The economics of data center outsourcing

Dec 13, 2000
Gartner

C. Da Rold, A. Erba

Economies of scale are expected to greatly benefit data center outsourcing (DCO) vendors. By providing services from large data centers, outsourcers are expected to be able to pass on savings to any client via low-cost MIPS (millions of instructions per second). But is this really the case?

This discussion of economies of scale in data center outsourcing (DCO) is part of determining the latest trends, issues, and events in the IT outsourcing market, and whether the critical success factors are achieving the desired benefits from outsourcing.

Join the IT debate

What have been your experiences with DCO? What kind of sales technique did the vendors use? What would you recommend to someone considering outsourcing? Post a comment to join the IT Debate. Next week, we'll feature the most insightful comments in our IT Debate column and award IT Debate T-shirts to the members who provided them.
The following Gartner research note is available online for one week. Subscribe to our IT Debate TechMail service to have this valuable information delivered to your Inbox each week.

The economics of data center outsourcing have changed dramatically in the past 10 years, mainly due to increasing competition and hardware cost reduction. The cost of the data center's raw material—the CPU MIPS—was once very high (thus sharing the data center yielded high profits), and has now dropped in price. The average data center TCO (the full MIPS cost, as measured through Gartner benchmarks, including hardware, software, occupancy, administrative and technical staff, operations, and disaster recovery) in Europe at the end of 1998 was about $28,000 per installed MIPS. At the same time, some DCO offers in Italy were in the range of $12,000 to $20,000 per installed MIPS, and sometimes less.

Does this mean that any client can save a lot of money through DCO? Does this mean that vendors have a full MIPS cost lower than $10,000 per installed MIPS? The answer to both questions is no. Client organizations must be aware that the large gap between data center TCO and "market price" for MIPS is closely related to what is included in, and what is excluded from, the services offered.

Economies of scale apply when the cost to deliver a certain amount of services grows less than proportionally to the volume of services delivered. They are at the core of the outsourcing concept. Services that cannot be sold in volume and that therefore do not benefit from economies of scale are out of the outsourcing game. Therefore, if an enterprise has a specific or unique service and it manages that service well, there is little or no opportunity for the enterprise to save money or improve service quality by outsourcing.

To analyze the effect of economies of scale on data centers, we consider the average cost per MIPS by size of data center as measured by our benchmark database. The benchmark data shows the following:

Economies of scale are a continuous effect, but significant effects start even at low levels (about 200 to 300 MIPS). Additional effects apply above 700 and 1,100 MIPS.
A data center of more than 1,000 MIPS costs on average about 30 percent less than a data center of about 200 MIPS; this difference is a measure of the effect of economies of scale on data center cost (Chart 1).

Chart 1

Chart 2 shows the effect of single-organization characteristics on the range of MIPS costs. The benchmark data shows the following:

The range between data centers in the same class is far higher than the effects of economies of scale (e.g., for the 200-MIPS class the cost ranges from minus 70 percent to plus 130 percent).
In every class there are more inexpensive data centers than there are expensive ones.
Very inexpensive data centers cost about 40 percent less than their class average.
Very expensive data centers cost from 1.5 to twice the average of their class and from three to six times the minimum.

Chart 2

Finally, looking at the benchmark results of some DCO providers (outsourcers), we can say that their fully loaded MIPS direct cost is quite similar to, or just below, the cost of a large, well-optimized client organization. Taking into account the vendor's gross margin (i.e., including its overheads and margins), its full MIPS cost cannot support as low a price as is sometimes offered, nor provide true cost savings to well-optimized clients.

Keep in mind these key facts:

Apart from very small data centers, the effect of economies of scale are relatively low, being in the order of 20 percent to 30 percent (i.e., often lower than the expected vendor gross margin).
The cost differences between organizations of the same scale are larger than the effect of economies of scale alone.
The DCO providers' MIPS full cost does not allow them—due to their overheads and margins—to provide real savings to well-optimized clients.
Only clients with a cost higher than the average, or clients with very small data centers, can gain savings buying full-cost MIPS from DCO providers.
Other clients asking for DCO will instead get an offer that relates to a partial DCO service and that demonstrates some kind of savings.
Clients must be aware that the general rule is: the lower the price per MIPS, the lower the real service content.

Acronym key

MIPS—Millions of instructions per second
TCO—Total cost of ownership

Bottom line
Client organizations need to be aware that the large gap between data center TCO and the market price for outsourced MIPS is closely related to what is included in the services offered. Vendors are often simply unable to provide savings to clients running optimized or large data centers. However, being more sophisticated than clients in dealing with outsourcing, they can often present some kind of savings through different outsourcing deal techniques, thus building appealing figures to offer.

Some outsourcing deal techniques used by vendors are:

Pick the service areas where the vendor can provide cost reduction more easily (through economies of scale, standardization and practices) and discard other areas where the client's costs cannot be improved (offer line design, deal shaping).
Build appealing entry conditions (e.g., asset payments), which the client will pay for during the term of the contract.
Build appealing cost reduction based on long-term financials (e.g., including all the expected MIPS cost reduction in the first-year price).
Reduce the service content or the service levels according to the cost reduction that is needed (e.g., reducing the staff planned for the service).
Delay any detailed description of service levels and responsibility matrix until after the deal signature and the first transition.
Play with economies of scale.
Lose money on the DCO part of a deal, making additional margins on other services for the same client.

Clients should save time and money by not starting a DCO project until they have carefully measured their data center TCO and compared it with the market.

Gartner originally published this report on Nov. 2, 1999.

Join the IT Debate

What is your experience with data center outsourcing? Post a comment to join the IT Debate. Next week, we'll feature the most insightful comments in our IT Debate column and award IT Debate T-shirts to the members who provided them.