By Timothy Prickett Morgan

Imagine my surprise last week when I received two confidential files outlining the competitive analysis that the special marketing teams who watch Hewlett- Packard and Sun Microsystems put together to help IBM’s sales reps and business partners sell IBM gear against HP and Sun equipment.

While I don’t think the documents I received are the sum total of IBM’s competitive analysis – at least I hope not for IBM’s sake – the documents are nonetheless informative about how IBM thinks about the competition in the Unix space and the information that it is relying on to characterize the server and workstation markets for the people who have to sell IBM kit.

By IBM’s reckoning, Sun really has most of its influence in seven market segments – ERP, supply chain, front office, data warehousing, design automation and engineering, software development and e-commerce – spread across five industry segments – retail, energy, healthcare, publishing and service providing. To a lesser extent, IBM is telling its partners that Sun has some presence in telecom, manufacturing, financial, education and government institutions. IBM’s competitive analysis team thinks that about 35% of Sun sales are for commercial workloads, 30% are for technical applications, 23% are for software development, and the remaining 12% are for industrial and other uses.

IBM reckons that Sun’s direct sales team is focusing on the top 1,500 accounts in its installed base, and that its remaining customers are being serviced by business partners. IBM is also touting the Unix server and workstation market share breakdown done for 1998 that shows Sun in the first or second spot in entry, midrange and high-end Unix server markets and in the number one spot for workstations (Unix plus NT) in terms of revenues but behind HP in terms of units shipped. The IDC numbers for the server markets at large, which include AS/400, S/390 and PC server revenues as well as Unix server revenues, show that IBM had a considerable lead in 1998 at the high end of the server market with 37% of the $16bn market compared to Sun’s fifth place position behind Hitachi, Fujitsu and NEC with its 5.6% of the market, Sun was, however, growing at 108% compared to IBM’s 17% growth and a 30% or 40% decline compared to the Japanese giants. Depending on how well these companies do this year, Sun could very well be the number two high-end server vendor behind IBM in 1999. IBM, says IDC, had 28% of the $18bn midrange server market in 1998 compared to Sun’s 13%, which was considerably behind HP’s 23%. The IDC numbers show IBM with 21% of the $25bn entry server market in 1998, with Compaq getting 18%, HP getting 12% and Sun getting 11%.

IBM is using a breakdown of Sun’s financials provided by Morgan Stanley Dean Witter that shows Sun growing from $8.6bn in sales in 1997 to $13.9bn in 2000. MSDW’s estimates show Sun’s entry server business growing by 21% to $2.5bn in 1999, its midrange server line growing by 17% to $1.19bn and the high-end Enterprise 10000 line growing by 126% to $316m. (None of those server numbers include storage or software). MSDW is projection flat revenues for the Starfires in 2000, but that entry server revenues will continue to grow to $2.9bn and that midrange servers will grow slightly to $1.21bn. MSDW reckons that Sun’s workstation revenue will decline by 17% to $1.7bn in 1999, and will decline further to $1.5bn in 2000. Storage revenues are pushing $3bn, up 40% from last year, and will break above $4bn next year if MSDW is right. The remaining $3bn in revenue is spread across Solaris, Java and other software sales, sales of microprocessors and service offerings. Services is growing at close to 40% per year, while software sales are growing at slightly lower rates as Sun on the whole – Java at 18% between 1998 and 1999, Solaris at 15%.

While IBM’s competitive analysis team says that Sun has good NT street fighters in the $5,000 to $20,000 price range with its Ultra 5S, Ultra 10S, E250 and E450 servers and that its E3500s, E4500s, E5500s, E6500s and E10000s offer mainframe-like reliability and scalability for prices that range from $37,000 to $2.5m, Sun’s sales reps are still too focused on selling iron, not solutions, because they want a shorter sales cycle. The one weakness IBM sees in the Sun line, especially on the entry servers, is that Sun doesn’t offer any native NT support. (IBM Netfinities can obviously run NT, and AS/400s have special inboard PC server cards called the Integrated Netfinity Servers that can run NT but which use AS/400 disk arrays and management software for storage.) Sun, of course, would say that not having native NT support was a plus, not a minus.

The IBM competitive analysis team is also apparently taking Sun’s foray into advanced disk arrays very seriously. IBM’s team points out that Sun will initially try to capture the storage associated with Sun’s own server sales, then move into other accounts and then offer true information storage across multiple platforms through Fibre Channel, much as IBM’s own Storage Systems Division is promising with its Seascape products. Sun is the number four storage vendor behind Compaq, IBM and EMC.

Over the next five years, IBM expects Sun to offer mainframe- class security and logical partitioning and that it will make good on its promises of having NUMA scalability that goes well above 1,000 processors in a single system image. IBM’s competitive analysis team also expects that Sun will build a professional consulting and services organization that will provide the full spectrum of services, from entry servers to large data centers. IBM apparently also is expecting that Sun will lose the low end of the workstation business to Wintel machines and that it will be forced up the food chain to offering very high-end graphics workstations and other value adds to stay in that market. IBM also concedes that Sun has clearly positioned itself as the computer industry’s standards visionary, that it has plenty of talent at AOL and Netscape to tap to expand its presence in e-business and that it is still positioned as the primary server supplier for ISPs, ASPs and other kinds of service providers. These IBM documents pre-date the most recent RS/6000 and Netfinity server announcements, which shift some of the ISP business back towards IBM.

The IBM competitive analysis team for HP did a lot more of its own work trying to suss out its competition. The team reckons that of the $39.47bn in revenues HP got in 1998 under its computer organization (distinct from the $7.6bn test and measurement business that is being spun off), $660m of it was for HP 3000 servers (down 8% from 1997), $2.82bn was for HP 9000 Unix servers (up 13%), $2.92bn was for workstations (down 7% thanks to an imploding Unix workstation business), $7.64bn came from PCs (up 13%), $1.56bn was for software (up 15%) and $1.6bn was for systems integration (up 19%). Printers, peripherals and supplies accounted for another $19.4bn (up 14%) and customer support brought in another $3.48bn (up 5%).

IBM quotes estimates from Salomon Smith Barney that say HP has gross margins of about 43% on HP 9000 servers, about 30% on PC servers and 25% on storage and workstations. IBM believes that HP gets 41% of its business from enterprise customers (more than 1,000 employees), 21% from medium sized companies (100 to 999 employees) and 14% from small businesses, with the remaining 24% mostly for PCs and printers going into the consumer market. IBM figures that in 1991, HP pushed 40% of its sales through indirect channels, and that last year the indirect channel grew to 63% of total revenues. Apparently HP had hoped to get this to 80% by this year, but won’t reach that level until next year by IBM’s estimates.

>

>Nq The IBM team went through the litany of recent HP announcements for e-business and apps-on-tap among other things, and took particular aim at HP’s five nines high availability campaign as a weak point, IBM says that the 5Nines:5Minutes campaign is all just a vision, that HP can only guarantee 99.95% availability today (which is equivalent to a standalone AS/400 with no extra mirroring or software features), and that its high availability offerings are for only unplanned downtime for a limited selection on bundled, pre-configured, pre-tuned and pre-tested HP 9000 mirrored servers. IBM says that the five nines deal only has a one-month service penalty (worth $12,000), which has very little business value to customers, but the campaign has great marketing value for HP

IBM is apparently trying to make as much hay out of Merced’s delays and the fact that HP had to extend its PA-RISC roadmap to 2003. IBM reckons that Merced will run at 733MHz and offer no significant performance benefits over PA-8700s running at 720MHz due in early 2001, and maybe, depending on how the HP-UX compilers work, not much of a boost compared to the 560MHz PA- 8600s due be fore year’s end. IBM is also touting Monterey/64 as the preferred Unix on IA-64, having essentially stolen whatever pole position HP was going to have with HP-UX on Merced and follow-on chips.

IBM concurs that by mid-1999 HP had installed 2,000 V-class servers in just under two years, with over $1bn in revenue derived from the boxes in 1998. IBM says that the average V-class server sells for about $700,000. IBM believes that the V class plays particularly well in financial services, manufacturing, telecom, government and health organizations, and that the boxes typically support OLTP, server consolidation and mission critical jobs as well as the standard ERP, SCM, and technical computing areas common in the Unix market. IBM says that HP has tremendous support from the ISV, consultant and systems integration communities. That said, IBM is leaning rather heavily on studies by DH Brown, which ranks AIX as the number one Unix operating system over HP-UX’s number four position when it comes to reliability, availability and scalability, and which also ranks IBM’s HACMP above HP’s MC/ServiceGuard for Unix clustering.

IBM says that HP is still, with such low software and services revenues, primarily a hardware company and that it may have lost 15% of potential customers last year because it lacked its own support and consulting services and is heavily reliant on EDS, Andersen Consulting, Cisco and HP resellers. IBM also says that HP’s $1.6bn acquisition of Veriphone two years ago yielded nothing. รก