How does this relate to site selection and business attraction? Articles about data center site selection tend to focus on one of two things, i.e., “scorecards” showing the best locations for data centers from a cost and/or risk perspective; and delving into the “scorecard” to show the laundry list of key variables associated with data center site selection. Both have value — to the end-users procuring new sites as well as to the economic development agencies and real estate developers looking to position their communities to attract data center business. Yet, in my experience, these guides often do a poor job of explaining the most important question: Why do data centers cluster in certain regions more than others? Why hasn’t a lower-cost location, e.g., Sioux Falls, S.D., actually attracted data center activity?
The answer lies in a better understanding of the overall data center marketplace and its competitive supply/demand dynamics. As data centers have matured as an asset class, with six REITs (soon to be as many as 10) and multiple publicly traded operating companies providing data center services, a real spectrum of options has developed in the marketplace, with different data center product types meeting different needs across the spectrum.
A Spectrum of Options
Equipped with this deeper understanding of the spectrum, site selectors and attractors can better tailor value propositions to find the right data center fits. The product spectrum ranges from unimproved land to fully built managed/cloud services offerings (with hybrid options falling in-between) as follows:
Unimproved Land: At the lowest end of the value spectrum, land is available for lease/purchase to a single tenant or buyer. Entry barriers are low (except in urban markets) and represent a small fraction of the data center budget (0.5 percent to 2 percent). Utilizing unimproved land, center development timing averages 24 to 36 months.
Existing Property for Retrofit: An existing structure (single-story, heavily built warehouse) is available for lease or purchase, typically to a single tenant or buyer. This may be an industrial property or commercial office building available for conversion. Some power may be in place with enhancement work needing to be done to the structure for data center use. This option represents a small fraction of the overall budget (less than 10 percent including land). Data center development from this type of asset may take 12 to 24 months.
Powered Shell: This is the first type of product truly unique to data center industries, available for lease or purchase to either multiple tenants or a single buyer. A true powered shell is a hardened facility with key elements of utility power distribution and telecom access in place, as well as enhanced perimeter security features and a master plan to show future use. This may be an existing facility that was a retrofit or a purpose-built new building designed to house data center infrastructure. It may reflect as much as 15 percent of the overall budget (including land). Data center development from this type of asset may take nine to 12 months.
Wholesale Data Center: Outside of hardware, this is where the bulk of the investment lies (over 80 percent). Major mechanical/cooling systems (cooling towers, chiller plants), electrical systems (emergency generators and UPS battery systems), fire detection and suppression (sprinkler or gas systems), and final space construction/fit out (air-conditioning units, raised floor systems, power distribution units). The space is “move-in” ready for the installation of equipment, and the wholesale data center provider is responsible for the maintenance and operations of mechanical and electrical equipment. This tends to be a multi-tenanted environment, with the key value proposition in the tenant benefiting from the economies of scale gained when a larger site is developed from both a capital and operating standpoint, as well as redundancy spread across a broader tenant base. Tenants will likely have dedicated infrastructure components, whether a demised suite, UPS modules, or other, with certain “major” infrastructure that is likely to be shared (e.g., generators and cooling towers).
Retail Data Center: This is also referred to as “cage/cabinet collocation.” The main difference in terms of business model is that the provider typically partitions space down to a single four-post cabinet or small “cages” with wire mesh walls separating different tenants’ equipment. The provider will typically make additional investments inside the data center room with respect to ladder racking, cabling, and power distribution (less than 5 percent of the overall investment). The provider manages the major power and cooling infrastructure but is also likely to provide more services at the cabinet/rack level, with technicians on-site to perform physical work (installations and power reboots/pressing buttons). There are many more tenants and the value of contracts is smaller but this tends to be more lucrative from a rate standpoint for a provider.
Managed Services Provider: This model is similar to the retail data center, but providers will typically carry more professional services capabilities (technicians/engineers with higher IT and networking certifications). There is more of an IT outsourced flavor to the business model, and when labor is managed properly, margins are superior.
Cloud and/or Hosted Services Provider: The provider makes an additional investment in the hardware as well as platforms/operating systems and user interfaces — which starts to look less immediately relevant to site selection but is an important user case and driver for data center business growth.
With these product offerings in mind, often the site selector needs to make a strategic decision early on — i.e., does the firm expect to invest its own capital to build and operate its own data center, or does it want to have the option of leasing the space from a provider? The response informs whether the end-user will want to evaluate multiple land options or multiple wholesale options.
Conversely, the data center site attractors simultaneously need to make another decision — i.e., how much capital are they willing to spend to attract potential tenants and buyers and what do they expect the product offering to look like? The majority of data centers are less than 50,000 gross square feet in size with less than 2.5 megawatts of critical IT load. At smaller sizes, building one’s own data center may cost as much as two times what it costs to lease space from a third-party provider, wholesale or retail.
Proven & Unproven Markets
As a result, the majority of new development domestically has been made by the data center providers in locations across the country that have a history of data center supply and demand, e.g., NY Metro, Northern Virginia, Chicago, Dallas, Silicon Valley, which are each key metropolitan areas with large populations and business density, and data centers exist everywhere people live and work.
“Proven” data center markets provide additional benefits including:
- Capital/financing availability — the more mature the market, the more willing traditional lending sources are to provide competitive financing and funding.
- Competition — which is good for tenants, but not always for providers. However, providers benefit from a deeper pool of vendor and labor pools.
- Deal structure alternatives — a mature market has competitive options for the above listed entities.
- Business clustering — data centers are core to generating revenue for businesses. For a buy side financial services firm, the data center may be a place where they transact with a sell side firm via a market-exchange matching engine.
In stark contrast, rural/unproven markets have unimproved land or existing properties but no history of actual data center development — a vision is being sold. This makes the most sense for companies labeled as “market-makers” such as Google, Facebook, Microsoft, Yahoo, and Apple. These companies each build massive “hyperscale” data centers to the tune of hundreds of millions of dollars of capital investment per facility. They have been the first movers for domestic locations such as The Dalles, Ore.; Cheyenne, Wyo.; Reno, Nev.; Lenoir, N.C.; Lockport, N.Y.; and Goose Creek, S.C. A challenge for the majority of site selectors and attractors is that these massive requirements don’t necessarily align with the needs.
Knowing Your Options
Which brings me to my final conclusion regarding why markets matter. For a business looking to make a major data center investment, whether at the construction or hardware level, it helps to know that if economic or business conditions change, you have options. An overbuilt enterprise data center in northern Virginia may have a second life as a retail data center or managed services provider. More liquid markets provide competitive options to meet different business needs, large and small, and offer downside protection if business conditions change.