Once upon a time, urban data centers had a reason for being. Network costs were sky-high, applications were unpredictable and difficult to manage remotely, and so server hugging was a legitimate approach to IT infrastructure management. Why would anyone move applications off site when there weren’t even mature data center operators with whom one could reliably deploy?
All of that has changed, with the end result being the death of the urban data center. Does anyone need any more proof beyond the moves by the Chicago Mercantile Exchange and the NYSE, moving their large data centers out of urban centers, to Aurora, Illinois and Mahwah, New Jersey respectively? It doesn’t get more mission-critical than commodities and equity trading, and the moves of these firms are clear evidence of the trend toward data centers being deployed outside of urban centers.
The following are the top reasons to believe urban data centers will rapidly diminish over the next five to 10 years:
1. Server hugging isn’t necessary. As far back as 1991, Foster Wheeler and other forward thinking enterprises were managing large portions of their (Japan-based) infrastructure remotely. Today, the tools are mature, the service providers are mature, andÃ‚ data centers are rapidly becoming a recognized and mature segment of industrial real estate.
2. Network costs are nominal. Today one can deploy gigabit ethernet across a metro at a fraction of the cost from just a few years ago. And the networks are more reliable.
3. Office landlords don’t want data centers. With a growing emphasis on sustainability and LEED certification, many landlords are simply unwilling to provide the extra power necessary to drive today’s data center environments.
4. Office users don’t want data centers. Managing and sometimes migrating office data centers is a herculean effort. Is it easy to scale when permits are required for installing new power? Most office managers have little insight into how much power their data centers consume to begin with. And how many firms can take advantage of lower real estate prices when it involves the dismantaling or re-deployment of their entire data center?Ã‚ In San Francisco, the average office lease is between four and five years, which means these issues must be addressed often times far before the equipment has even been depreciated.
Based on the above, what can operators, investors, developers, and users expect? The answer is more consolidation. As urban data centers disappear, there will be a natural tendency for users to aggregate in established regions, such as Virginia, New Jersey, Dallas, and Chicago (suburbs), with Iowa, North Carolina, and Nebraska acting as secondary markets.
In the long term, there may be some twists to the historically robust Silicon Valley and Los Angeles regions, owing to their earthquake risk and high power costs. I find it difficult to advocate for any company to deploy in those regions, regardless of their size. Witness Twitter and Facebook, who are paying between 8.4 and 12.4 cents per kWh, versus as low as 4 cents or less in neighboring states.Ã‚ They both started with just a few racks, but owing to architecture and resource costraints have been unable to diversify (until recently).