Two years after the debut of the first, beleaguered Itanium server chip, Hewlett-Packard ( HPQ) insists the time has finally come to start making good on the silicon it co-developed with Intel ( INTC). It won't be an easy sell. IT spending remains drowsy, and businesses must overhaul their software applications before they will run on the newfangled Madison architecture. On top of that, Itanium's brief but checkered track record is bound to hinder marketing. Yet in the longer term, the potential rewards for H-P are tantalizing. Though server sales have tanked over the last couple of years, the market is still massive -- valued at just under $50 billion in 2002 -- and it's expected to resume growing in the single digits over the next five years. "That's still a lot of money being spent," points out Mark Melenovsky, an analyst at IDC. "If H-P can attack that profitably and take share away from IBM and Sun, that's a great opportunity." Aside from the strategic potential for share gains, the latest version of Itanium offers a more assured benefit: H-P can stop funneling so much money into its own high-end PA-RISC and Alpha chips. On Itanium, it shares development costs with Intel, which manufactures the silicon. "Instead of us having to invest in a processor, or multiple processors, some of
the savings will go to the bottom line and some we can reinvest" in new areas of research, says Mark Hudson, vice president of H-P's enterprise servers and storage. Though he declines to say exactly how much H-P will save, one analyst has pegged the cost in the hundreds of millions. This being the sharkish hardware industry, of course, a rival is already vying to exploit the shift. Sun plans to make a grab for customers of H-P's discontinued proprietary chip lines, wooing them over to its own Sparc hardware. "We'll offer a sweet financial deal," promises Larry Singer, vice president of Sun's strategy office.
Dysfunctional FamilyMeanwhile, peddling Itanium won't be a cakewalk. Merced, the seemingly hexed first version introduced in 2001, was beset by delays and proved resoundingly unsuccessful. Its successor, McKinley, has been found to trigger
To be sure, while new buyers may be reluctant to make big investments now, it wouldn't be the first time they've had to rework applications for new technology. "When people ported from mainframe to Unix in the early '90s, that was a cost they had to eat," notes IDC's Melenovsky. "When you do long-range planning, you start looking at what platform offers the best return on investment over time." Case in point: Raymond James plans to run Itanium on Superdome, H-P's high-end server, for a new project to store terabytes worth of data from customers and financial advisers who work for the company. The company won't need to port anything again at the outset, but acknowledges it will have to modify applications on existing hardware that will later be shifted to Itanium. "This architecture is our strategic direction," explains Raymond James CIO Tim Eitel. "As technology changes, I think you have to be prepared to change with it." He says he was inclined to go with H-P's new offering in part because his shop was already stocked with hardware from Compaq and H-P. Yet H-P may find it much harder to win over entirely new customers to Itanium. "If it was you running a data-intensive
shop , which would you buy, an Itanium or IBM? Probably IBM," says Whalen. "It has a more robust system architecture, better software support and deeper software. Itanium is an odd duck" because it requires applications to be ported again, he adds. Still, H-P is gamely hoping to steal some share from IBM's pSeries chips, based on Big Blue's Power architecture. H-P argues that in a test of performance, Itanium cranked through more transactions per minute. IBM doesn't dispute the numbers, but says it will come out with its own upgrades as market demand develops.