IBM Buys Blade Network to Control Ethernet Switches
October 4, 2010 Timothy Prickett Morgan
The buying spree at IBM continued again last week, as the company spun the acquisition wheel of fortune and landed on blade and rack Ethernet switch maker Blade Network Technologies for what is rumored to be $400 million in cash. The deal marks the re-entry of Big Blue to the networking business–if you don’t count the OEM agreements with Blade Network, Juniper Networks, and Brocade Communications and the reseller agreement with Cisco Systems.
I don’t count these as being in the networking business, but before last year and the Dynamic Infrastructure blitz, IBM didn’t even try to look like it was in the networking racket. (Remember Dynamic Infrastructure? That was before Smarter Planet and after On Demand and e-Business.) And now IBM has a big problem, or rather, three of them. Cisco Systems has jumped into the server racket with its Unified Computing System blade and rack servers and their converged server-storage switches (which allow for Fibre Channel storage protocols to run over the same switches that link servers to each other and the outside world). Hewlett-Packard has bought 3Com to become a credible number two player in networking behind Cisco, has a converged networking strategy for blade servers that has allowed HP to kick the tar out of IBM, and now has the largest server business in the world (in terms of both shipments and revenues) and the second largest IT services biz on the planet. Oracle owns databases and middleware, has a clear number two position in applications, and has its eyes and heart set on peddling integrated stacks of application and systems software and all-Oracle hardware–including InfiniBand and Ethernet switches.
Quite hilariously, the Blade Network acquisition comes on the heels of Big Blue’s president, chief executive officer, and chairman, Sam Palmisano, bashing HP for not doing enough of its own research and development and paying a $2.4 billion premium to acquire disk array maker 3PAR. Well, it could turn out that the 3PAR acquisition pans out a lot better than the XIV deal did for IBM a few years back, although the XIV clustered disk arrays are selling well, by IBM standards.
IBM sometimes has a knack for going in precisely the opposite direction it needs to go, as all companies do. IBM did this with application software in the early 1990s and did it again with networking in the late 1990s. And exiting the PC business may prove equally disastrous for Big Blue over the long haul, if HP can create the whole stack from handheld or PC back to the data center, control the whole “end user experience,” and sell that, soup to nuts, to IT shops. (Which is precisely the plan now that it owns Palm and 3Com.) This is HP’s vision, as I explain in detail over at my other day job at The Register.
Rather than compete in the networking business, IBM sold off its networking business–including all of the intellectual property in its switches and routers–to Cisco for an unknown sum and stopped development on new products. Right when a whole new wave of networking innovation was coming down the pike. IBM did manage to get a $2 billion, five-year technology and reseller agreement out of Cisco as part of the sell-off. IBM’s chip business got some love and money, making chips for Cisco. That was the end of the Token Ring protocol and Systems Network Architecture (SNA) for networks because it was clear that Ethernet and TCP/IP had won. Well, except for all those places in the data center and in the supercomputer centers of the world where InfiniBand still wins hands down.
The upshot of IBM’s exiting the networking hardware business in 1999 is that it did not have its own switches when it entered the blade server market in 2001. Whoops. And as it turns out, not only can you charge a premium for blade servers compared to rack servers, you can charge a premium for blade switches compared to rack switches. Whoops again. IBM tapped Cisco and Nortel Networks to supply its blade switches, but Nortel went into bankruptcy in January 2009 and Cisco is now, no matter how much the two companies smile at each other, an enemy. Luckily for IBM, three years before Nortel went bust, the company sold off its blade server switching business in early 2006 to private equity firm Garnett & Helfrich Capital, which relaunched it as Blade Network and eventually erased the Nortel brands off its products.
Blade Network gets most of its money selling switches for HP, IBM, and NEC blade servers, which are rebadged by the vendors. Blade Network has created four Gigabit Ethernet and two 10 Gigabit Ethernet modules for IBM’s BladeCenter chassis, and according to Big Blue, about half of its BladeCenters are equipped with one or more of these switches. HP has been emphasizing its own VirtualConnect Gigabit and Flex-10 10 Gigabit Ethernet switches over the three Blade Network modules it also sells; NEC has two Gigabit and one 10 Gigabit Ethernet switch from Blade Network in its product catalog.
This is the second time in two weeks that IBM has bought a company that has an OEM relationship with NEC, the prior one being the $1.7 billion acquisition of data warehousing appliance maker Netezza.
In the past two years, Blade Network has moved out to rack-based Gigabit and 10 Gigabit switches, called RackSwitches, oddly enough, that are resold by IBM and rebadged by Juniper.
Blade Network is a privately held company and thus does not release formal financials, but Vikram Mehta, the company president and chief executive officer, told me this time last year that as part of its contributing to a $10 million round of funding for Blade Network, NEC had assessed the switch maker to have a value of around $230 million. NEC, Juniper, Garnett & Helfrich, and an unnamed third party kicked in that $10 million. Everyone believes that the third party was IBM, of course, since only a few months prior to the funding round, IBM and Blade Network inked a patent cross-licensing agreement.
If the rumor reported in Barron’s is correct and IBM paid $400 million for Blade Network, then it paid a pretty high multiple to get its hands on it. Last September, Mehta told me Blade Network was at an annual revenue run rate of around $100 million and growing at around 30 percent, which means it should be on track for around $130 million in 2010. If that is the case, then IBM paid a factor of three to get control of a company that probably gets more than half of its revenues from IBM at this point. The deal will no doubt kill off the HP OEM agreement and NEC may or may not want to partner with IBM for the switches in its future generations of the SigmaBlade servers, so the real comparison to figure out the premium IBM paid is the revenue that would have been preserved versus the price of the deal. Say IBM accounted for $65 million of Blade Network’s sales in 2010. Then the multiple doubles to six to one.
Why would IBM do this? Because now all of the profits from future blade server switch sales come directly to Big Blue, and ditto for rack-based Gigabit and 10 Gigabit Ethernet switches. IBM can now tightly integrate the Blade Network products with its Systems Director systems management tools and its VControl virtualization management plug-in. The RackSwitch switches are VM-aware, in that they can preserve network settings for virtual machines as they are live migrated around a pool of servers. This is something IBM needs to add for Power Systems. And speaking of which, IBM needs to wake up and offer live migration for IBM i partitions and, why not, subsystems. If a subsystem is like a virtual private server, then prove it and offer live migrations for these, too.
It is remotely possible that NEC, Oracle, or HP makes a counter-offer to Blade Network, but this seems unlikely given the tight ties with IBM. The deal is expected to close in the fourth quarter, and soon after that, you can expect a new converged networking push from Big Blue on its various systems.