Exing-out X: Chinese Company Poised to Become World’s Leading Volume Server Maker

On January 23, 2014, IBM announced its intention to sell its System x product line (System x is IBM’s brand name for its x86-based servers) to Lenovo for $2.3 billion dollars. This acquisition, pending U.S. government approval, will position China’s Lenovo as the number three x86-server maker in the world. And I believe that within five years, this acquisition will enable China to become the number one server maker on the planet from a volume-of-sales perspective.

I’ve been tracking the progress of IBM’s System x (formerly known as Netfinity and then x series) since the mid-1990s. When IBM first introduced System x, I initially found the Netfinity product line to be bland and lack luster – a me-too entry in a crowded market of server players that included Hewlett-Packard, Digital Equipment Corporation, Dell and numerous many, many others. But IBM engineering rapidly changed my opinion by “cascading” functionality from its more advanced System z mainframes and Power Systems onto the System x platform – and by improving the System x design and regularly adding advanced management hardware and software features. System x rapidly moved from relative mediocrity to today’s industry flagship for advanced scale-up x86 designs (referred to as “high value systems”). And the company has also demonstrated leadership in other x86 market segments including high-performance computing, blades, and more recently in the growing converged “expert systems” market.

But even with excellent x86 systems designs, IBM has had to ask itself one very tough question over the past few years: “should we be in the high volume/low margin volume x86 server marketplace?” IBM makes its money by selling high-value systems, software and support – and IBM eschews the high volume/low margin product model (that is typified by Hewlett-Packard and Dell). So why continue to invest in a product line where lower margin profits are derived by the sale of huge volumes of low cost servers?
IBM’s proposed deal with Lenovo essentially hands-off IBM’s high-volume low-end and midrange, high-value (scale-up) and converged “expert systems” System x servers to a company that embraces the high-volume/lower margin model (Lenovo the world’s leading PC maker) – while enabling IBM to continue to architect specialized appliances as well as future hybrid system designs. By making this move, IBM can rely on Lenovo as a partner for building mass market, commodity and high-value x86 servers – while IBM developers continue to architect specialized servers that Lenovo will presumably build for IBM.

Comments on the New Arrangement
Over the years I’ve watched IBM architect numerous x86 systems designs that have been optimized to process specific workloads. For instance, in September, 2013, IBM introduced a new x86 system design known as NeXtScale designed to deliver maximum performance (by reducing management overhead and eliminating the internal backplane). Most recently IBM has introduced its sixth generation of X Architecture, an architecture that helps scale the size of x86-based server configurations, extends memory, to improve input/output per second (IOPS) performance, improves virtual server performance, and improves system resiliency. This new generation makes it possible for System x to address up to 12 TB of flash memory – enabling System x servers to finally compete with Unix and Linux servers in the high-end of the server marketplace. And a year and a half ago IBM introduced its innovative “PureSystems” designs that focused on providing turnkey systems that are tuned by experts to offer fast deployment and optimized performance – thereby reducing deployment and tuning costs and helping customer reduce the time it takes to deploy solutions.

So the big question in my mind when I was briefed about the proposed sale of System x to Lenovo was “how is IBM going to transition its engineering talent to Lenovo in order to ensure that IBM customers can continue to gain access to advanced and innovative x86 system designs?”

The answer to this question is partially contained in IBM’s transition plan which calls for 7500 people from development and marketing from its System x organization to join Lenovo. In other words, a lot of the talent that Lenovo will need to keep building advanced systems designs will transition to Lenovo as part of the overall acquisition activities. But – and this is important – IBM will continue to design specialized servers (also known as “appliances” such as its PureData Systems for Analytics, for Operational Analytics and so on) to address the needs of specific workloads (in this case, to meet advanced business analytics needs), or to address hybrid computing needs (with specialized chassis designs such as its zBX offerings) – and even to create new designs that use multiple different types of processors such as FPGAs (field programmable gate arrays) or graphical processing units (GPUs – as found in IBM’s new The Now Factory servers). So, in short, IBM will hand off traditional x86 server development to Lenovo – but will also continue to design advanced specialized server offerings that will contain x86 processors that will be targeted to optimally address specific workloads – and these specialized designs will be sourced from Lenovo or elsewhere.

As I considered this scenario, I came to realize that IBM is providing Lenovo with a “push start” — enough engineering talent to enable Lenovo to compete more effectively in the volume low-end and midrange x86 server markets (where, as a result of the acquisition, Lenovo would suddenly become the number three x86 server supplier) — while at the same time providing Lenovo with enough talent to continue to push high-end, high-value x86 server development (where Lenovo can readily differentiate its offerings from Hewlett-Packard and Dell offerings). But this is where it gets really interesting: with this push-start of talent coupled with China’s deep developmental expertise, growing system design expertise and manufacturing prowess – this acquisition will make it possible for China to become a world-class server manufacturer. This will be good for the computing industry as a whole because Lenovo should be able to create new, innovative x86-based designs while also helping to lower computing costs (due to low cost manufacturing advantages). And, as a result of this acquisition, I expect Lenovo to replicate what it has done as a result of its PC acquisition of IBM’s Thinkpad nine years ago – move from relative PC maker obscurity to the number one position in the PC market. Only this time, I expect it will only take five years for Lenovo to become the market leader in x86 high-volume and high-value server markets (as opposed to the nine years it took to claim ownership of the number 1 spot in PCs).

In addition to continuing to invest in x86 appliances and other advanced x86 systems designs, IBM is making huge investments in its Power Systems line while also continuing strong investments in its System z mainframes. This year I expect IBM to announce its POWER8 architecture – a chip design that will feature vastly superior thread handling and the ability to address extremely large memory configurations (for supporting very large in-memory databases as well as high-end enterprise applications). Further, IBM will make a major push to expand Linux adoption on Power Systems – starting with the availability of its Integrated Facility for Linux (IFL – specialized packaging that allows Linux to run in an optimized fashion of POWER processors) – and continuing with mainframe advancements in scale, security, and performance. The work that IBM is doing with both of these server lines will provide more powerful alternatives to x86 architecture – enabling IBM to differentiate its systems and software in the areas of cloud computing, business analytics, security, scale, and more. And, when IBM needs to sell x86 solutions, it will clearly partner with Lenovo first and foremost due to its tight engineering and business relationship. But, interestingly, for customers who want to deploy Hewlett-Packard or Dell x86 servers, IBM will no longer be conflicted by having its own competing x86 product line – so IBM will be able to build and support integrated heterogeneous cloud environments consisting of multiple vendors’ platforms and storage.

Summary Observations
A few months back I wrote a trip report that talked about a speech that Steve Mills, IBM’s Senior Vice President and Group Executive, Systems and Software, presented in Stamford, Connecticut. Mills talked about the need for companies to constantly reinvent and transform themselves in order to remain competitive and retain their leadership positions. Since that speech I have come to realize that IBM is becoming a new company – one that is focusing on building cloud, mobile, social and Big Data solutions. The IBM of the past has been a big iron company, a major software maker, and a professional service provider. And these business units sometimes acted in a siloed fashion. But the new IBM is now focused on building integrated solutions (described in the previous sentence) that include optimized hardware, infrastructure and application software solutions, and expansive professional services.

As I look at Hewlett-Packard, Oracle and Dell, I see companies that are primarily focused on selling x86-only solutions. As I look at IBM, I see systems and solutions company makes optimized solutions available on general purpose computing environments (just like its primary competitors) – but I also see extremely innovative solutions built on distinct hardware platforms (such as mainframes and Power Systems) that offer significantly more processing power than is available in the x86 world. IBM’s growing suite of Watson Foundations is a good example of this phenomenon.

IBM will still be able to play the x86 game with Hewlett-Packard, Dell and Oracle – but IBM is also positioned to offer much higher-value systems with tightly-integrated infrastructure, a highly tuned database (with BLU acceleration), and a series of industry-specific solutions that its x86-only competitors won’t be able to match in terms of integration, performance and throughput. I think this IBM/Lenovo deal will be a good move for both companies, and from a competitive perspective I am very much looking forward to seeing how IBM’s new appliance-focused x86 approach works out.