What Is A Data Center? Admit It, You Still Don’t Know How This Multibillion-Dollar Sector Works

Data centers are entering the mainstream of commercial real estate, and industry professionals need to be ready.

Data center capacity in North America has more than doubled since the start of 2018, according to PwC.

This explosive growth reflects the degree to which these facilities support nearly every aspect of our increasingly digital lives, whether we realize it or not. Every time we make an online purchase, send an email, use our phone for directions, attend a Zoom meeting, watch Netflix, track a run on our Apple Watch or use the internet in any way, it happens because of the data flowing back and forth at the speed of light to and from servers housed in data centers. 

Until recently, data centers were something of a siloed alternative asset class — hugely technical buildings that were regarded more the realm of telecom or tech than real estate. They also only existed at any significant scale in a handful of locations like Northern Virginia and Silicon Valley, not in most local markets. But all that is changing. 

Data centers were one of the few asset classes to thrive due to the pandemic, drawing in a broader range of investors who have continued to see the sector as a growth opportunity and relatively safe bet during turbulent economic times.

The industry’s growing scale has also attracted developers from other areas of CRE, particularly those who specialize in logistics or other industrial uses. At the same time, the data center landscape is becoming far less centralized, with large-scale data center developments increasingly coming to markets, such as Nevada and Minnesota, that previously saw few such projects.  

This means that many professionals across the commercial real estate space are going to have to become familiar with at least the basics of a sector that can seem a bit impenetrable for those without a tech, telecom or IT background. The data center industry has its own jargon and set of assumed knowledge that can be quite different from other areas of real estate, but it is a language that is going to be ever more crucial to understand. 

With that in mind, here’s a primer to understanding data center real estate and development: 

What are we talking about when we talk about data centers?

On one hand, data centers are essentially just large buildings housing thousands of servers and the equipment needed to support them. But the companies that own and operate data centers have an increasingly complex array of business models that are shaping investment and development across the sector. 

Imagine a large, Fortune 1000 company with complex IT needs: perhaps a car manufacturer, with its own IT systems to support everything from its email, online sales platform and dealer portal to its design software and financing platform.

Twenty years ago, a company like this would likely house the servers supporting these systems in data centers it owned and operated. But while these self-operated facilities — often referred to as corporate, enterprise or on-premises data centers — are still utilized by certain industries, their numbers are dwindling. Today, the vast majority of companies have found that it is far more efficient to outsource their IT hosting needs, if not the IT infrastructure itself. 

Often, this means leasing space for servers in a facility operated by a third party known as a colocation data center. There’s not just one colocation model;  these can be turnkey facilities where a tenant has to do little more than plug in its own servers, sometimes called retail colocation.

In other cases, larger tenants will lease a section of a data center or an entire facility that they will mostly outfit themselves, with the landlord providing little more than the building shell, power and basic infrastructure. This model is often called wholesale colocation. 

Companies also have a third option for their IT infrastructure needs: the cloud. Instead of worrying about managing individual servers or IT infrastructure at all, a company can simply purchase a set amount of computing power with specific performance characteristics from a cloud provider, the largest being Amazon Web Services, Google and Microsoft. 

Of course, all that computing power is still happening on servers hosted in a data center, just one managed by the cloud provider. The cloud industry has been the biggest growth driver for the data center sector, with the major cloud providers — often collectively referred to, along with social media giant Meta, as hyperscalers — utilizing an ever-growing share of the world’s data center capacity.

Because they need to be able to expand their computing power quickly to meet unexpected demand, hyperscalers are increasingly leasing or self-building massive campuses with room for multiple data centers. 

In reality, the divisions between corporate data centers, colocation and the cloud aren’t so clean cut. Most large companies utilize some combination of the three in what is known as a hybrid architecture. Major data center companies also generally have a hand in each of these models, creating a tangled web of relationships between data center providers, cloud companies and the enterprises using both their services. 

Providers like Digital Realty and Equinix operate traditional colocation data centers, but they also offer cloud services, along with a whole spectrum of so-called managed services or infrastructure-as-Service that fall somewhere in between. This includes services like bare-metal cloud, where companies lease specific, preinstalled servers run by a third party. 

At the same time, that data center provider might be building out a build-to-suit data center or an entire campus for a hyperscaler, who may also be leasing space in some of the provider’s colocation facilities. The data center provider is also likely leasing some of the data center space in which they operate from other operators.

To use a phrase popularized by a certain hyperscaler: It’s complicated. 

You can’t just build data centers anywhere.

Although data center vacancy rates are at record lows and demand for these facilities has continued to outpace new development, the data center building boom of the past three years has largely been limited to a handful of markets.

Why haven’t data centers been popping up everywhere? The answer lies in the two main site considerations for data centers: power and connectivity. 

All the servers in a data center need constant power, and a lot of it. The data center industry generally talks about the size of facilities not in square feet but in megawatts, as the amount of electricity a site has access to serves as a better measure of its ability to produce revenue than physical size. 

 The amount of energy needed by each individual facility has also been going up — where a 2MW or 3MW data center was considered big just five years ago, developers now regularly build campuses exceeding 100MW.

Because of their enormous power usage, data centers need to be in markets with a low cost of power to be profitable.  They also need to be sure the local utility can generate the power they need and successfully deliver it to the site, a process that generally requires high-capacity transmission lines and large substations. 

Such sites are rare, and they’re getting rarer. Traditional data center hubs like Loudoun County, Virginia, are experiencing significant power constraints, as the industry’s power demands exceed the capacity of aging transmission infrastructure. Virginia utility Dominion Energy had to unexpectedly withdraw power commitments to some data center developers last summer, and experts expect more of these issues to arise in the months ahead. 

So why aren’t data centers flying up in every market where there is cheap power? 

Data centers don’t just need power, they also need connectivity: access to the fiber cable networks carrying data between data centers and to end users at the speed of light. Data center operators need to be able to ensure certain processing speeds for customers, known as latency. That requires being near  large concentrations of customers, or near a level of robust fiber infrastructure that most markets lack at the moment.  

But cheap power markets with poor connectivity may be seeing data centers soon. Power constraints in traditional data center hubs are making the expense of building out the fiber needed for new markets more tolerable for developers, while subsidies for such middle-mile fiber were included in the 2021 federal infrastructure bill. 

A few use cases are also emerging for data centers with access to lots of cheap power but poor connectivity. Cryptocurrency mining was the most prominent driver of this kind of data center, sparking development in nontraditional markets as far flung as North Dakota and Wyoming on sites tied directly to renewable energy sources. The crypto mining sector has crashed, but some new industries are trying to fill the demand gap at these facilities: from AI training to rendering movies and video games. 

Everything is moving toward the “edge,” even if no one can agree what that means.

The edge is one of the data center industry’s most common buzzwords, even as its meaning remains fluid. 

In the most general sense, the edge refers to data storage and processing that takes place close to the end user.  For a large cloud provider, a large data center in Denver or Boston might be considered an edge facility if it serves customers only in that area, as opposed to a core data center in Loudoun County that stores and processes data from all over the country. Cloud giants like AWS are increasingly building out this kind of edge infrastructure to provide low latency options for certain customers in major population centers. 

But “edge” is also commonly used to refer to much smaller, often unmanned facilities used to quickly process certain types of data produced in their immediate surrounding. This includes processing units on cell towers — sometimes called the wireless edge — that could see widespread adoption due to the emergence of 5G, or roadside units used to quickly process data for driverless cars. 

While the exact definition of the edge remains murky, its widespread use reflects a certain trend within the data center and digital infrastructure space: The industry is becoming increasingly decentralized. Companies and individuals everywhere are using and producing an ever-growing flood of data that increasingly needs to be stored and processed near them. 

And that means data centers will be everywhere.