
Every so often, I’ll get a call from my friend Paul. Although he is a nurse and has good job working in a busy clinic, he always has his eye on the ‘next big thing’. In the late 1980s, it was tanning salons and then in the late 1990s, it was tech stocks. Ten years later, in the mid-2000, it was flipping condos, especially in Phoenix and Las Vegas. Each time that he called to discuss these ideas, I would warn that the bubble may be close to busting. And each time, he would quickly dismiss this notion using the same rational – everyone wants to jump on board, the growth will continue to skyrocket, and this opportunity is a sure thing. A few weeks ago, I got a call from my old friend Paul and, just like many of our past conversations, I heard the words ‘skyrocket’, ‘jump on board’ and ‘sure thing’ ringing in my ear. But the topic of this conversation wasn’t tanning salons or condos, it was data centers. This month, we are going to take a deeper dive into the current darling of the real estate sector by discussing what they are, what they need and why they could be the next bubble to burst.
Data centers may seem like they just recently showed up on the scene, but they have been around since the 1940s, when the very first electronic computers were born. As the computer trend accelerated, these mainframe rooms, as they were then called, where characterized by dedicated electrical power, specific cooling systems and restricted access. The term ‘data center’ gained traction in the 1970s, led IBM and the seven dwarfs, which included Burroughs, CDC, GE, NCR, Honeywell, RCA and UNIVAC. The introduction of the internet era in the 1990s, and associated widespread use of the word wide web, email and cloud-based storage, threw more gas on the data center fire. And while the specific uses evolved, the basics needs of a data center remained consistent; heavy electrical power, specific cooling systems and restricted access. There are familiar examples across northeast Ohio that many of you may pass all the time; PNC’s facility on W 150th just north of I-71, KeyBank’s facility on I-480 at Tiedeman Road, several of the Progressive Insurance facilities located along I-271 and, an oldie but a goodie, the NASA Glenn research center on the peripheral of Cleveland Hopkins International Airport.
The sector really began to ramp up with the advent of artificial intelligent (AI). While we discussed the latest AI developments in last month’s column, the AI boom really goes back to the introduction of Apple’s groundbreaking application known as Siri, way back in 2010. This is generally regarded as the first widespread public use of AI. Although Apple maintained strict control of this technology, others quickly jumped on the voice-activated technology bandwagon. This led to an increased demand for data centers, as the technology was real time and interactive. Then, in late 2022, along came ChatGTP and the data center powder keg fuse was lit. While specific uses continue to evolve, the needs remain constant; heavy electrical power, specific cooling systems and restricted access. Let’s dive into these needs as a result of AI.
Traditional data center functions such as web site hosting and cloud-based storage can be accomplished using conventional central processing units (CPUs), but AI systems are significantly more robust. Rather than request and retrieve functionality, AI requires specialized processers known as graphics processing units (GPUs) that can perform thousands of calculations interactively and simultaneously. Electric consumption is typically defined by a kilowatt (kW). One kW will power a TV for 20 hours or a household refrigerator for 3 hours. A traditional CPU server rack consumes between 150 and 250 kW per day but a GPU rack consumes between 1,500 and 2,500 kW per day. A medium size data center contains 1,000 racks, a large corporate datacenter contains 5,000 racks but a hyperscale facility for companies such as AWS or Meta contains at least 30,000 racks.
All of these racks of GPUs being powered by a huge volume of electricity not only creates a quick and detailed response to your queries on the best way to lose 10 pounds, but it also creates a significant amount of heat. Traditional data centers use dedicated cooling systems to circulated air around and under the processing units. But the heat generated by AI facilities would quickly overwhelm those systems. As a result, modern data centers use liquid cooling, with chilled water circulated directly around the hottest components, which removes heat much more efficiently. Access to a significant source of water is just as important as access to a significant amount of electricity. Finally, a modern data center looks like a big concrete box. There are no windows, no lights, no landscaping and no signage. Have you recently driven on I-71 between Columbus and Cincinnati and wondered what those grey bunkers were just north of the outlet mall in Jeffersonville? The largest online retailer in the world prefers that you don’t.
The dramatic rise in demand for modern data centers has resulted in a spike in demand for suitable vacant land. Last year, Microsoft and Amazon made headlines for vacant land purchases in northern Virginia. The former acquired 124 acres for $465 million or $3.75 million per acre while the latter acquired 189 acres for $700 million or $3.7 million per acre. SDC Capital partners, a private investment firm focused on digital infrastructure, upped the ante by purchasing 97 acres for $615 million or $6.3 million per acre. These locations were chosen not as a result of traditional metrics, such as traffic counts, population density or access to rail lines. Rather, they had the three key attributes that modern data centers covet – proximity to heavy electric power, an ample supply of water and enough acreage to provide securable scale. While the raw numbers here in Ohio aren’t as eye-popping, a similar spike in prices clearly occurred. Columbus has emerged as a leader in the state, both in activity and pricing. In Licking County, Meta acquired a total of over 1,300 acres, with the most recent purchases exceeding $1 million per acre for vacant farmland that ordinarily traded for $10,000 to $15,000 per acre. Other areas in Ohio that are considered to be data center hotbeds include Hebron, Lancaster, Wilmington and Massillon.
Despite the bravado, my friend Paul thankfully has never actually acted on his impulses to get involved with tanning salons, tech stocks or condos in Pheonix at their peak. But there are plenty of people that are rushing in to stake their claim in the current data center gold rush. As with any booming industry, there will be winners and, then the inevitable bust happens, there will also be losers. Hopefully, Paul continues to be more focused on IVs than kWs.
Boom and Bust – Last month, the latest chapter was written on a four-building office portfolio in Cleveland’s eastern suburbs. The three-property office campus known as Landerbrook Corporate Center along with Metropolitan Plaza in Highland Heights sold for $12.2 million or $22 per square foot. The same portfolio last sold in 2017 for $53.8 million or $99 per square foot.
Alec J. Pacella, CCIM for September Properties Magazine




















































