Merrill Lynch Cases IT Spending, Server Buying Patterns

The analysts at brokerage house Merrill Lynch have been taking the pulse of data center managers to get a sense of how things are going out there in IT Land. A couple times a year, Merrill Lynch polls 75 chief information officers in the United States and 25 in Europe to pick their brains on various issues. According to those polled, IT spending among these large enterprises is expected to rise by an average of 5.2 percent in 2006 and by an average of 4.8 in 2007.

Last November, these same people were polled, and they expected spending to increase by an average of 2.7 percent this year. Clearly, there is a little more optimism out there. Merrill Lynch analyst Richard Farmer said in a report released last week that he expects actual IT spending to increase by 7 to 8 percent in 2006, since the small and medium business market is under represented in the polling data.

Across all geographies, spending on laptops was expected to increase by nearly 5 percent, and spending on RISC/Unix servers was expected to go up by nearly 2 percent. Mainframe spending, however, is expected to decline by 1.5 percent among those polled by Merrill Lynch, and spending on X64 servers is only expected to rise by 1.1 percent. On the storage front, spending on network attached storage averaged a 6.3 percent rise in the poll, while storage area network spending is predicted to rise by 5.2 percent in 2006. CIOs say that, on average, spending on application software will grow by 5.6 percent, business intelligence software will grow by 4.8 percent, database software will grow by 4.2 percent, and middleware by 3.3 percent.

Merrill Lynch did not ask specifically about IBM’s proprietary servers, but those polled said they expected, on average, that 6 percent of their server budget would be spent on AIX servers, 11 percent on Solaris servers, 19 percent on HP-UX servers, 9 percent on Linux servers, and 55 percent on Windows servers. Some 75 percent of those polled said they believed that Solaris was a better operating system than Linux, but almost all of the companies who said this also said they would not change their spending plans in favor of Solaris. That would be a big bummer for Sun Microsystems. These same CIOs said that they expected to boost their spending on Hewlett-Packard iron in all of the key markets: Windows, Linux, and HP-UX.

While 28 percent of those polled said that power consumption was important to them when it came to servers, 23 percent of those polled said that price/performance was more important and 72 percent said power consumption was not an important enough issue to change their spending plans when it came to servers. About 90 percent of those polled said that power efficiency alone would not overcome “vendor stickiness,” and only 7 percent said that power was the main issue.

I have heard a lot of talk with CIOs lately about how power efficiency and virtualization (which drives up server utilization and therefore makes servers yet more power efficient) do not engender server consolidation so much as an expansion in computing capacity. While this is true over the short term, since CIOs can then add lots more computing power to an existing data center without resorting to water-cooling and other exotic measures, over the long haul power will become a central issue. At some point, adding one more server forces a company to build a new data center. At the Red Hat Summit last week, John Williams, director of server and workstation product planning at AMD, jokingly called this the $10 million server, because that is about what an enterprise-class data center costs these days. Once companies squeeze efficiencies out of their existing data centers, provided their workloads are growing like crazy, at some point, power issues force construction. And CEOs and presidents are going to try to put that off as long as possible. That’s why I think it is wise to be proactive about virtualization and power issues now; you will get to look like a hero later.