At CRE.Converge 2019, attendees had the chance to learn about the unique aspects that define data centers and the factors driving their development and leasing. Kristina Metzger, leader, CBRE Data Center Capital Markets, led the discussion along with panelists Rafal Rak, vice president, portfolio management, Digital Realty, and George Rogers, principal, Skyline Capital Advisors, LLC.
Metzger started with an overview of data center basics, beginning with the definition: “A large group of networked computer servers typically used by organizations for the remote storage, processing, or distribution of large amounts of data.”
These servers used to be housed in the basements or store rooms of office buildings, she pointed out, but have since evolved to be housed in more purpose-built structures located separately from office space.
The demand drivers for data center development include the global use of data creation, collection and storage. “Internet browsing, pulling remote data, Artificial Intelligence, the internet of things, everything in your house that is connected to internet,” said Metzger. “Your Ring doorbell, Nest thermostat… these are all driving data creation and that data needs to be stored.”
The annual size of the “global datasphere” has grown from less than 20 zettabytes in 2010 to a projected 175 zettabytes by 2025. “And 90% of the world’s data was created in last 3 years,” Metzger added. So the demand for data centers seems to be without ceiling for now.
Data centers can be divided into two core asset types: powered shell and turnkey/co-located. Powered shell data centers are facilities with exterior construction completed, available power and connectivity, but with the interior left as raw space to be finished, maintained and operated by the customer.
“The rent structure is similar to a traditional industrial facility,” said Metzger, “and we see an average cap rate of 5-7%.”
Turnkey or colocation data centers are typically multi-tenant facilities that offer fully built-out infrastructure from the Power Distribution Unit or greater demarcation point. These solutions are typically operated and managed by the landlord/operator, include 24/7 security and are most often subject to a Service Level Agreement (SLA).
“Power allocation is what drives the rent in these facilities,” said Metzger. Tenants pay for kilowatts used per month.
Primary data center colocation markets across the U.S. are Northern Virginia, Dallas/Fort Worth, Silicon Valley, Chicago, Phoenix, New York tri-state area, and Atlanta.
The primary location drivers include electricity costs (“The single greatest expense for data center owners,” said Metzger), taxes, latency (the amount of time it takes from data to travel from one area to another), and natural disasters.
For Northern Virginia specifically, Rak noted there are several factors that make it an appealing location for data center development: subsea cables to Europe, the amount of fiber, sales tax abatement, and low development costs and power costs.
Phoenix is one of the fastest-growing markets, Metzger said, with major drivers behind that being the area is a low-cost alternative to SoCal with less proximity to natural disasters, appealing tax incentives, and a sizeable amount of land.
“And sales tax is also important for data centers compared to other asset classes because they put $100 million to $150 million worth of technology in there,” Rak pointed out. “For these bigger companies like Microsoft or Amazon, that can have a huge impact on their bottom line.”
Attendees asked the experts about specific obsolescence concerns for the product type. Metzger said she’d include pounds-per-square-foot load and clear heights. Servers continue to get smaller, density increases as server racks grow taller to accommodate more hardware, and ample clear heights are needed to ensure the cooling equipment can keep up with the increasing amount of power being used (and therefore heat generated).
“I believe we’re in early stages of demand and the market will continue to move for this asset class,” said Rogers. “With AI, blockchain, and so much data in the medical field and beyond… there is so much more information out there that we’re still in the early stages of storing and using.”
“The reality is, the majority of companies still store their data within their office building,” said Rak. “They might have difficulty migrating that to a cloud service because it takes so much time and could potentially impact their business. So I think were in the early stages when it comes to widespread adoption.”
“With 5G and the way we’re going, I don’t see the trend going backward where we use less data – were going to become more connected,” added Rak. “We’re at about 15 years or so of data collection just from Facebook alone, and that data keeps accumulating – it won’t go away. My wife still has a Hotmail email address with 16,000 emails in it, and she can still pull up the oldest message from 1999.”
Brielle Scott is Senior Communications Manager at NAIOP.