With the sale of its x86 server division for $2.3B, IBM exits the marketplace it started in the early 1980s. Some have argued that this is a good move for IBM and tolls the death knell for the x86 server marketplace. “Well, if IBM is closing shop, surely the end is nigh. What with all that virtualization malarkey, nobody is purchasing x86-based servers anymore.” While it is true that the glory days may be behind for a company based on server hardware due to the consolidation of large numbers of compute entities on single-host servers, there is still a significant marketplace for x86 servers: according to IT Candor, over $46B in 2013. Although the server market suffered its third quarter of revenue decline, dropping 4%, actual numbers shipped increased by 2%. The plain fact is that people are still buying x86-based servers.
So, why has IBM exited this market? Servers like its x3950 and x3650 models make excellent ESXi servers, have a large memory capacity capable of running four CPUs, and offer excellent expandability options. Strategically, as a company it is moving to a more services-and-software model, and it is still low overhead and large margin. That said, IBM is still the dominant player in the non-x86 server market, with well over 50% of the market share, or an estimated $13.7B.
If you do the math on the graphs above, you’ll see IBM is walking away from over $2.3B in revenue, but this is not high margin, and IBM does not and cannot compete on price. It makes very good economic sense for it to walk away and concentrate on its larger and much more lucrative hardware market in zSeries, pSeries, and iSeries machines, where it still controls the whole stack, from hardware to end user presentation.
It is easier to understand why Lenovo has purchased the division. IBM will still sell the xSeries brand, so its market is fairly assured, with one possible caveat to be discussed later. Lenovo gets an immediate $2.3B slice of the server pie. Lenovo purchased IBM’s PC division in 2005 and has taken that failing brand to bigger and better places, so it has the obvious experience of turning around the brand.
What is interesting in the graph above is the size of the “other” section: over 36%. This section refers to the market for white box x86s—unmarked custom-built servers the likes of which Google, Amazon, and Facebook are putting in their data centers by the bucketful daily. This is a very large proportion of the market now, and it is a major thorn in the sides of the two major incumbents, HP and Dell, as there is no way that either can compete on price alone.
How will this affect the other vendors?
My original thought was that this would change the status quo dramatically. On reflection, though, I believe that the world will continue as before. I originally thought that IBM’s sale of its server division to Lenovo would debar it from the US federal market, thereby opening greater opportunities for HP, Dell, or Cisco. However, with IBM continuing to sell the brand, I now believe that these opportunities will not develop. Lenovo has had success with the PC market that eluded IBM when it owned the PC division; if Lenovo can replicate this success with the server division, then it may even open up greater opportunities for IBM Federal, as the latter’s hardware offering will be much improved.
What is the future of this space?
In the enterprise, HP, Dell, and to a lesser extent Cisco and Fujitsu should be worried about the growth of the white box. With the virtualization of the data center, the choice of hardware has, to a large extent, become irrelevant. It really is all about the software and the management of large arrays of servers running multiple virtual machines. This is where the likes of Dell with OpenManage, HP with SIM, and Cisco with its UCS platform can shine with their integrated hardware management software. White boxes do not have that seamless and integrated management and monitoring platform that comes with the big vendors; white box users have to go out and purchase products that are potentially less functional, or they must spend on customization of open source products to gain the same functionality.
That said, in today’s cloud-based service world, hardware is increasingly becoming a commodity. With resilience built in further up the stack, the loss of a host server is not as catastrophic as it previously was, given that the ability to restart VMs automatically is built in on all hypervisors. When a server goes down, all that a cloud provider really requires to provision and install a replacement is a couple of SNMP traps indicating that the server is down and its location.
The fact is, x86 hardware is and was commoditized with the introduction of virtualization. Has IBM done the right thing? Time will tell. I do believe that it was the correct decision for IBM as a company. It could not afford to hold on to that sector. Due to the way that IBM as a company is structured, it could not be as agile as HP and Dell. As a result, it has suffered greater from the effect that Cisco’s ingress and the white box expansion have had on the overall market.