Le Misancalculation: A one act Itanium tragedy by IDC
Deus ex muckupina
As the artist's rendering points out, IDC does not have a sparkling track record where Itanium sales predictions are concerned. In fact, its 2004 prediction was 1,700 per cent above the actual sales mark. Some analyst firms can survive on results like that, and some can't. (As a courtesy, we're not even going to mention IDC's original 1999 Itanium sales forecast that put revenue at $33bn - !!!!!!! - in 2002.)
IDC this week told the world that it sees Itanium sales growing at about 35 per cent per year to reach $6.6bn in 2009. The major driver behind this forecast for Intel's 64-bit chip stems from a recent announcement by the Itanium Solutions Alliance (ISA). This grouping of Itanium backers held a bake sale and managed to pull together $10bn to support the chip over the next five years.
"IDC believes that the formation of the ISA may ultimately prove to be one of the most significant elements in driving future sales of Itanium servers," IDC said in its report.
"The ISA provides an opportunity to unify marketing messages for Itanium across the multitude of constituents involved, and to deliver a cohesive and well-thought-out story to entice customers to the Itanium architecture as they consider a variety of server platform alternatives for their IT organizations. In essence, the ISA represents the best opportunity for creating an Itanium brand in the marketplace that aligns the capabilities of the technology with customer requirements and meets expectations."
Given that two of the IT industry's largest companies - Intel and HP - have been hammering away on Itanium marketing and heaving billions at the chip and customers for the past ten years, we can't see why IDC would be so impressed that the likes of SGI, Bull and Unisys have decided to chuck about $200m per year at the chip too. SGI's market moving influence ain't what it used to be. And, let's face it. About half of the vendors in the ISA probably won't exist in five years.
IDC, however, disagrees.
"The ISA will aid in bringing attention to the platform in a manner that the smaller server vendors (compared with HP and Intel) would not be able to accomplish on their own."
And still, we stand by our statement. In IDC's own report, for example, you'll find HP mentioned 88 times. Intel earned 9 mentions. NEC got 5, SGI got 3 and Bull grabbed only 1 mention.
That's an interesting spread when you note that HP took 70 per cent of the $1.6bn in 2004 Itanium server sales, while NEC grabbed 11 per cent, SGI grabbed 8 per cent, IBM grabbed 4 per cent and Unisys grabbed 3 per cent. The mentions to sales ratio matches up well enough, and we expect that will hold true for years to come, since HP customers are the ones who really need to move to Itanium.
With that in mind, let's take a closer look at how Itanium's future might stack up from HP's perspective.
Today, HP sells about $4.8bn worth of Unix boxes per year with the vast majority of those sales coming from PA-RISC servers. In an ideal world, HP would transition 100 per cent of Unix sales over to Itanium and then pick up new Linux and Microsoft business. IDC happens to think "Longhorn" Server on Itanic could be a real hit.
Under IDC's scheme, however, HP doesn't look to be doing that great with the Itanium transition - even by 2009. It will have to split the $6.6bn in sales with all Itanic vendors. Even if it maintains 70 per cent share of the Itanium market, HP will just be selling $4.62bn worth of Itanium servers in 2009. In this perfect world scenario, Itanium serves as a mediocre PA-RISC replacement at best.
Add to that IDC's idea that the other ISA vendors will sell more Itanium servers over the next few years. That means that HP's 70 per cent share should drop, according to IDC, although we suspect it will actually rise.
IDC seems to ignore its own evidence to conclude that HP is in better shape than expected.
"The current market perceptions of Itanium-based servers are quite positive," IDC said. "Both HP and non-HP customers display high levels of awareness of the platform, and market perceptions do not appear to reflect the negative press that has recently been part of the public domain."
IDC points out that HP customers are some of the most loyal around and insists that most of the flock will move to Itanic. But its numbers don't back that up.
Only 70 per cent of HP's customers know that the company is phasing out PA-RISC in favor of Itanium, according to IDC. That's not great. You might expect 100 per cent to know after a decade of HP and Intel harping about this transition.
Then, IDC tries to put a pleasant spin on where the non-Itanic followers might go.
"The more important news here is that the PA-RISC customers who do not intend to migrate to Itanium are more likely to choose a standard architecture moving ahead rather than another proprietary platform such as IBM POWER or Sun SPARC," IDC said.
That's not exactly IDC's shining moment of clarity. Looking at the detailed results, you'll find that 36 per cent of PA-RISC customers plan to move to Xeon, 25 per cent plan to move to Opteron, 25 per cent plan to move to SPARC and 11 per cent plan to move to IBM. Sure, x86 takes the lead, but losing 36 per cent of sales to Sun and IBM is horrible news for HP and seeing 61 per cent of sales go to a non-Intel platform is even worse for Intel.
Of course, the IDC figures are as confusing as ever when you keep examining the numbers on which vendor PA-RISC customers plan to move to. Close to 22 per cent of customers picked HP, 17 per cent picked Sun, Dell and IBM - and the rest fell into white box or "don't know." So, IDC is telling us that 25 per cent of PA-RISC customers plan to move to SPARC but only 17 per cent of PA-RISC customers plan to move to Sun. That's a neat trick. (This would only be possible if Fujitsu picked up a lot of the PA-RISC business. Given Fujitsu's small market share and the fact that IDC talked to US customers only, such a scenario seems close to impossible.)
A footnote at the end of IDC's study says that the margin of error ranges "to around (+/-) 18 per cent of small groups such as of PA-RISC respondents migrating to an alternate platform."
So, er, anywhere between negative 1 per cent and 35 per cent of PA-RISC customers plan to move over to Sun. And these are the exacting figures IDC uses to provide the "more important" news.
Nowhere in the IDC study does the analyst firm address the negative effects that IBM and Dell abandoning the Itanium market might have on the chip. Instead, we're told that investments from SGI, Bull and others have proved strong enough to counter negative impressions of the chip. In fact, much of IDC's thesis hinges on the idea that "Consistency in Messaging Will Be Key to Future Itanium Adoption." And here we thought sales mattered.
IDC also fails to discuss Intel's roadmap slips in detail. And it doesn't address the affect a rising interest in 64-bit x86 boxes will have on the high-end server market come 2009. Why bother?
The truth about Itanium is that it has been doing quite a bit better over the past 18 months - for HP. That's no surprise, since HP stopped producing fresh batches of PA-RISC chips. The migration is on.
For other vendors, however, Itanic remains a tough, tough sell. Despite massive efforts to port software to the chip, many high profile customers remain dissatisfied. In addition, not having Sun, Dell or IBM on board makes the product look pretty bad. You can't have an industry standard when you're missing about 65 per cent of the industry.
IDC never explores these issues. Instead, it relies on flaky data and embarrassing conclusions to create a study that could outdo its 2004 effort - a feat we thought impossible. ®
Sponsored: DevOps and continuous delivery