Subscribe
Share
Share
Embed
Current Season: First Draft Live
Between economic whiplash, shifting policies and market volatility that changes by the hour, you need industry insights that cut through the noise. That's exactly why we're launching First Draft Live, a new weekly series that breaks down what's happening, why it matters and what you need to know to do better business.
Join us live on Bisnow.com every Friday at 12:30 PM ET / 9:30AM PT for conversations with the industry's sharpest minds discussing the week's most critical stories, or catch the replay right afterwards — here on your podcast app of choice.
Alright. Welcome to First Draft Live. It's Thursday, April 9. I'm Mark Bisnow, editor in chief coming to you live from New York. So for years, data centers are one of the quietest corners of commercial real estate, highly technical, highly specialized, mostly invisible.
Mark Bonner:Well, these days, that's clearly no longer true. Over the past few years, the explosion of artificial intelligence layered on top of surging demand for cloud streaming and digital infrastructure has turned data centers into one of the most aggressive development stories around the globe, and America is the tip of the spear. The scale is enormous. At this very moment, The US data center pipeline is now estimated at roughly $1,700,000,000,000. Buildings that once required 20 or 30 megawatts are now demanding 60 to 100, and entire campuses are pushing into the many hundreds.
Mark Bonner:At the same time, the geography is shifting. For decades, development clustered in Northern Virginia, Silicon Valley, and Dallas. Now roughly 64% of new projects are happening in so called frontier markets, small towns, rural communities. Entire regions are being pulled into the digital economy almost overnight. And that shift is being driven by one constraint above all else, power.
Mark Bonner:But as fast as this build out is happening, something else is happening just as quickly. Resistance. Projects are being canceled. Communities are pushing back. And in some cases, developments are getting blocked after years of planning.
Mark Bonner:So what was once a sleepy asset class has suddenly become a political issue, which raises a bigger question. What happens when one of the fastest growing sectors in real estate collides with the communities it's trying to enter? To help us unpack that question, I'm joined today by Graham Williams, president of Tract and senior managing director at Tract Capital. They're behind some of the largest data center campus developments in America right now. And a quick note, today's conversation builds on Bisnow's Data Center Nation special report.
Mark Bonner:You can currently check it out on biznow.com, which Tract helped make possible through their generous support. However, the story was independently reported. Graham, welcome to First Draft Live.
Graham Williams:Hi, Mark. Happy to be with you.
Mark Bonner:So, Graham, let me start here. We've gone from a relatively niche asset class, believe it or not, ten years ago, most people had not heard of this, to a $1,700,000,000,000 development juggernaut. From where you sit, are we still in the early innings of this build out, or are we already starting to run the real limits, whether that's land, power, or politics?
Graham Williams:Well, Mark, I think we we are definitely in the early innings of the demand for build out. There are about 50% of enterprise workloads in this country have migrated to the cloud, which leaves another 50% yet to go. And AI, I think we're just now trying to wrap our arms around what is possible with both the large language models that we've become accustomed to, but also a whole wave of new, more targeted models. And so the demand, we believe, is durable easily through the next decade. We think we grow from around 40 gigawatts in The United States today to north of 150 gigawatts in 2034, and it really calls into it causes the need to be thoughtful and planful about this new infrastructure.
Graham Williams:These are we'll we'll talk a lot today about the changing nature of of data center development, and it really requires a rethink about how we we position and preposition that infrastructure.
Mark Bonner:So look, we're gonna get into land power and politics a little later, but let's talk about where this growth is actually happening because the data sent the data shows a pretty dramatic shift away from the traditional hubs into some really new markets, I think, would surprise people. Places like Abilene, Texas Jackson, Mississippi, just to name two. And that's largely being driven by access to power. So when you're evaluating a site today, what actually comes first today? Is it power, full stop, or are there still other factors that can override that?
Graham Williams:Well, Mark, as we've looked at United States in particular, we have looked at all of the different attributes that are required for large scale data center development. And as we've parsed the entire country, we found that less than or right about one half of 1% of the land area in The United States is viable for large scale data center development. And that's before you get into zoning constraints. And so that's where we think about proximity to high tension transmission lines, I. E.
Graham Williams:The power, an intersection of fiber, municipal infrastructure like water and sewer and access, and increasingly important proximity to a workforce. These are large campuses that require lots of skilled men and women to build them. And when we're 300,000 electricians short in this country, it's hard to get someone to drive two and a half hours to a job site no matter what it is. And so all of those things mean that there are actually not as many locations for data centers as people might think. It's a very niche universe, and so you have to look for all of those things.
Graham Williams:I think we see a newly minted data center expert born every week, but too many of them are thinking it's just about power or it's just about network or it's just about land. It's about all of those things together. And so I think you you're seeing this migration of of, the cloud platforms to new markets largely in search of that combination, and power clearly is a huge, huge part of that.
Mark Bonner:But, you know, I mentioned Abilene, Texas and Jackson, Mississippi. Now I'm I'm well aware of Jackson, Mississippi. I lived there for two years. These aren't places that are known for their large workforces. These aren't places that are known, frankly, for their their higher ed.
Mark Bonner:They these are places that are known for brain drain. Right? Mississippi loses all of its talented people to other states. Texas, at least in their small towns, maybe to larger markets within Texas. I mean, can it really be true that in Abilene, Texas or Jackson, Mississippi has the workforces to support a data center or a big data center campus as you state?
Graham Williams:Yeah. I think some of these deployments are testing the the the bounds of smaller workforces and are, I don't think anyone would call Abilene unsuccessful. It certainly has been successful, but it's had its challenges, and I think workforce is one of those. I also think it's important to call out the importance of the workloads that each data center is trying to serve because those have different location requirements. And it's not something that, you know, the general public is generally aware of.
Graham Williams:You know, we think of AI today as the big models being trained, and that's really where that's going to places like Abilene and Jackson, in some cases, North Dakota. That's really location agnostic demand. That can go anywhere as long as it can be built and powered. The other but that's ultimately grabbing a lot of headlines, but is gonna be maybe 10% of the overall workloads over the next decade. What's really driving the growth that happens year over year over year is cloud growth and inference growth.
Graham Williams:And AI inferencing is how we interact with the models and all the applications we use. And those two, cloud and inference, are really location specific, and so those generally fall into markets that are NFL markets, if you will. And so we have seen a broadening of that fabric. So if you look at Amazon and Google and Microsoft and Meta, in 2013, they each averaged three regions, three cloud regions: East, West, Central. Today, they're averaging north of 12.
Graham Williams:So you can see that as they grow, they reach a point where they have to find outlets to grow in addition. And so a good example is we had big deployments in Ashburn and in Chicago in the early 2010s, and the clouds realized they needed to spread that risk and find new power. And so they drew a figure eight between those markets. And right in the middle of the figure eight was Columbus, Ohio, which happened to have really good infrastructure and good network. And lo and behold, Columbus became a huge data center hub.
Graham Williams:And it's really because it was what we call splitting the rings between Chicago and and Ashburn.
Mark Bonner:Yeah. Still though, power feels like it is the defining constraint. Right? Utilities are struggling to keep up coast to coast. Interconnection timelines are stretching out.
Mark Bonner:In some cases, we've reported a Bisnow and developers are having to rethink projects entirely because the grid just isn't there. So how constrained are you right now practically speaking? What is what is what does it look like today and where do you see this heading in the next couple of years?
Graham Williams:Well, think we we recognize that there was gonna be a big constraint in both zoned land for data centers at the right scale. And as you mentioned, these are now multiple 100 megawatt campuses instead of individual buildings. So that requires I think the average right now hyperscale campus is 300 acres. Some are getting to be thousands of acres large. Those are hard to find.
Graham Williams:That's one constraint. The other is clearly power. And I mentioned earlier, we think The U. S. Demand is going to grow to 150 gigawatts.
Graham Williams:That doesn't exist today on the grid. That's going to require net new generation, net new transmission infrastructure, and that elongates the predevelopment process. And I think our industry has been, for a lot of reasons, very shortsighted and just trying to keep up with the demand. What can I build in the next three years? And that mentality has led to a lack of thinking about years five to fifteen.
Graham Williams:But when we're talking about gigawatt campuses that need new generation and transmission, need new water infrastructure, need zoning, need fiber builds, all that takes five to seven years. And so you had better start seven years ago if you want to have access to power today. And so we believe that if you're not thinking longer term in a master plan way, we're going to run into even more constraints and we're going to have greater scarcity going forward.
Mark Bonner:That's strategy, but let's talk tactics. Right? Because we're moving in towards a world where data centers increasingly are being asked to solve for their own power. How do you do that?
Graham Williams:Well, there are a number of ways, Mark. I mean, first, we're certainly as we think about site selection and where the industry is going and where the workloads need to be, You have to think about the existing system and how it needs to grow regionally. And then within those kind of regions and markets, you're looking for locations that have existing grid capacity to the extent that there's any left, a utility who's willing to invest into the future and make incremental investments. It's a tough spot for a lot of investor owned utilities because they were really set up to have predictable cash outlays and 9% returns, and their investors put their money down for that profile. When in reality now, we're coming into a big growth cycle, and they're being asked to deploy growth CapEx, and they're not used to taking growth risk and looking for growth returns.
Graham Williams:So we're having a realignment with the utilities. Some are willing to make those investments, others are reticent. We're also looking for states and localities that have policies to facilitate, one, competition in the energy generation side so that you don't just have a single utility that's a monopoly and you have to work through them, And two, an ability to do on-site generation. You know, if we're going to require, gosh, 110 new gigawatts of power, and those are really focused on very specific loads that are concentrated in specific areas, it makes a lot of sense to try to build that as close as possible. So you take out the transmission costs and the transmission risk and policymakers can either make that easy or make that challenging.
Graham Williams:So we think about that certainly. And then last is, you know, there's a growing importance of new types of core infrastructure to enable that on-site generation. So obviously, you know, access to gas pipelines that have capacity in them, thinking about geothermal resources, thinking about pump storage in certain cases, access to water, thinking about policies and localities that are interested in SMR, Small Modular Nuclear Reactors going forward. So all of those things have come into the frame of site selection and important criteria as we move forward. Ultimately, for very large loads who have sensitive workloads and customers, we think they're all gonna wanna be grid connected for resiliency, but we think we're gonna have to supplement that with, on-site generation.
Graham Williams:And increasingly, that's gonna be privately financed.
Mark Bonner:Look. A theme of my conversations with developers for the last two years is that at at some point in time, you've gotta talk to the government. You gotta talk to a state government or you gotta talk to the federal government, Uncle Sam. And that's been deeply frustrating. And so, you know, you kinda mentioned it in passing, but where is friendly right now for data centers on the map and where is challenging on this front?
Graham Williams:Yeah. It's a good question. I I think it's it's obviously hard to generalize, because so many of these rules are hyper regulations are hyper local. There are clearly I think there's between 30 and thirty eight and forty states that have passed certain sales tax incentives that are really important to very large scale data centers, and so that becomes a bit of a threshold test. And then there are localities who manage the zoning, and I think what we've seen is and I'm sure we'll talk about some of the conversations that are happening in localities, but the most challenging areas are where a data center project is net new to a community.
Graham Williams:They've never seen one before. They can't go drive and look at one, and that brings in a fear of the unknown, certainly. And then on the other side of the spectrum where there are a lot of data centers and they've kind of reached a saturation point. So, Prince William County is one of those, Phoenix, Arizona is one of those where you say enough is enough. But in the middle, there's actually a pretty interesting friendly batch that says, you know, in the numbers and if planned the right way, they're really attractive members of a community.
Graham Williams:They drive really good jobs. There's an incredible tax base that gets kicked off. There's a whole ecosystem that supports them. And so it's that middle that has been the most friendly. And then I'd say the last thing is communities that are that are undergoing rapid growth in population and need to find ways to fund infrastructure for their growing populations.
Graham Williams:Data centers are generally a really good match there because they don't bring in they don't exacerbate the growth in population. They're very light touch on traffic in schools and roads, but they provide a wallop of economic development and tax base. So that's how that's how the general in general, the landscape's evolving.
Mark Bonner:But the growth story is running into something else, which is this public backlash. We're seeing more organized oppositions, lawsuits, moratoriums, ballot initiatives, projects that would have sailed through approvals a few years ago are now getting blocked. And in some cases, this is becoming a political issue in local and even state elections. Look. Take Virginia governor Abigail Spanberger.
Mark Bonner:She sailed into office last year in part because of her stance that the data center industry should, quote, pay their own way. For the uninitiated, Virginia is home to Loudoun County, the data center capital of the world. Graham, do you think the industry underestimated how quickly this would become so political? This is everywhere on television these days as well with with the tech companies.
Graham Williams:Look. I I I think the the growth has been, dramatic, and I I do think that the industry has been slow to recognize and engage in the conversations where they need to be had. Look, these are important conversations to have. The growth rate of just the public cloud is roughly 25% a year. And when you add on AI on top of that, the infrastructure and land are growing at 35 plus percent a year.
Graham Williams:And that means every data center that's been built to this point is going to double in the next three years. And it's going to double again in the next three years. There were 26,000 acres of land purchased in the 2025 for data centers. So that demand is important, and it has it's driven by all the applications that we all use in an increasingly digital world. And that means these conversations are relevant for national competitiveness, national security, the economy.
Graham Williams:A Harvard professor last year wrote that in the second half of 'twenty or in 2025, their data center investment accounted for north of 90% of GDP growth. So the 4% that we grew, lion's share of that was driven by data center activity. So that's really important. And then the quality of life, the services that all of us enjoy. Now that bounces up against these require a lot of resources.
Graham Williams:We need to be honest and planful about that. It requires water, it requires power, and it requires a social license from the host And so I think that kind of the shortsightedness of our industry of just trying to keep up because it's really hard to keep up and double every every two or three years. Too many of us in the industry didn't throw our headlights far enough out. And I think we're now seeing the impact of that. But I also look, this is also new and we have to we're in a changing environment, and we're in a change like that we have real conversations to have about public policy, and we need to lean into them.
Graham Williams:I think we there are are when you go through rezoning cases, for example, there's generally a population that's supportive. There's a population that's indifferent. There's a population that wants to learn, and then there's a population that is just opposed. The ones who are indifferent and want to learn can bring very reasonable concerns that we need to address directly, and we need to do a better job on that.
Mark Bonner:And I think one of those reasonable pushbacks would be a reasonable person might be able to say, I see the top line national economic benefits of the data center industry. I I see that this is a powerful, turbocharged economic juggernaut, right, that that is creating jobs and is is contributing to GDP. But if it means that my electrical bill is higher, if it means that I'm at the kitchen table and I'm looking at, wait a second, my electrical bill is 14% higher this year than it was last year, you could understand how a regular Joe could could be anti data center just on that alone. Right? Even if they also wanna have faster Internet and they want their Netflix load and not being interrupted.
Mark Bonner:Right? That's a that that's that's sort of a difficult crosscurrent to navigate for the data center industry.
Graham Williams:Yeah, I think that's right. Look, and no one should say that data centers don't consume a lot of power. They do. But that is a conversation that we want to be able to have in a constructive way around the cost of power, which is the following. Not every utility grid or transmission system operates the same.
Graham Williams:They have different rules about how costs and investments are passed on either in a very specific way or socialized to a rate base. In many locations, we've had you can find CEOs who've recently been talking to people like the Federal Board of Reserves and in the media who are saying actually data center load brings marginal rates down for the rest of the rate base in many places. You know, PG the CEO of PG and E, Excel, DTE, Alliant, Entergy have all referenced the fact that if you can bring very consistent load and even out underutilization, it actually can bring costs down. Also, but there are instances where like in PJM or other places where big investments into the grid get socialized or historically socialized. And so there are a couple of things we should talk about that.
Graham Williams:One, the hyperscalers themselves who are driving a lot of this load have recently made commitments that, yeah, we're going to pay our own way for net new generation so that the rate base doesn't get impacted. And that was I mean, there was a ratepayer protection pledge that they signed with President Trump along those lines. But it's also important to note for that average Joe in a county in Pennsylvania or Virginia that if a data center goes 50 miles away from your house and is on the same grid, you're still going to see the increase in cost, but you won't see any of the incremental property tax that comes because that's going to the county, four counties over. And so yes, these are not unreasonable concerns and they're conversations that need to be had. And I think that the industry is now realizing there just needs to be more engagement early, more transparency, and frankly, the size of these campuses should be thought of as bigger because that allows us to plan and design infrastructure that goes to one place in a part of the community that the community wants it.
Graham Williams:So you don't have a spider web of transmission lines going everywhere. You just build to one place and that you can isolate development toward the center of larger pieces of land to create more buffers. And frankly, just from a developer perspective, if you have bigger developments, then you reduce the amount of risk of a number of them falling through. So I think all of those things are are clearly playing out right now in real time.
Mark Bonner:I mean, look, some developers have already begun to respond to these dynamics. And look, it's starting to look less like traditional real estate development and more like a campaign. We've reported on strategies around community outreach, local coalition building, messaging strategies. Has getting a data center approved effectively become a political process onto itself, Graham?
Graham Williams:Yes. It has. You know, with
Mark Bonner:the And how much does that local sentiment now factor into whether you even pursue a site at all?
Graham Williams:Well, certainly, at at Tract and part of our site selection process, we're looking for communities that are interested in this sort of development, that are welcoming to that on the front end. And even in those communities, there's going to be opposition and we're going have to go through a process. But we have very little interest in rolling a rock uphill. If the will of the people is, we don't want that in our community because ultimately it's their decision. And so we spend a lot of time working with elected officials and economic development leaders and community leaders before we make a decision to buy land and apply to get a sense for is the will of the people such that the benefits of this, if done properly and responsibly, outweigh the concerns.
Graham Williams:And so even then, it has become a political challenge. We count about a third of zonings, a third to a half of rezoning cases for large data centers have been successful in the last twelve months. Transparency is important too. I mean, that is that's an area where our industry has done ourselves a disservice over time where, you know, there's cloak and dagger around who the developer is, who's the hyperscaler code names, all of those things, I think, while not intended to obscure information from the public, have just layered on a level of mistrust. And frankly, I think a lot of the opposition that is just against no matter what is rooted in a fear of the unknown and a disquiet about how technology is changing our lives rapidly.
Graham Williams:Then the data centers become the manifestation of that concern.
Mark Bonner:Right. But we've I I hear you loud and clear, but we've also seen a meaningful number of projects canceled after opposition builds. We've even seen political violence. In Indianapolis, I think, this week or last Yeah. We saw a politician who was supporting a data center development shots into his home.
Mark Bonner:He was uninjured, thankfully. But, I mean, this is starting to reach a crescendo across the country in a lot of different places. Big markets that are established, but also some of these emerging markets that we've talked about. Graham, I mean, the political violence aside, which is a porn, and should never be a part of anything, you know, in in America, What usually goes wrong? You know, if if you do your due diligence and you decide, I'm gonna go forward here, because I I I think there's a pathway, after you've done your your research about a market.
Mark Bonner:If it goes off the rails, what are some of the things that get things off the rails? What goes wrong? Is it a failure to engage early enough, or are there cases where the scale of these projects just fundamentally doesn't fit the community?
Graham Williams:You know, I think a lot of it where where projects have had to be canceled is that there hasn't been enough transparency early in the process, and there hasn't been enough engagement early in the process process to address many of those very, very rational and those concerns that we talked about. And that starts to build on itself if you if you can't directly answer questions. Now, certain members of the community might not like your answers, but if you're not there in person and having direct conversations, then that's where we've watched others really stumble. And I think this is going to get harder for a while because, like anything, the loudest voices make it sound like they're a bigger group than they are. A lot of the polling we see is that you know, the majority of folks are supportive of projects that bring these sorts of benefits.
Graham Williams:And the last thing I'd say is there are a lot of projects that get approved that don't make the news in the same way that the projects that don't get approved. And so there lots of communities that are very happy with having data centers as part of the fabric of their community and a pillar of the community, frankly. And, you know, in our the way our news cycle work, those stories don't get a whole lot of traction.
Mark Bonner:Okay. That's all the time we have today. Graham, thanks so much for joining us.
Graham Williams:It was a it a lot of fun, Mark. Thank you.
Mark Bonner:We'll be back soon with another episode of First Draft Live. You can also find today's episode and all of our past conversations on your favorite podcast app. I'm Mark Bonner. This is First Draft live.