The Q3 Server Market And IBM’s Place In It

December 9, 2013 Timothy Prickett Morgan

The figures are out for server shipments and revenues for the third quarter from Gartner, and they shed a little more light on what is going on in the IBM customer base and in the market at large. A number of different factors are making it tough on Big Blue and these have short-term and long-term ramifications for IBM i shops and indeed any customers who are deploying Power-based systems.

The most important thing to realize is that while server shipments are continuing to grow, revenues are not and I suspect that profit margins have fallen even faster than revenues. If this trend persists, we could see a massive shakeout in the server industry as vendors try to shift away from iron and toward software and services. This is, of course, folly. All of the major IT vendors sell systems, which means integrated collections of servers, systems software, storage, and networking, and the day they forget that and start thinking in terms of arbitrary business unit lines is the day they lose their minds.

So, for example, the fact that IBM talks about Software Group and Systems and Technology Group as if they were some kind of separate entities is just plain silly. To be sure, IBM has a fairly large database and middleware business that is sold for competitors’ systems, there is also a lot of its systems software that is sold on its own machinery and, in fact, much of this software is the only option available for its machinery. We could argue about whether hardware drives software or software drives hardware here, but even that doesn’t make much sense. People either invest in a system or they don’t.

The real problem with the X86 server market, from the top-tier server makers’ perspective, is not that it is commoditized, but rather that no one vendor controls it as a system. IBM, Hewlett-Packard, Dell, Fujitsu, and others engineer the systems and sell them, but Intel, Microsoft, and Red Hat get most of the profits. When the server makers cut price to compete in a sluggish market, they cut into their own margins and the dominant chip and operating system makers don’t really feel much.

Add to that the build-your-own-server crowd, like Google, Facebook,Baidu, and Amazon, and there is even more pricing pressure on the overall systems market. In this case, the hyperscale datacenter operators buy so much of the volume that they can command Intel and Advanced Micro Devices to give them custom chips and then they don’t generally engage the key server makers but original design manufacturers that work at margins that are more like retail. And, these big cloud buys tend to use their own operating systems, so there is no money in it for Microsoft or Red Hat, either.

What is a server maker to do except to try to stay in the X86 server game and make the best of their particular situations with other platforms? This is precisely the situation that IBM is in as rumors have been ongoing since earlier this year that it might be selling off its System x and BladeCenter server businesses, which no doubt had an impact on its X86 server sales in the third quarter. While IBM’s efforts to expand the use of Power chips in servers–through the OpenPower Consortium and by pushing more Linux applications on Power machines at prices that are comparable to X86 iron–are commendable, this was a strategy that IBM should have taken decades ago, not this year, if it was to blunt the X86 attack.

As you can see from the table above, this time last year Big Blue had more server revenues than HP, and this was the normal state of affairs for decades. In the third quarter of this year, despite respectable growth for its System z mainframe line, the revenue gap between HP and IBM widened considerably. IBM had a alarming 18.9 percent revenue drop, to $2.82 billion, while HP eked out 2.2 percent growth to $3.4 billion. But it was not all good news for HP, either. In that company’s most recent financial report for its fiscal fourth quarter ended in October, it was very clear that HP boosted its revenues and shipments by cutting into profits. Because IBM is most concerned about its profits, it may simply be walking away from deals that a newly aggressive HP is willing to make.

Whatever is going on in X86 Land, it is affecting Dell, too, which just recently went private and probably gave its customer base a reason to stop and wait to see how that all turned out. Dell has been growing revenues and shipments for the past several quarters, eating market share, but this time around its revenues fell by 3.5 percent to $2.03 billion and its shipments dropped precipitously by 14.1 percent, as you can see in the table below:

IBM’s overall server shipments fell by a stunning 28 percent, and if you drill down into these numbers, as Jeffrey Hewitt, the research vice president for servers at Gartner did with me, you will see that the System x business was the culprit, and in the United States in particular. Hewitt says that IBM’s X86 server sales were off 39 percent in the quarter, and worldwide, were down 29.8 percent. Some of that is companies and government agencies in China buying from indigenous firms like Huawei (up 202 percent), Sugon (up 61 percent), and Inspur (up 64 percent). Lenovo is also a Chinese company, but it does not do many hyperscale deals and only saw 1 percent shipment growth, according to Hewitt. IBM is not in the hyperscale deals that big Chinese firms are doing, either, and it is missing out on the opportunity there. If you can call it an opportunity when you do all the work and make no money at it. (I personally think it is OK to run a business at breakeven if you are doing good work. But I am not a public company beholden to shareholders.)

Gartner does not provide public data on the Power Systems business as a whole, so we have to use the AIX-Power combination as a sort of proxy. In the third quarter, IBM actually managed to boost its Power Systems shipments for machines running AIX as the dominant operating system by 19 percent to 14,611 machines. That drove $726.1 million in revenues in the third quarter, down 29.8 percent. IBM talking about future Power8 chips, which will have a lot more computing performance for enterprise workloads, probably didn’t help matters here, but not talking about Power8 might have made things worse. (Look at HP and the Itanium chip, for instance.)

The overall Unix market on RISC and Itanium processors fell by 31 percent to $1.29 billion and shipments were off 4.5 percent to 27,649 machines. IBM pulled up the shipment class average big-time. HP had a 21.9 percent decline in shipments of Itanium-Unix iron to a mere 3,855 machines, according to Gartner, and Oracle fell 22.3 percent to 8,039 boxes. That gave Oracle $265 million in Unix server sales and HP $243.7 million. IBM has won the Unix Wars, and in fact did so several years ago. But this market has no bottom until it finds new workloads above and beyond big database serving for ERP applications that run on fleets of Linux and Windows servers.

At this point, it doesn’t cost IBM all that much to keep IBM i and AIX going or to tune up Linux for Power chips. So don’t think that these steep declines mean the impending doom of these operating systems. That said, there needs to be a bottom that Big Blue can stand on, and thus far, we haven’t been able to find it. There is a certain level of revenue that is necessary for IBM to sustain its chip plant in East Fishkill, New York, and to continue to do processor design and system building. At some point, it may just be smarter to do what Unisys has spent the better part of a decade doing: gradually porting its MCP and OS 2200 mainframe operating systems to run in an emulated fashion on Xeon processors and yet still be able to run the same binaries that worked on Unisys CMOS mainframe chips. As long as IBM i applications run unchanged, this is what really matters to IBM i shops.

I don’t think it has come to this yet. But I worry when I consider how IBM is less interested in being a hardware supplier than it is in making its profit targets for the 2015 Roadmap. IBM should be interested in hardware, and had it focused on this, it might have created some truly amazing things in the past several decades. It is never too late to take another whack at it. I can envision IBM i running on an ARM server processor controlled by Big Blue and fabbed by GlobalFoundries or Taiwan Semiconductor Manufacturing or maybe even Intel now that it is throwing open its foundries to any and all. IBM must be sorely tempted to get out of the foundry business, and it would not be surprising to see IBM sell it off its fabs to any of the three companies above to free itself of the investment burden while still being able to design chips and build systems.