Musing: Observations on IBM sale of Server Division to Lenovo

Reuters reports that Lenovo has bought the server division of IBM.

(Reuters) – Lenovo Group Ltd, the world’s largest PC maker, agreed to buy IBM Corp’s low-end server business in a long-awaited deal valued at about $2.3 billion, the biggest-ever tech acquisition by a Chinese company.

Lenovo will pay $2.07 billion in cash and the rest with stock of the Beijing-based PC maker, the company said in a statement to the Hong Kong exchange on Thursday.

Some observations from my perspective:

Good for Cisco, HP and Dell in the short-term – Companies who prefer to purchase branded servers from large US Multinationals will turn to the remaining server vendors as second suppliers. Cisco UCS is particularly well placed because a selling feature of IBM X-series was the custom silicon and features. IBM customers are also willing to go against convention and buy ‘off brand’. Customers who like those messages will like Cisco UCS products with its smaller market share but unique silicon features.

Probably Bad for Branded Servers Overall – If IBM can’t make high profits from servers with their captive customer base, then the progress of commodity manufacturing of IT infrastructure gets a significant boost. Traditional suppliers like HP, Dell and Cisco need high profit margins of >50% to meet shareholder expectations where as a manufacturing company can make a simpler product that is good enough for most requirements at <25% margins. Expect sales of branded servers to shrink in the years ahead.

Same Logic Applies to Networking – hard to ignore that the same logic applies to networking hardware. Expect sales of OEM / Whitebox switches to grow in the years ahead. It won’t happen this year but look for signs that it will happen. Note that Cisco UCS is a part of the ACI Networking in the longer term and Cisco appears to be positioning servers/networking infrastructure as a single strategy to resist white box encroachment.

A sale of UCS servers supports/reinforces sale of networking products and protects future revenue. This line of argument leads to the conclusion that Cisco ACI is a closed / proprietary system internally but an open system externally i.e. open to software on top of APIC but not a solution that is open standards between the network & server or network to network. Other parts of Cisco are pursuing an open strategy like OpenStack/ODL/XNC or even DFA and we will need to wait to see which strategy wins though customer purchases. This assumes that customers are unwise enough to commit to any strategy where so much is at stake.  Cisco isn’t staking its future on any SDN strategy and attempting to offer companies all possible solutions and let them decide.

IBM Moving Away From Full Spectrum IT Company – IBM will partner with hardware suppliers to meet their service contracts and outsourcing. But the days of IBM as a full spectrum IT provider look to have passed. Expect some companies to leave  IBM as impact hits their future business over time.

The Rise of Whitebox Servers

According to IDC, IBM had 23% market share in Q2 2013 for Worldwide Server Factory Revenue of $3.5B. Cisco has less market share than ODM vendors (admittedly combined).


3Q13 Revenue

3Q13 Market Share

3Q12 Revenue

3Q12 Market Share

3Q13/3Q12 Revenue Growth

1. HP






2. IBM






3. Dell






4. Cisco






4. Oracle






    ODM Direct












All Vendors






It should be clear that

  1. IBM second largest server manufacturer is not able to make a sufficient profit and had 20% decline in sales.
  2. That ODM or Whitebox server grew by 50%.
  3. Cisco UCS has relatively small market share but it’s growing nicely.

Comments always welcome. Your thoughts ?


IBM’s press release is here

  • Pär Björklund

    I just feel it’s worth mentioning that the timing seems good on this. With the low trust in American companies currently this could be good for Lenovo, of course trust in Chinese companies is a longtime issue but still.

  • returnofthemus

    Also worth mentioning the only server business IBM will be pullling out of is the highly commoditised, low-margin, low-growth x86 server business, while retaining its high-value, high-margin System z and Power-based systems.

    • Etherealmind

      Doubtful. The Power-based systems saw a 30% decline in sales last year. System z was down 15% too. IBM would sell off those divisions as well but no one wants them.

      Who would want a hardware platform that is shrinking in sales where it’s only value is that it runs an operating system (AIX) that no one wants. In fact most customers are replacing AIX with Linux as quickly as possible.

      The mainframes remains a valuable source of services revenue so that’s a different proposition.

      • returnofthemus

        Yes and you’ll find that Linux runs on Power too, just as it does System z, more securely and with better performance.

        • Etherealmind

          Yes, it also runs on Raspberry PI and Intel EUC. With equal security and at much better price point. The key point is that relatively few people run Linux in situations that require vertical scale where the AIX architecture is useful.

          And those companies that use AIX are finding that Cloud applications like are more attractive which further shrinks the overall market.

          AIX is approaching an extinction/survival point where revenues are neutral or shrinking slowly. In the current financial environment, this means that AIX is irrelevant.

  • Etherealmind

    What tosh. It’s part of commoditisation process. You can’t sell a product at 75% gross margin when it can be manufactured in volume and with the same functionality. Any decent sized computer manufacturer can make a product identical to IBM, UCS, HP or DELL at fraction of the

    The market is changing and IBM is reaping the fruits of moving it’s workforce offshore. Their services and products are being commoditised too.

    • returnofthemus

      IBM is more software and services than it is hardware, a global company with revenues inxs of $100m and 400k employees, exactly where is offshore?

      • Etherealmind

        My best working definition: When it’s not in legal jurisidiction of customer.

  • Mark Berly

    As usual great article – the server market is definitely moving in the direction of ODM manufacturers. You are seeing the same type movement in the networking space toward merchant silicon (COTS) it reminds me of the various processor manufactures trying to keep up with x86 many years ago. The realization that merchant silicon provides better functionality / performance at a lower price is really taking hold as we are seeing public / private cloudification of the data center infrastructure. Creating closed / proprietary systems to justify high prices is becoming a thing of the past, if companies are focused on the proper business model and engineering process to support the expected margins while providing customers high quality product at highly competitive prices its going to be a long down hill slide for them…

    • returnofthemus

      I can only assume your post was somewhat cloudified (excuse the pun), by some of the inaccuracies in the blog, one of which was that Cisco’s UCS and IBM’s System x are somehow non-industry standard, yet both are based solely on x86 (hence the ‘x, in system x), more specifically Intel x86.

      Now when I think ODM I get Déjà vu and the rise of Dell during the whole build your own PC phenomenon and companies that fell by the way side, Compaq, Toshiba, Olivetti, Apricot, Tulip to name a few.

      Yes its true we are seeing a rise in the use of merchant glass, by every single major networking vendor, but from my observations this is mainly in the DC, the big benificiary here at the moment appears to Broadcom.

      Isn’t it strange that we don’t refer to Intel x86 architecture or the Broadcom chipset as closed or proprietary?

      As a side note IBM is major proponent of Open standards not only has it benefited from it, its very survival depended on it, in 2013 it surpassed Redhat in its contribution of code to OpenStack, it spearheaded the creation of OpenDaylight, as well as the Open Virtualization Alliance and made significant investments in Linux.

      Furthermore all its ToR DC switches are based on merchant silicon and support OpenFlow.

      • Mark Berly

        The real benefit of UCS (and now ACI) from Cisco is the proprietary aspects – agreed the servers themselves are x86. The reason people do not call x86 proprietary is it is not companies such as AMD, Cyrix, NEC Corporation, IBM, IDT and Transmeta have manufactured CPUs conforming to the x86 architecture. The reason none thrived / survived in that market was they could not produce chips at the scale or price that Intel can, we are seeing the same thing with the networking ASICs. The differentiator is the ability to integrate the various silicon into a single operating system release, if vendors can not do that they will eventually be over taken by those who can….

        • returnofthemus

          Other than ACI being a complete fantasy (just like Cisco’s original SDN play, the Self-Defending Network), I’m afraid you lost me.

          Just about everyone knows that Intel won that battle, which is why the dominent architecture is Intel x86, I would’ve thought the real differentiator to be chassis design/construction and systems management, not its ability to run Windows or Linux?

          • Etherealmind

            The differentiation derived from chassis design, systems management and other fancy hardware add ons does not seem useful or valuable going forward. Who wants a blade server at premium pricing when you can buy the same outcome with an OpenCompute/Super Micro/ODM form factor.

            Same points apply to networking although the change in networking is yet to come.

          • Mark Berly

            There are dramatic differences in the hardware design, aside from power (which directly impacts OPEX) things like MTBF is often overlooked. Providing well constructed hardware at a fair price with an operating system that is stable, feature rich and allows for both traditional network management as well as DevOps and customizable features / functions is what is needed – IMO….

          • returnofthemus

            Couldn’t agree more, while the hyperscale environments have licence to experiment, I’m not sure many enterprises are ready to start experimenting with unknown ODMs with their mission critical workloads just yet.

  • Michael Leonard

    Cisco UCS is only growing since it rides on the back of the Nexus sales effort and installed base, plus Cisco put a whole offer around it for cloud orchestration and also uses VCE to drive in to cloud service providers. Technically it has a slight advantage in being able to address more memory to up the VM per server ratio. Other than that the UCS sales increase does not negate the growing impact of the white box trend which will continue. This week Open Compute is at the San Jose Convention center. It should be interesting.

  • Etherealmind

    The numbers vary, but it is significant. What is more important is that the stock market rewards growth not sales. So server sales have shrunk by about 4% on $10B business per quarter between HP, IBM, Dell and the other minor players. Gartner numbers suggests that Whitebox amounts for the sames sales as Cisco UCS.

    So the cloud companies are not buying much. And what they buy is cheap so the numbers are low. The impact of cheap means that 50-70% profit margins on servers are being eroded.