The first of a two-part article by Timothy Prickett Morgan.

IBM held a meeting in Beantown recently to make the case to industry gurus that it has not only straightened out its server marketing strategy, but that it has also figured out how to actually start growing its server businesses and, perhaps, even grow faster than the server industry at large for a nice change. The arguments that IBM presented in favor of such growth are as complex as its customer base, but it all comes down to this: IBM is going to figure out what customers want before it starts selling it to them, sell those products to its customers and then support the living daylights out of them to not only keep those customers happy, but keep them coming back for more IBM goods and services.

New markets

This new strategy sounds simple enough, but it is actually quite complex to implement and even if IBM does get it to work, it will not start working overnight. It will take time for IBM to gain the knowledge it needs and to forge the partnerships that will be essential for it to move into new markets, as it plans to do this year. It will take time for IBM to train its personnel and for them to get used to the new way that IBM sells and supports its products. More importantly, such a broad and sweeping strategy as IBM is undertaking will have to be modified as conditions change in the marketplace. That said, it seems that Bill Zeitler, IBM’s new server marketing czar, and his marketing team have finally figured out what everyone else has known all along – IBM’s server line is too complicated and customers often feel that IBM doesn’t know what it is doing. This sort of thing does not inspire confidence, especially when there’s a Hewlett-Packard or Compaq sales rep knocking on the door with a simple plan every other day. IBM is getting its server act together just in time. According to statistics provided by IBM as well as other market analysts, IBM continues to dominate the server market, if having a little less than a quarter of yearly sales can be said to be a dominant position. For 1997, IBM estimates that it had about 23% of the $50bn worldwide server market with nearly $12bn in server sales. (These numbers do not include the sale of peripherals and software for those servers.) IBM’s biggest competitor, the new Compaq with Digital Equipment and Tandem thrown in, had about 11% of the market with $5bn in sales. Like IBM, this new Compaq has a complex set of product lines that are often at loggerheads. Hewlett-Packard came in third with about 8% of the market and $4bn in server sales and Sun Microsystems was fourth with 7% of the market and $3.2bn in sales. The big three Japanese vendors – NEC, Fujitsu and Hitachi – plus all the other server makers like Dell, NCR, Gateway and myriad smaller firms accounted for the other $25bn in sales (no vendor in this pack has more than 5% market share, however.)

Losing ground

IBM may have the biggest market share in the server arena, but it is losing ground. Big Blue is the first to admit that it hasn’t seen the kind of growth that Compaq, HP and its other competitors have enjoyed in the past few years. IBM’s chairman, Louis Gerstner, says that, due to its size, there is no way IBM for IBM to grow as fast as smaller players. This opinion is just plain wrong. IBM could and can and may grow as fast as Compaq and HP – it will just have to change what it does to accomplish the task. IBM may not mend its marketing ways and therefore not gain market share, but its size will not be the reason why it won’t grow. Protecting one IBM server base – say the S/390 mainframe or AS/400 mini bases – at the expense of others – like its Unix and PC server lines is what has always limited IBM’s growth, and it could continue to do so in the future. This attitude, more than anything else, explains why IBM’s server revenue has grown only in the single digits over the past few years while HP, Sun and Compaq have enjoyed 18%, 19% and 20% server revenue growth. This year, Bill Zeitler says that AS

/400 and RS/6000 server revenues are already in the double digit growth range; now all he has to do is keep it up and get PC and mainframe revenues growing at the same rate to accomplish the task that Gerstner has set out for him – double-digit server revenue growth on par with IBM’s competitors.

e-commerce and groupware

If the market data compiled by Merrill Lynch in a recent server market report is any indication, IBM has its work cut out for it. Merrill basically believes that the Unix market is going to continue to gain market share from AS/400s and mainframes as well as eat up all the extra business that comes from supporting new workloads such as e-commerce and groupware; if Unix platform providers do not get that business, Merrill figures that Windows NT server sellers will. In 1995, Merrill says that NetWare accounted for about $3.8bn in server sales, about 10% of the $37bn worldwide market. (The Merrill numbers include sales of full system configurations, not just raw servers.) Windows NT servers were half that at $1.7bn. Unix servers accounted for a whopping 46% of sales at $17.3bn, and Others – which includes AS/400s, S/390s, OS/2 servers and DEC VAXen – had about 40% of the market with $15bn of sales. By 2002, Merrill expects the server market to more than double to $81bn; NetWare will more or less hold with its 10% share and Unix will more or less hold with 46% share. But the proprietary Other category will actually shrink from $15bn to $10bn, dropping to 12% of the 2002 server market, and Windows NT will explode to account for $27bn in sales or 33% of the market. These numbers indicate the kind of pressure that is on IBM’s mainframe and AS/400 businesses. Not that these machines do not have their advantages. They do. Customers want 99.999% availability – the so-called five nines – that translates into less than five minutes of downtime a year. On a good day, NT servers offer 97% availability, which equates to about 11 solid days of downtime a year. Mainframes, through their crafty through archaic systems software and tight coupling of that software with error and crash protection hardware not found on any other systems in the world, offer 99.9% availability in standalone configurations (about 9 hours offline) and five nines in parallel sysplex clusters (five minutes offline). Customers have to cluster Unix servers to get the same availability as is available in a standalone S/390. Unix servers offer about 99% availability – which means they are down about 4 days a year – in standalone configurations; single AS/400 servers are on par with standalone mainframes as far as availability is concerned, being offline 8 to 9 hours a year. Obviously, Unix and NT vendors have some work to do on their systems software in this arena. That’s why HP is working with Hitachi and NEC to develop self-healing implementations of HP-UX that will give that Unix variant fail- safe capabilities similar to those found in IBM’s flagship MVS mainframe operating system. In looking at the entry and midrange server markets alone – the ones that the AS/400 is a dominant player in – IBM faces even bigger challenges. By 2001, Merrill expects that the NT server market will be twice the size of the proprietary server markets (which is split pretty evenly between the DEC VAX and IBM AS/400 bases). Merrill similarly expects that companies will invest more than three times the amount they spend on AS/400s and VAXen in Unix servers. E-commerce, the big driver behind server sales in the coming years, will apparently play into the hands of the Unix vendors and Bill Gates, not IBM, at least in Merrill’s opinion. There is a good reason why Merrill expects that Unix will dominate the midrange despite the best efforts of Bill Gates and IBM. By the time Microsoft gets its NT house in order, the few remaining flavors of Unix – Solaris, HP- UX, SCO and maybe Digital Unix and maybe AIX – will be mature operating systems with most of the benefits of mainframe operating environments. These big Unix boxes will provide today’s mainframe functionality at prices even lower than PC servers. And that, more than any other factor, will drive the Unix market to new heights. In Merrill’s future server market, most new applications will go on a Unix server unless they are more appropriately supported on an even cheaper Windows NT box. Of course, this is not exactly how IBM sees events unfolding. While IBM believes that both Unix and NT have a big place in its server strategy, it does not by any stretch of the imagination believe that customers are about to abandon their AS/400s and mainframes. Au contraire, IBM believes that it will be able to grow these businesses just as fast as it can grow its Unix and NT server lines. IBM’s growth in the mainframe market will pretty much have to come at the expense of Hitachi and Fujitsu, and with the new G5 series of CMOS mainframes, which offer nearly the same power as Hitachi’s Skyline bipolar mainframes, IBM can make a very convincing case for a G5 machine. During 1999, IBM could even see an uptick in mainframe sales (MIPS growth has been steady for years, but prices fall so fast that IBM has seen revenue declines for many years). Moreover, with the G5 machines, IBM will be able to convince many customers who might have gone to a Unix box to support PeopleSoft, SAP R/3 or Baan ERP suites on a mainframe instead. there won’t be thousands of such customers, but rather several hundreds, and that could add up to an extra $500m in revenue for IBM’s mainframe line. That would be a very large increase, and one that would go straight to the bottom line. There are also thousands of mainframe customers who have resisted moving to IBM’s CMOS mainframes – and for no good reason, either, except they don’t want to be bothered. Most of these customers are, like customers using vintage AS/400s, wasting bundles of money supporting their old equipment. For the same money, they could have a more modern machine and use more modern software.