IBM’s Plan for an Adjacent, Custom Systems Market
March 12, 2007 Timothy Prickett Morgan
In seven years, server maker IBM has come a long way. As the millennium turned, Sun Microsystems was growing so fast that it looked like it would overtake the top spot in the server racket in a year or two, and then Hewlett-Packard bought Compaq, and looked like it would unseat Big Blue as the world’s largest revenue generator in the market for general purpose servers. Back in 2000, IBM’s revenue share had fallen to 22 percent, and now it is pushing up toward 40 percent.
To be fair, IBM dodged two bullets, and that was because Sun and HP had each shot themselves in their respective feet with them just as IBM had locked and loaded its own guns with Power4 servers.
Sun’s business model was based on the dot-com boom going on forever and on the timely delivery of ever-faster UltraSparc-based systems. The dot-com boom ended around March 2000, which is also when it became clear that Sun was going to be woefully late with its UltraSparc-III platforms. Only now, after seven years of struggling, is Sun able to field UltraSparc-IV+ machines that can meet or beat IBM’s Power5+ servers in terms of price and performance. Having a relatively captive Solaris installed base, which is zealous about Sun’s Unix variant, Sun kept Sparc server prices high to maximize revenue during the downturn that followed the dot-com blowout, and resisted moving to cheaper X86 iron and even went so far as to kill off Solaris 9 for X86 platforms. And, predictably, Sun’s fortunes fell as customers in that downturn went to AIX–IBM was selling machines at 50 percent off list in Sun accounts as a starting negotiating position, and was doing so with its Power4-based “Regatta” servers, which absolutely smoked Sun gear on all the benchmark tests. IBM delivered three times the bang for the buck compared to Sun, and it reaped the benefits.
By buying Compaq in 2001, HP stepped over a crashing Sun and a declining IBM to briefly have the dominant revenue market share. But HP’s plan to kill off its AlphaServer and PA-9000 servers and move HP-UX and OpenVMS to Itanium-based Integrity servers hit some snags as chip partner with Intel had delay after delay with the Itanium chips. HP didn’t have new, inexpensive Unix and OpenVMS systems to sell to customers during the dot-com bust and the recession that followed the 9/11 terrorist attacks and customers did not want to over-invest in the sunsetted AlphaServer and PA-9000 systems. While a lot of HP’s customers started buying ProLiant servers running Windows and Linux, these machines are much less costly and much less profitable. It has been six years of disappointment for HP as applications were slowly ported to the Itanium architecture. Only now, with the delivery of dual-core Itanium 9000 chips, HP-UX 11i v3 and OpenVMS 8.3, and a new set of chipsets for the Integrity servers, can HP say that it delivers performance and price/performance comparable to IBM’s Power5+ machines.
And now, of course, Power6 is right around the corner, and IBM is set to open up a substantial lead again in the Unix market. Given this, and the fact that mainframe sales are up substantially from a few years ago and IBM is doing pretty good selling X64 servers, you might think that the company would pump some marketing dollars into the System i5 and i5/OS platform, maybe cut prices and boost sales and stop the move of vintage OS/400 shops to Windows. If you were thinking about IBM needing to boost server revenue market share to 40 percent as a goal of its own, you might do this.
But this is probably not how IBM thinks. What IBM knows is this. By attaining its roughly third of the Unix revenue pie as Sun and HP are getting their acts together, every incremental dollar spent to increase Unix market share brings back less in profits than dollars in past years. Gaining market share at the expense of profits is not something IBM is interested in. That’s why mainframe prices are still very high compared to other alternatives, and that is why IBM is still charging a premium for i5/OS platforms compared to Windows and Linux servers in all but a few cases (the user-capped i5 520 announced late last year being a notable exception, where it costs the same as a similarly configured Windows box).
This is also why IBM cannot honestly believe that it can take a lot more revenue market share in the X64 server space, either. Big Blue has more than half of the blade server market with its BladeCenter machines, but HP has hit back with its new c-Class BladeSystems machines. IBM and HP have the lion’s share of the blade market, which in my estimation has been stunted by the lack of standard blade server and blade chassis form factors. With HP, Dell, and IBM being the dominant X64 server players, and Sun and Fujitsu-Siemens not that far behind, and a slew of whitebox vendors who fight for every dollar, IBM can’t really expand substantially except through acquisitions. IBM could and did expand as it delivered better-engineered machines, especially at the high-end of the market, but IBM’s own sales force and reseller channel can only push so much iron–just as the channels from its competitors can. It seems unlikely that IBM could grow much beyond its 20 percent revenue market share in the X64 market.
So, with the server market pretty much capped for IBM and revenue market share kissing 40 percent, what will IBM do? Create an adjacent, custom systems market that other vendors can’t replicate so easily. This, according to a presentation I saw recently by Bill Zeitler, who is senior vice president and general manager of IBM’s Systems and Technology Group, is precisely the plan.
Given IBM’s positive experience with its Power-based game console partnerships with Sony, Nintendo, and Microsoft, the company desires to build a custom systems that are distinct from its general-purpose servers and their software and related services, which are growing about 6 percent organically, according to Zeitler. In 2005, hardware and financing made up about 27 percent of IBM’s sales, with services making up 53 percent coming from services and another 20 percent from software. (When Zeitler gave this presentation at a UBS server conference I participated in, IBM had not yet reported its fiscal 2006 financials.) IBM’s pre-tax profit picture is a little more balanced, with hardware and financing making up 28 percent of profits in 2005, compared to 35 percent for services and 37 percent for software. “It is our goal to have this profit mix, more or less,” explained Zeitler. “And we expect to gain a point of market share a year or so going forward.” But there is only so much share to be gained in this market, as a chart Zeitler presented revealed. Take a look:
As you can see from this chart, IBM’s expectations for growth in the server and storage business are not all that great. The total addressable market opportunity for servers was around $57 billion in 2005, according to IBM’s own internal estimates, and will only grow to around $60 billion by 2009. That’s 5 percent revenue growth total in five years, and IBM is committing to grow its market share by 1 percent a year, which means it has to have sales growth that is twice the miniscule level that the company is expecting for the server market between 2005 and 2009. In this market, that is quite an achievement. Storage is doing a bit better according to IBM’s estimates, growing from a $36 billion market opportunity in 2005 to a $43 billion pie in 2009–that’s 19 percent growth over five years, and about four times the growth rate for servers. This growth will come even as disk arrays use much cheaper capacity each year, which suggests that IBM is expecting, as are many players in the storage market, the continuing explosion in storage capacity among the enterprises of the world. Still, the server infrastructure market opportunity–servers plus disk arrays–is only going to grow by 11 percent over five years, rising from $93 billion in 2005 to $103 billion in 2009.
Now look at the two stacked bars that show what Zeitler called Technology Collaboration. In IBM’s estimation, this market will grow from $76 billion in 2005 to $115 billion in 2009–that’s 51 percent growth over the same five years, and five times the growth in the server/storage infrastructure space. IBM is rubbing its hands together, and considering its deep expertise in chip manufacturing and hardware and software design as it relates to the commercial, general purpose systems it makes, Big Blue smells money.
“We want to take the expertise we have in the systems business and move it on over into this adjacent market,” said Zeitler.
So what exactly is this technology coloration market? Well, to start, for IBM it means the Power chips that are at the heart of game consoles. But it is more than that. These custom systems are next-generation network devices (switches and routers), retail machinery, aerospace and defense systems, and HPC clusters with electronics and architectures tuned to solve very specific life sciences problems, just to name a few. These are not the kinds of machines you push out through the System x, i, p, and z sales channels. Zeitler says that IBM has a $4 billion or so business in this area today, and hopes to make it a $10 billion business by the end of the decade. That would go a long way toward filling in the revenue gaps IBM has since it sold off its disk drive and PC businesses, and presumably this will be more profitable, too. (It would be hard for this adjacent systems market to be less profitable than IBM was in PCs and disks were for Big Blue, to be honest.)
According to IBM’s analysis, the potential custom systems business is a lot larger than the piece it thinks it can get right now, and if IBM is saying publicly it can reach $10 billion in sales, you can bet that inside the company, Big Blue’s top brass are hoping to do a lot better than that. In any event, the dominant pieces of this $76 billion adjacent systems market that Zeitler has his eye on are custom systems, where IBM makes a whole unit (such as a radar system or a machine for doing DNA analysis on the fly), and components, such as the chips that go into game consoles, set top boxes, routers, and cars. IBM expects the custom systems business to grow by 58 percent, rising to $49 billion in 2009 from $31 billion in 2005. The component part of this business will also grow nicely, with sales of $30 billion in 2005 and growing by 67 percent in 2009 to $45 billion. While software and design services for custom systems may be small revenue contributors, these are also growing fast and it is my guess that they will be very profitable–especially when it comes to design services, where IBM has lots of skills and few competitors. IBM reckons that software sales on custom systems came to $4 billion in 2005, and will grow by 75 percent to hit $7 billion in 2009. That’s an average of 15 percent growth per year, and that is almost three times as much growth as IBM is getting out of its commercial software business today. Design services for custom systems brought in about $11 billion across the market in 2005, according to IBM’s estimates, and is expected to grow by 27 percent to hit $14 billion in 2009.
So, if you are wondering why IBM doesn’t do more to help promote one or all of its general-purpose server lines, now you have an idea of what IBM is really thinking about as a means of growing its overall business–and why.