Mad Dog 21/21: Not As Big Blue
May 18, 2015 Hesh Wiener
In 2011, IBM‘s revenue exceeded $106 billion. Every year since, the company’s intake has fallen, sliding 10 percent to under $95 billion by 2014. Analysts at investment bank Credit Suisse predict IBM’s 2015 revenue will drop another 9 percent to $86 billion. The CS analysts believe IBM’s revenue will dwindle to $81.6 billion by 2018. Despite this, most investors are hanging tough. The same cannot be said of IBM’s customers, who are impressed by superior systems engineering not clever financial engineering; they understandably see IBM’s failures in the server business as a danger.
In 2014, according to IDC, IBM’s server business was second in monetary size to that of Hewlett-Packard and slightly larger than that of Dell. But IBM’s 18.4 percent share of the total server market included, for the first three quarters of the year, the X86 server business that was sold to Lenovo. By zooming in on the year’s fourth quarter, observers on and off Wall Street can see a substantially different picture. IBM’s share of the server market in the fourth quarter was half the size of HP’s slice and quite a bit smaller than that of Dell. IBM remained ahead of some other key server suppliers, including Lenovo and Cisco Systems, but even that lead is suspect. IBM’s glass house servers are often sold during each year’s fourth quarter. This means that IBM’s market share during most if not all of 2015 could well be less than it achieved during the final quarter of 2014.
IBM’s shrinking market share affects customers both directly and indirectly. When any type of server loses popularity, customers find that the pool of expert talent diminishes; sources of parts and service, particularly third party sources, dry up; opportunities to trade in or remarket unwanted equipment became scarce; software developers and support companies are tempted to put their best efforts into more popular market segments; the direct and indirect competition that keeps IBM on its toes fades away; and computing professionals begin to migrate to situations that provide more promise and opportunity.
In the case of glass house servers, the process is likely to be very slow. The history of former IBM systems such as the System/34 for small business and the 24- and 31-bit mainframes formerly used by big business shows that old equipment that performs valuable (sometimes irreplaceable) work and applications will stay in use for years, even decades. But customers inevitably move on, and when they do finally migrate, not only the old equipment but also personnel whose talents are wedded to that equipment become subject to disposal.
Until quite recently, IBM never discontinued a line of systems until it had in place forward migration paths for its customers. So in the case of small business, the System/32 wasn’t killed off until the System/34 provided a suitable upgrade path. The System/34 similarly gave way to the System/36, the System/38, and ultimately the AS/400. Sure, customers didn’t always make the leap, but IBM managed to achieve a very high migration rate time after time. This was true even in cases where the transition was pretty difficult for many users, as was the case when IBM decided to press users of the System/34 and System/36 toward the System/38. Similar migration issues arose in large systems market, where the precursors of the System/360, such as the IBM 1400 and 7090 lines, were discontinued and customers were obliged to either migrate or leave the IBM fold.
IBM’s methods worked very well for a very long time. IBM led the computer industry and enjoyed a very high customer retention rate as computers evolved during the 1960s, 1970s, and 1980s. But during the past twenty or so years there was a sea change in the server business, one that IBM failed to fully understand. That change grievously, perhaps fatally, wounded IBM’s server business and damaged the software and services operations tied to its servers.
What happened was that servers based on X86 technology began to augment and later displace IBM’s alternatives. These servers first showed up in departmental offices rather than glass house data centers as PCs supplanted green screens. When IBM saw the effects of an apparently irresistible force, it decided to join the crowd rather than fight. IBM was able to supply X86 servers and even to get its machines recognized as top quality products. But it was not able to exercise account control, as it did with proprietary computer families. IBM didn’t have suitable server software of its own, weakening its grip. At its best it resembled a flabby Dell or HP.
To make matters worse, IBM failed to adequately appreciate how disruptive X86 equipment, with its low cost and wide choice of suppliers, could be once it made it into the glass house. Only now, long after the most ambitious users, such as Google and Amazon and eBay fled IBM’s world, long after IBM was forced out of the chip business, has the company decided to try opening up its Power architecture to other hardware makers. While a number of companies are pleased by gaining access to IBM’s excellent technology, there is no sign their efforts will help the Power architecture become significantly more popular. Power isn’t going to challenge X86 or ARM. More likely, chip makers will cherry pick IBM’s ideas and use what they learn to improve circuitry that will ultimately compete with IBM’s proprietary architectures. But even if some chip makers decide to produce Power-compatible circuits, IBM is unlikely to enjoy new opportunities. Years ago IBM’s processor chips (or more accurately chips from both IBM and Motorola) powered Apple desktop and laptop computers. Apple switched to Intel chips for Macintosh machines and adopted ARM chips for iPads and iPhones from the get-go . . . and never looked back.
If the server market in 2015 and beyond looks like the market during the fourth quarter of 2014, IBM will have been reduced to an also-ran in hardware. If contrary to our expectations some vendors in addition to Chinese PowerPC chip maker Suzhou PowerCore license the Power architecture and create servers that can run IBM’s AIX or IBM i software, IBM’s hardware market share will be further reduced. (It seems unlikely IBM will allow such chips to run its own operating systems.) Additionally, as IBM reviews its designs for storage controllers, which still include some Power technology, the force of pro-Power politics inside IBM is waning, and in the end IBM might see that X86 or ARM or MIPS processors will help it create more competitive subsystems. Similar considerations may ultimately drive IBM to consolidate its mainframe and Power hardware, or to move all its server technology to more economical platforms.
IBM’s share of the server market is heading toward 10 percent. Its cloud operations increasingly uses systems acquired from Taiwanese or Chinese manufacturers. Eventually customers are bound to believe IBM when it says its future is the cloud rather than the glass house. As a result, IBM’s server business might not go through a graceful, gradual decline. Instead, IBM’s hardware operations might suffer a catastrophic collapse.
At first, the impact is likely to be confined, because users of IBM Power and mainframe machines will continue to pay for software licenses and stick with their services arrangements. But that can only go on for a few years. IBM will try to keep customers marching forward. Mainframe shops will be propelled by software licensing arrangements that force users to move to newer platforms every three years. Power platform users will face similar pressure wherever IBM can exert its influence. But such customers are bound to be unhappy living in a state of intimidation. As opportunities arise for them to move to more competitive parts of the IT industry–into the X86 world or to the cloud, whatever the iron inside it–they will break loose.
And once they unplug the familiar systems they have long used, they won’t turn back. Many will reassign or perhaps dismiss technical experts that know how to run legacy systems. Some will shift the legacy workloads they don’t know how to move onto machines run by service companies, IBM included, while maintaining a core team of legacy systems experts to manage their newly remote computing. And the remainder of the IBM hardware customer base, becoming smaller and in some ways an exotic relic of bygone times, will cease to provide a significant business opportunity for IBM.
IBM could emerge from this transition with new vigor and wide horizons. But it won’t be the kind of company that old-timers think of as IBM.