But how long can it prosper by filling gap left by IBM?

When Steve West took over as CEO of Santa Clara, California- headquartered Hitachi Data Systems (HDS) in 1996, he must have had serious doubts about the very basis of the company’s business. As the largest supplier of IBM plug-compatible mainframes, distributing systems everywhere except Japan, the model was to mimic any IBM product within 18 to 24 months of its introduction; recycle the product line every four to five years; and improve performance by about 15% to 20% a year. But by the mid-1990s, that model was breaking down: many customers were no longer willing to wait for HDS (or the other major PCM vendor Amdahl) to catch up with IBM’s technology innovations; product cycles were shrinking to two to three years, and price/performance improvements had to be more like 15% to 20% a quarter. At the same time, HDS was wrestling with a two-headed technology problem. It was having to deal with a fundamental change that IBM was making to its mainframe architecture – the switch to commodity CMOS processors which are less expensive to both manufacture and run. And, alongside that, it had to embrace IBM Sysplex technology, the introduction of which allowed multiple mainframes and CMOS processors to be hooked together as if they were one system. That flux was reflected clearly in HDS’s revenue picture. After years of steady growth, revenues stalled in 1993 at around $1.6bn and stayed flat for another three years. But Hitachi, without fully realizing it, was sitting on a goldmine. The company bet that IBM’s 1992 decision to switch away from the traditional water-cooled, ‘bi-polar’ mainframe processor architecture to CMOS technology was premature. With the introduction of its Skyline range of mainframes in April 1995, Hitachi reckoned it could soak up on-going demand for bi-polar machines for at least three years. There was sound logic behind that decision. CMOS-based systems, despite their potential, were not capable of providing the raw processing power that bi-polar could muster. And the CMOS/Sysplex architecture required users to convert their applications – something many were unwilling or unable to do.

HDS revenues

But the seam Hitachi hit was much wider and deeper than even the company had hoped. According to Frances Walker, VP in charge of the Skyline Business, in the six months to September 30, the unit shipped 120 Skyline mainframes worldwide (with an average of about four Skyline processor ‘engines’ per machine). That is about a 30% to 40% growth, says Walker, something the company expects to repeat in the second half of its fiscal year. The hit product has changed perceptions at HDS. Executives have even stopped saying that bi-polar is a stop gap technology. Skyline, is here to stay, says Walker. The largest bi-polar engine today at 150 MIPS [millions of instructions per second] is still two and one-half times more powerful than the largest CMOS engine from IBM. Mid to late 1998, CMOS will grow to 85 MIPS and to 124 MIPS in 2000, but Skyline can still be twice as fast, he says. With the only bi-polar system available, HDS charges roughly 15% more for a Skyline machine than an equivalent IBM CMOS mainframe. It was clear even as far back as late 1992 that there would continue to be some demand for bi-polar upgrades from customers who were unwilling to rework their applications. Back then, IBM made the decision to go to CMOS under extreme pressure to bring mainframe pricing more in line with client/server systems. But what has surprised analysts, IBM and HDS itself, was the desire by users to simply add more and more power to their current applications. We took the low hanging fruit in the first year of Skyline shipments and saturated that initial [upgrade] market, says Clif Warner, director of marketing for Skyline. As customers became more conscious of the benefits of Skyline, then we opened up a new market [of users simply needing more and more processing power]. According to Walker the notion that these are just traditional HDS customers is not true: 52% of shipments are to new customers, many the deepest shade of IBM blue, so our expectations for growth is still very strong. That has made Skyline HDS’s flagship product. Sales have surpassed the $1bn a year mark, says Walker, and now account for more than half of HDS’s total revenues. The company’s mainframe market share has also ballooned – from 7% in 1995 to 22%. However, some analysts think the Skyline bubble will burst sooner than HDS expects. Says Bob Djurdjevic of Annex Research: They had a two year window before more powerful CMOS machines appeared from IBM and before Sysplex was widely accepted. We are at that point now. It is going to be a tough 1998 for them. Djurdjevic points out that the performance peak of a bi-polar system is around 1,200 MIPS, as a maximum of eight engines can be hooked together. In comparison, the CMOS architecture with parallel Sysplex can employ hundreds of processors and produce a (theoretical) performance of 15,000 MIPS. Of equal importance, at some point, the sheer cost of manufacturing bi-polar chips will outweigh the power advantage that they have over CMOS – a point made more pertinent by the ferocious mainframe price erosion that is cutting the price per MIP by about 2.5% to 3% a month. As prices continue to decline, HDS has less and less room to absorb the shock of that, while CMOS manufacturing can, says Djurdjevic. Skyline is a one time event, the ‘Spruce Goose’ of mainframes, he says, referring to the single flight of the mammoth aircraft built by media magnate Howard Hughes in 1947. If that is the case, and the Skyline phenomena fizzles out around the turn of the century, at least its success will have bought HDS some space to build alternative revenue sources. The company has its own line of CMOS mainframes, the Pilot Series, which uses CMOS processors sourced from IBM. However, that agreement is due to come to an end in 1998. After that, HDS executives hint privately, Hitachi will be ready with its own CMOS chip. HDS also has a booming large systems disk business. IDC says that HDS is continuing to steal market share from IBM in this sector and will see its slice of the market grow by 6% to 19% in 1998. At the same time, CEO Steve West has plans to diversify the company’s business beyond the mainframe area. His plans include the launch of HDS into the midrange server market with a range of Unix and NT machines. There is also considerable scope for further expansion of the HDS systems integration operation, he says. But those initiatives are not going to fill the kind of gap that Skyline will leave when it goes. To survive in the mainframe market, Hitachi will have to craft a winning CMOS strategy. As Walker says: Life is pretty good right now for HDS, we hope it will be good for ever. It is going to take more than hope.

From Computer Business Review.