.
TRANSCRIPT END