Ask TPM: IBM’s Dependence on Hardware for Software and Services
February 23, 2009 Timothy Prickett Morgan
The experts who put together Four Hundred Guru answer a lot of tech questions for the i community about programming and administering IBM midrange gear. But some of your questions are of a strategic or tactical nature, having to do with broader issues in the information technology space or with the i platform in particular. If you have a question you want to get my take on–or you want to use me to get a straight answer out of the people in IBM who know what is going on–then send me an email and I will see what I can do.
Here’s a very good question I got recently concerning how depending IBM’s software and services revenue streams are on Big Blue continuing to sell hardware.
I have had a great pleasure reading through some of your recent analyses on IBM’s quarterly earnings releases and am curious about one thing you’ve alluded to multiple times: How dependent would you say IBM’s services and software businesses are on its systems sales (and, I suppose, its Microelectronics business given the dependence of System z and System i/p on IBM’s own chips)?
Are a significant percentage–50 percent? more? less?–of IBM’s software revenue and services contracts for IBM hardware? Is most of that for its X86/X64 server business?
I dug through their 2007 annual report and wasn’t able to find it and am just curious what you thought.
Thanks, and please keep up the good work!
First of all, the only people who call me Mr. Prickett-Morgan are bill collectors or someone trying to sell me something I don’t need, and since you are neither, call me Timothy or TPM, like everyone else does. And because you asked a very good question, you can even call me Tim if you want. HA!
That is a tricky question, indeed, but kind of central to trying to understand Big Blue. I have contended for years that IBM’s sales didn’t change over time in its hardware, software, and services categories so much as revenue streams have been captured in different buckets that were more suited to the conventions of a given decade. Let’s take the extreme examples to illustrate the point very quickly. It’s 1970 and you just ordered a System/360 mainframe. It cost millions of dollars to acquire, and millions of dollars a year to support. Hardware comprises most of the cost, at least based on the sticker price items, and you have to depreciate that investment over time. But in conjunction with that System/360 mainframe, you get an on-site set of IBMers to help you code your applications and to keep the box humming. You agree to pay lots of money, and you get a ridiculous amount of service that is directly proportional to the risk you are taking in being an early adopter of advanced mainframe technology. You are doing this, of course, so you can fire lots of human clerks and offer new kinds of products and services to your company.
Now, fast-forward to 2000, only 30 years later. The mainframe hardware itself is maybe 20 to 25 percent of the system sale, storage is maybe 25 percent, and if you count memory and disk together as storage, they account for maybe 25 percent of the cost of the underlying hardware cost. Every service you want from IBM for installing and upgrading, performance tuning, application integration, tech support, hardware support, and anything else has a cost–and not only that, it has a cost that is not a matter of public record. And the software is only available on a monthly rental basis, and if you toss that into the mix, the software is easily a lot more expensive than the hardware over a three-, four-, or five-year term.
Did you spend any less money? Adjusting for Moore’s Law improvements on CPU speed and memory and disk capacity as well as adjusting for inflation, my guess is not. People still cost money, even if you outsource some of the costs to India and China. Granted, you got a lot more capacity for the money. Orders of magnitude. In one of my more prescient moments, when Louis Gerstner was taking over a battered IBM and hung the company’s future on services, I once quipped: Welcome to the new computer business, where customer support is not an attitude but a profit center.
So that’s the shape of the change. Obviously, what applies to mainframes applies equally well to i and AIX boxes and to a certain extent to IBM’s Windows and Linux platforms. Two and three decades ago, if IBM didn’t sell the iron, it didn’t get any of the other software or services money. A decade and a half ago, IBM had some systems integration and support businesses that were not entirely dependent on its own platforms, and that was because the Federal and some state governments insisted that services providers be platform agnostic; large customers followed suit, since they had a mix of iron and they wanted to simplify their support issues even if they were left with a spaghetti mess of code and iron. After IBM wiggled out of the consent decree in the late 1990s, Global Services could come into the fullness of its being. But back then, I think IBM was still very dependent on hardware sales to drive software and services sales, and for mainframes in particular. In the 1980s, the dependence was nearly 100 percent, and it started slipping as Global Services chased deals that might have other IBM platforms in them. The business process optimization and re-engineering consulting practice that IBM got when it bought a piece of PriceWaterhouseCoopers has allowed Big Blue have a very large–and fast-growing–consulting practice that has very little to do with iron at all, even if it does feed back into IT systems.
Take a look at IBM’s Q4 numbers. Global Technology Services–the IT part of what used to be called Global Services–accounted for $9.6 billion in sales, but Global Business Services–the PWC business that really doesn’t relate directly to IT–accounted for $4.7 billion in sales. So right there, here’s 17.4 percent of IM’s $27 billion in Q4 sales that have nothing to do with its own system sales. Now, let’s take a stab at Global Technology Services. IBM’s share of the server business is somewhere around 30 percent, give or take. Assume that IBM’s services customers have a slight preference for IBM iron, or that IBM server buyers have a greater tendency to employ IBM’s services, even on non-IBM iron. Let’s split it down the middle and say that 50 percent of sales at Global Technology Services is driven by companies where IBM platforms–a lot of mainframes, but also lot of AIX and Windows iron with a smattering of IBM i and Linux platforms–are the dominant machinery. That still leaves $4.8 billion driven by other platforms, and I am assuming the same spread on strategic outsourcing sales, which were 34 percent of the $14.3 billion in Global Services sales in Q4. Just because you no longer run on your own mainframes and Unix boxes does not mean you are not just as trapped by your applications, right?
Now, let’s pick apart Software Group, which had $6.4 billion in sales. Operating systems accounted for 10 percent of sales, or $640 million, but it is my guess that Windows and Linux (and mostly Windows) resales were half of that. IBM has been peddling its DB2, Informix, Lotus, Rational, and Tivoli tools on other platforms for years. I think IBM’s software is a lot more dependent on IBM hardware than is its services, so let’s say that 65 percent of the remaining software is related to IBM’s own hardware. That’s $3.7 billion that we will say is dependent on IBM’s iron. But that also leaves $2.66 billion that is not dependent.
Within Systems and Technology Group, which had $5.4 billion in sales, about $4.7 billion was for IBM’s servers and storage, and I think IBM’s storage sales are mostly for IBM systems, even if they do support other systems. But, if you want to be conservative, let’s chop it in half and say IBM has a really big business allocating storage for other systems. And I don’t believe you can think of the Microelectronics Division, which had around $488 million in sales, as anything other than a way to keep the Power architecture evolving and IBM’s chip plants in East Fishkill operating at an acceptable cost level as they make z6 and Power6 processors. We’ll see how strong IBM’s commitment to chip making is the minute Sony, Toshiba, Microsoft, and Nintendo all go for X64 processors. IBM will sell its chip plant to the same Abu Dhabi investors that Advanced Micro Devices is trying to sell its chip foundry operations to so fast it will make Tom Watson’s head spin in his grave.
Here’s the point, just among these three groups, you have $3.64 billion in server sales that are laying the groundwork for current and future software and services sales. The ratio was $3.64 billion in server sales to $9.5 billion in software, services, and storage sales directly tied to IBM platforms. But here’s the interesting bit. Across just those three groups–Systems and Technology, Global Services, and Software–there is another $12.3 billion in sales I would estimate are not tightly coupled to IBM’s platforms. So you get a 1:3 ratio of sales that are dependent on the underlying platforms and a 1:4 ration for those that are not. If you wanted to analyze this like IBM does, you would look at profits, not revenues, and I think the mainframe and i platforms and their software and services represent an even larger proportion of profits and services on non-IBM platforms represent a smaller piece of profits.
Now, let’s talk about the words dependence and correlation for just a second. IBM has about 500,000 global customers worldwide, and a little less than half of them are i shops that spend a lot on their systems but do most things themselves. About 300,000 shops use mainframes, Unix and Linux, or Windows boxes, and they tend to engage IBM across a broader spectrum of products. (IBM used to sell application software to its midrange shops, but started exiting that business in the early 1990s so it could be an agnostic supplier of platforms. I think this was foolish, of course. Where is the money today? In application software.) My point is, just by having such a large customer base, IBM could take the BPO expertise at PWC and turn it loose on a vast base of customers. The mainframe or Unix or i box didn’t drive those BPO consulting revenues, so there is no dependence. But there certainly is a correlation. Just turn it around and you’ll see what I mean. You think a customer that is ripping out mainframes is going to let IBM reorganize and rejigger its entire business? Seems unlikely. That sounds like a job that Hewlett-Packard and its new EDS unit gets.
This is all, of course, just a lot of guessing. But I think it is safe to say that IBM is less dependent on hardware sales to drive software and services sales than it was a decade or two ago, but is still very much dependent on staying in the systems racket at this point in its very long corporate life. That said, I can see the day in the distant future when the company is just called International Business and we leave it at that, forget the Machines.