What The System x Selloff Means To IBM i Shops

January 27, 2014 Timothy Prickett Morgan

As has been rumored for nearly a year and suspected for the past five years by many of us, IBM has finally gone and done it and sold off its System x server business to Lenovo Group for $2.3 billion. The details of the deal are complex, and there is no guarantee that the U.S. government will approve it, but given the strong presence that Lenovo has in North Carolina by virtue of its acquisition of IBM’s PC business in late 2004, the odds favor the nod from Uncle Sam.

With that move, Lenovo becomes the number three shipper of servers in the world and starts a many-year race to try to catch and pass Hewlett-Packard and Dell. It has taken Lenovo nine years to accomplish that feat in the PC market, which Lenovo has learned to make money from thanks to its low-cost manufacturing, its wringing of costs out of its supply chain, and its dominance in its indigenous Chinese market and strong presence in North America and Europe thanks to the IBM PC deal.

I went through the details of the IBM-Lenovo deal at my other publication, EnterpriseTech, last Thursday when the deal was announced, and I also did a piece in HPCwire about the ramifications for Lenovo taking over the X86 server and switching businesses from IBM for the supercomputer market. Frankly, I would appreciate it if you would give them both a good read so I don’t have to go over all the same material here. (There’s only so much time in a seven day work week.) For The Four Hundred, I wanted to take a bit of time and mull over how this affects IBM i shops both directly and indirectly.

First, this deal doesn’t surprise me, and it probably does not surprise you either. But does disappoint me and, to be honest, it saddens me. The idea that IBM would walk away from the dominant server business, at least by shipments and now by revenue considering how the RISC/Unix market has collapsed and will not likely rebound. IBM is selling off the System x business, presumably because it is not profitable, but also because it is something it can sell while at the same time getting approximately 7,500 employees off its payrolls. Lenovo’s Peter Hortensius, who is president of the Think Business Group that sells servers and storage into enterprise accounts, said that buying the IBM System x business accelerated its plan to become a dominant system supplier by about five years, and would actually boost Lenovo’s profits once the deal is done. After thinking about it for a bit, I reckoned that IBM can’t get economies of scale in manufacturing and, because it doesn’t have a PC business, it lost volume pricing leverage with Intel and Advanced Micro Devices the minute it sold off the PC business to Lenovo. This had to be apparent many years ago, and the wonder is what took so long. My guess? The advent of vanity-free, custom or homemade servers have taken about a quarter of the systems market, and there went the last remains of juicy profits for IBM, which, unlike Dell and HP, does not have a play here. Those cheap servers put margin pressure on everyone in the X86 server business.

Now, Lenovo has all of the intellectual property and engineers who worked on iDataPlex and NextScale hyperscale machines, the new System X6 modular machines that will sport Intel’s Xeon E7 processors, and surprisingly the Flex System modular and converged systems, which are sold as PureSystems. IBM is going to keep manufacturing the Power-based Flex 260+ and 460+ nodes based on Power7+ processors and presumably will partner with Lenovo to resell the Flex Systems as a single entity. Presumably IBM’s business partner and reseller channel will be able to work through IBM and sell a single integrated solution, much as they have in the past. If not, this deal will be very disruptive to the IBM i accounts that are moving towards PureSystems as well as those Windows and Linux shops who were doing the same. What I can tell you is that IBM has retained the PureSystems brand and IBM is a reseller of the portfolio of System x, BladeCenter, iDataPlex, Flex System, and NextScale products. That does not necessarily mean that IBM’s channel will stay downstream from Big Blue–they will very likely work through their existing Arrow Electronics and Avnet relationships and will now become Lenovo partners. Neither IBM nor Lenovo were not clear about this, and the wonder is why IBM didn’t just subcontract manufacturing for the Power-based Flex System nodes to Lenovo.

But why stop there? Why is IBM even bothering to keep manufacturing for Power Systems at all? High-end Power Systems machines and System z mainframes are made in Poughkeepsie, New York, for the Americas region and Singapore for Europe and Asia. Low-end and midrange Power gear is being moved to Guadalajara, Mexico, for the Americas region and is already in Shenzhen, China, for Europe and Asia. Just about all System x gear is made in China and design centers are located in Shanghai, China, and Taipei, Taiwan. The United States government, and possibly a few European governments, might have preferred for X86 and Power systems to be made locally, and in the past would have put pressure on system makers to build at least something locally–or at least do final assembly. But those days are long gone because, to put it bluntly, the server chip makers and operating system makers and disk drive makers and flash chip makers are the only ones who profit in a reasonable manner from the server business.

IBM separated its software from its hardware, and used commodity components in the System x line and commodity Windows and Linux operating systems, and rode down Moore’s Law and the volume curve for each, and there is not enough margin in it to meet Big Blue’s obsessive focus on attaining $20 in operating earnings per share by 2015. This number, which IBM’s top brass calls its 2015 Roadmap and which many employees refer to as the 2015 Roadkill, is the one and only metric that matters. Basically, former CEO and chairman Sam Palmisano staked the job of current CEO and chairman Ginni Rometty on that goal as a parting gift. (It could turn out to be a curse.)

Plenty of IBM i shops are used to buying X86 servers from Dell and Hewlett-Packard, and they are used to mixing different kinds of form factors inside of a rack to support their diverse workloads. For these companies, nothing at all has changed. Lenovo, to its credit, has retained a certain American and European flavor and will no doubt continue to have a sizable presence in the Research Triangle Park area of North Carolina–at least if it wants to continue to sell machines to the big banks, insurance companies, and government agencies that are IBM’s current System x buyers. Lenovo is not shelling out $2.3 billion just to buy some intellectual property, but an existing X86 server business with perhaps as many as 100,000 customers (that is my estimate off the top of my head) and reasonably solid relationships with those customers.

What IBM gets out of this is less clear, and we will see how it manages to fulfil the needs of customers who have Power Systems and System z mainframes and want to add in X86 gear all in one deal and increasingly in one system. IBM had planned to make System z nodes for the Flex System to accomplish this, by the way, and it could still do it. IBM gets out of making servers, which it cannot do profitably, and it becomes like Teradata, EMC, Terascala, Unisys and others that go to Dell for their underlying hardware and slap their own brands on it. In this case, IBM may not even bother with changing the brands.

IBM was clear that it will continue to invest in its Power Systems and System z products as well as its various disk and flash storage products going forward, and it will try to give the impression that it has more funds now available to concentrate here. I think this is not going to turn out to be true. IBM will invest precisely as much as it planned on such endeavors to chase analytics and other enterprise workloads and will use whatever money it was spending on System x development on share buybacks, dividends, acquisitions, and anything else that allows Rometty and her team to make that 2015 Roadmap.

The good news is that IBM did not sell off the Power Systems business and seems committed to continuing development on Power processors. It would be helpful if IBM outlined its plans for Power8+ and Power9 at this point to convince customers it is in it for the long haul and that this business is not going to be raffled off, too. If anything, making Power chips is even more expensive than buying X86 chips, given the huge foundry costs IBM must bear. Something will give here, too, at some point. Hopefully many years hence. Maybe not. We’ll have a better sense by the end of 2015 for sure. . . .

In the meantime, I will try to get some clarity on how this all might work in mixed platform shops. IBM and Lenovo may not have worked this all out yet.