The Rack Report

It’s the Return on Equity; Financial Comparison of Data Centers

 Email Article  

Tweet  

  

By Peter Tippen, Partner, Rackhouse Group LLC

Two things are true about the cost data center facilities.  First, the total value of equipment and non-real estate assets is commonly 4 to 5 times the value of land, bricks and mortar.   Second, with some exceptions, there are no two data centers alike. 

Occupancy cost analysis models developed from the real estate industry are frequently used to compare data center alternatives.  By "occupancy cost analysis" we mean a method of comparison whereby alternatives are lined up, total costs are computed and the total costs are compared, one alternative to another, by using the lowest common dominator, sometimes the square foot of space occupied.   Real estate metrics confuse the measurement of non-real estate assets. 

As an alternative, practice the following procedure.  We agree - it is more involved.  A great deal of the input has to be developed and, depending on obtainable data, can be imprecise.  But, the process itself is valuable and gives focus to issues important to the data center facility decision. 

• 1. Estimate the value or total cost / kW of critical load for each data center alternative. [This is the hard part, be patient, it will be worth the exercise. We know engineers and construction professionals are not available for every analysis. However, do your best on this. Rackhouse has seen situations where even gross estimates of costs / kW of critical load can be used to make this point. So, do your best.] 
• 2. Compute the cost to finish out, move in and commission each alternative.  
• 3. Identify economic lives for the component parts of each facility. For example, what are the chances that any equipment will have to be replaced during the contract or the ownership for the data center? If leased or contracted, what will be the cost, if any of replacing it? If owned, what kind of amortization of the asset represents an annual cost of equipment vs. real property? Comprehend that cost on an annual basis.
• 4. Understand the amount of power and its cost. Include power required to run the data center (the mechanical load) as well as the servers (the critical load). [This, like number one above, may not be an exact science. We have seen benefit from best guesses and estimates.]
• 5. Finally, capture all rent, contract and expenses. For owned data center alternatives - interest on borrowed capital, cost of equity capital, expenses and residual values.

Then,

• 1. Sum the costs from numbers 2-5 above for each year of the data center's life or use. This is the annual cost ("Annual Cost").
• 2. Compute the Annual Cost / kW of critical load.
• 3. Divide the Annual Cost / kW of critical load by the total cost / kW determined in the first step[1].

The annual quotient represents the cost of the asset per year / kW of critical load.  It is superior to any comparison that does not attempt to take the value or total cost of the asset into consideration because, as we began, the data center's cost is largely comprised of equipment, and every data center is different from the next.

Conclusion - Compare to the Cost of Using your Own Money to Own a Data Center

Using the information developed above, the analysis can be manipulated to calculate the data center total cost / kW of critical load in either a real or hypothetical example of ownership.  To do this, assumptions will be required regarding the amount, cost and amortization of available debt. In addition, return on capital assumptions are made on the equity required to own.[2]  Whether the data center user would consider ownership or not, this comparable number can be used to judge the reasonableness of the cost / kW of critical load for the data center being examined.  In other words, "Why would I pay you X if I could own if for Y?" or "This looks like a deal to me."  We hear more of the latter than the former these days.  When it's all boiled down, return on equity is king.

[1] The cost per year should be analyzed per kW of critical load.  There may be even better comparison metrics.  How about cost per server or cost per unit of data?

[2] Return on Equity (ROE) - The annual return on shareholder equity targeted for investment and used to analyze investment decisions.  Each company or situation will have its own application of the ROE. 

Cool image by:  http://www.flickr.com/photos/daviddmuir/