Will Data Center Consolidation Lead to Higher Cloud Prices?

All the merger and acquisition activity in the data center market has some worried about cloud computing costs.  


No-Size-Fits-All! An Application-Down Approach for Your Cloud Transformation

Posted December 11, 2017

Guest Author

Everywhere you look, massive data center consolidation is underway.

Digital Realty recently purchased DuPont Fabros. CenturyLink is both buying Level 3 (along with its data center assets) and shedding some of its legacy assets to Cyxtera. Equinix is also moving on two fronts, concurrently finalizing its purchase of Verizon’s data center assets and its acquisition of Telecity’s European data centers. CyrusOne completed its purchase of Sentinel. ViaWest is consolidating its assets with Peak 10.

Where does all this consolidation lead? Does the activity portend a spike in cloud computing and hosting bills as fewer and fewer players offer inventory for hosting assets?

First, it’s important to understand what data centers are. While IT folks have experience working in these facilities, developers are largely abstracted from what is a very important piece of the cloud computing ecosystem — many have rarely or never seen the inside of a datacenter.

Data centers are buildings that are specifically built to provide services to hosting assets such as servers, storage, and networks. They often look like windowless Best Buy stores from the outside, with multiple levels of redundancy for water, power and electrical systems so they can maintain service even in the event of extreme weather. For example, many data centers in Houston were built above the 500-year flood plain (the area impacted during Hurricane Harvey) and were able to avert flooding. In addition, these big boxes have an enormous amount of power utilized both for cooling and operating equipment. Finally, they have large generator and UPS banks to support loss of power for extended periods of time. They are basically bunkers for cloud computing gear that are designed for the specific needs of IT equipment — reliability, security and stability, along with adequate cooling and power. Ultimately, data centers are the physical foundation on which all cloud services are built.

On one hand, the demands of data centers and the equipment going into them aren’t static. Each revision of hardware tends to drive more efficiencies; for example, more cores per processor or more storage per drive. These technological improvements mean that in the same space, you can pack more data and utility in each successive integration of data center gear. Of course, there is a wall of how dense users will go — limits of revenue and liability per rack, etc., come into play. But for the foreseeable future, there is deflationary cost of data center space as more and more compute fits in the same racks. Similar to Moore’s law for processing power, which states that computers grow faster over time, data centers grow more productive over time.

At the same time that consolidation is taking place, new entrants are appearing. Sabey Data Centers is founded by the Sabey brothers, who come from the building construction industry. Lincoln Property Company offshoot Lincoln Rackhouse is a real estate play focused on the data center market (which includes refreshing enterprise data centers into something more commercially viable). Of course, there are specialists such as Rob Roy’s Switch, centered on large-scale customers wanting efficient cooling and power.

On the other hand, data centers are creating “moats” (a concept from Warren Buffett that denotes ability to maintain competitive advantage), giving them pricing power and make it more difficult to locate elsewhere. Equinix and Telx (now part of Digital Realty) have heavy cross-connect revenue, which is derived by connecting two entities via cabling and charging a monthly fee for it. Because these facilities are “the places to be,” more and more cross connects are sold.

A new entrant to a marketplace faces a challenge — how to get people to locate with them when everyone with whom they want to connect is in another data center. Known as the “network effect,” and popularized by the rise of Facebook’s acquisition model, a data center becomes exponentially more valuable with each customer added, because it allows the existing customers to potentially make another connection in an efficient way (via a cross connect). Isolated data centers have an uphill challenge in getting customers who want to be where all their network neighbors are to locate with them instead, even if these firms make the price compelling.

Similarly, many data centers are offering on-ramps to public clouds — with some proudly announcing they have a degree of exclusivity for a region or area that gives them an edge with customers looking to reach these clouds efficiently.

The current market is definitely headed toward shortage and a spike in underlying cost. With the tidal wave of demand, mega-clouds are paying huge premiums for space as they need it now. On the receiving side, data centers operators are building with sizeable “spec” footprints (i.e., build it and they will come), often selling out before construction is completed.

Because of the long lead times and exacting specifications to build these facilities, prices are abnormally high. In Ashburn, Virginia, for example (outside Washington, DC), rural land has spiked over $1 million per acre as demand outstrips supply.

However, this tighter land market is unlikely to lead to a direct price spike for cloud services. The more likely case (as we have been seeing) is that prices aren’t dropping as rapidly as they would have had there been a more balanced market, especially since large-scale hosting providers are signing long-term leases with large pricing escalators (because they have no other choice) thereby locking in the higher costs for years. The market will rebalance, but that adjustment is probably still many years away.

What does this ultimately mean?

In the short-term, or at least for the next few years, the embedded cost for data center space in the cloud market is higher than it should be. The good news is that the efficiency gains from the hardware that operators are using allows those entities to defray some of that cost and should be able to mitigate it over time as the market ultimately normalizes. For the end users, it just means prices aren’t dropping as rapidly as they might have if the demand weren’t so high.

The great news is that prices will keep dropping, and potentially faster, as the market for space rebalances. Ultimately, the embedded cost for data center space with today’s prices is a fraction of the total cost, and other efforts at efficiency will yield more gains than trying to battle to save on data center costs in today’s market.

Supply and demand will ultimately decide how this situation evolves. While we should reach equilibrium over time depending on the speed of rollouts and possible shortages, the current climate could lead to higher than normal prices for some time. The most likely outcome is that prices won’t drop as fast as they would if the market were efficient.

by Marty Puranik — CEO of Atlantic.Net, a leading cloud hosting provider.

Photo courtesy of Shutterstock.

Tags: cloud computing, data centers, cloud prices

0 Comments (click to add your comment)
Comment and Contribute


(Maximum characters: 1200). You have characters left.



IT Management Daily
Don't miss an article. Subscribe to our newsletter below.