The Business Case for the System iWant
December 11, 2006 Timothy Prickett Morgan
The Excel spreadsheet associated with this article is available for download
Last week, I fleshed out a hypothetical user-capped System i5 line that is based on the machines that IBM announced in October and that, as far as I can tell, represent the most cost-competitive alternative to Unix, Windows, and Linux servers that Big Blue has fielded running OS/400 and i5/OS in the past decade. I would be the first to admit that the so-called System iWant user-capped machines have some issues, but still, I wanted to build a business plan and see how these machines might help or hurt the revenue and profit stream of the System i5.
I don’t have an MBA, and I don’t have to try to sell the idea to IBM chairman and chief executive officer, Sam Palmisano, so I built the product line first and worried about the business plan last. I know a thing or two about markets and spreadsheets, and I could have done it the MBA Way, which would have meant looking at the installed base, what customers are willing to pay and which ones are willing to pay, it and then done a min-max calculation to minimize efforts and costs and maximize profits. Having done that, I could have then designed a System iWant product line that would fit the bill.
I could have done that, but I think MBA-think is completely backward, and is one of the reasons why the world is so screwed up. Which is why I did it the other way around. And, incidentally, I had no idea until I finished the spreadsheet on Saturday afternoon if the business plan would actually pan out or not. Which is living on the edge a little, since I could look like a fool if the pricing wasn’t right on the machines. But, I trust my gut, the larger IT market, and both tell me that if you give customers a fair deal, they will give you money and keep coming back.
Luckily for me–and perhaps for the System i5 business if IBM does something akin to what I am proposing–the numbers work. (Whew!)
Just to recap a bit. I don’t think IBM can wait until Power6 and V5R5 to fix the economics and packaging of the System i5 line. With the System iWant 2007 Edition, I created a line of machines that can support from 13 to 421 users and cost from $8,600 to $347,300. Those prices include hardware plus base disk and memory, as well as full licenses to i5/OS V5R4 and DB2/400, including WebSphere for Java workloads and full-on 5250 processing for green-screen workloads. No governors, except on the number of concurrent users on the system. The boxes cost $3.60 to $4.60 per CPW, if you want to think in terms of CPW, including a very reasonable amount of memory and disk: 1 GB of memory and one 35 GB, 15K RPM disk for every 10 users on the system. The System iWant to Buy is a perpetual license, but you will have to pay maintenance on it (call it 10 percent per year maybe). The System iWant to Lease has a 36-month, $1 payout lease based on the list prices at a 5 percent interest rate. No negotiating on the price. List price is street price. Memory costs $500 per GB, and the disks costs $1,000 for the 35 GB unit. This is plenty of a premium to street price, and perhaps a bit too high, but it is better than what IBM charges. I used both Power5 and Power5+ processors to create a line with 66 different configurations, with a fair amount of granularity on the user count.
So, what would happen if IBM took my advice and actually launched the System iWant in 2007 instead of sitting on its hands until 2008, as the company is rumored to be thinking about doing? The short answer, if you believe that IBM can get the entire installed base to move to new gear in five years, is that it can make significantly more money than it currently does now on the basic hardware and software, and have a much more engaged installed base. And then it can make even more money on selling extra memory, disks, other peripherals, maintenance, software tools, and services to customers who either buy or lease a new System iWant machine.
To build the System iWant business plan, I started with the two simplest statistics: the number of OS/400 and i5/OS customers in the world and how many machines are in the installed base. IBM and the companies that count customers and boxes are a little vague on the number, so I picked three different scenarios based on installed bases of 450,000; 500,000; and 550,000 machines worldwide. As a data point to check reality against, I know that in the early 2000s, there were 22 million end users sitting at AS/400 and iSeries servers in the world, because IBM said so in talking up its various DB2 databases. The scenario I posit leaves us with between 16 million and 20 million end users seats, and given the attrition in the i5/OS and OS/400 community in the past five years, somewhere around 15 million to 17 million seats is probably where the real world is currently at. The model I laid out stipulates that everyone moves from their current machine to a System iWant within five years. All of those machines get unplugged and all those seats get moved over.
Knowing the size of the installed base, I then tried to figure out what System iWant machines the base would move to. In the real System i5 world in 2005, the i5 520 entry servers accounted for 96 percent of sales. I don’t think the System iWant would be any different. So my model says that 95 percent of the AS/400, iSeries, and System i5 machines replaced customers will be 520-class boxes. I then have 4 percent of the base moving to System iWant 550 boxes, and 1 percent moving to larger System iWant 570 machines. In the Excel spreadsheet that I built to come up with this model, which I am sharing with all of you so you can pick it apart and maybe improve it, you can change the percent of sales going to the iWant 520 and iWant 550 machines and it will cascade all the changes throughout the spreadsheet. You can do what-if analysis.
Knowing the machine breakdown, I now factored in how many machines would be bought and how many would be leased. I reckoned that 35 percent would acquire and 65 percent would lease, mainly because of the attractiveness of the price, the stinginess of the OS/400 customer base, and the low finance rate I am making IBM charge. But again, if you change the percent that want to buy (which are in bold blue text in the spreadsheet), the model will cascade the changes and you can see the effect.
Having done those two steps, you have to reckon when people will upgrade the machines in the installed base. I think it starts out slowly, then accelerates in year two, peaks in year three, and trails off in years four and five. This is the cycle we saw in the transition from CISC to RISC AS/400s from late 1995 through late 1998. Then, you need to figure out what the price of the machines they will buy is. I just took a mathematical average of the three different types of machines in last week’s story and plugged them into the sheet, along with the average number of users each class of machines would support. I realize that this is speaking in averages, but this will give us a rough feel for how the money will add up for base boxes and their integrated system software. (My base box has enough memory and disks to be useful, just like IBM’s user-capped Solution Edition machines and unlike most of the other models in the real System i5 lineup.)
So, we have model, price, type of acquisition, number of machines acquired, and time of acquisition over a five-year span for three different sizes of installed base. From here, it is just some simple math build it out into a five-year sales matrix, and then do the math to see how it all adds up.
Click here to see a static image from the spreadsheet with my initial conditions set.
With the model I have set up, IBM only makes $715 million peddling 67,500 machines in the first year if the base has only 450,000 machines. If the base has 550,000 machines, it is boosted to $874 million. That is less than IBM makes selling iSeries and System i5 hardware today, but probably not too much less. And, IBM would be selling regular System i5 machines for people who want unlimted user System i5 machines and their perpetual software licenses in the first year, remember, which I ignored in my analysis.
In year two, sales bump up to just under $1.5 billion because IBM has converted a cumulative total of 180,000 machines, with two-third of them on leases. By year three, annual sales have grown to $2.5 billion and there are now 337,500 machines moved over. In year four, revenue holds at $2.5 billion, and the base grows to 405,000 machines. In year five, sales taper off a bit to $3.2 billion and 450,000 machines have all been converted to the System iWant user-capped boxes. And here’s the funny bit: this is very aggressively priced iron, which can compete and beat Windows. Moreover, these numbers do not include a hardware upgrade cycle for the customers who buy or lease machines (which could easily add $300 million to $400 million each year as customers add users). This model does not take into account maintenance fees, upgrades for peripherals, memory expansion and disk capacity expansion–all of which would add maybe another $200 million to $300 million to the kitty once the base is converted. And services–imagine if IBM offered migration services, or if customers could easily justify high availability cluster?
In short, IBM might be able to pull off a $4 billion System i5 business and have 16 million happy end users and 215,000 happy customers who do not have to cost-justify every thing they do on the box. And it could get considerably larger, too, with cheap HA clustering that is used by much more than a tiny fraction of the base, as it is today.
My model above assumed people will rather lease than buy, and that makes more money for IBM, of course–as well as its resellers. But if the split turns out to be 50-50 on buy versus lease, the revenue only drops by several hundred million dollars a year. In year five, the System iWant business still ends up being a $2.7 billion business on the conversion of the 450,000-strong installed base–not including all that other extra revenue.
This is considerably better than IBM does today. And it is the best reason I can give for IBM to get in gear and get a revamped product line out in 2007. It is time to innovate in a way that makes the System i5 distinct, easy to use, and easy to sell again. If IBM can do it, now is your time and chance to tell IBM to get up off its, er, AS/400 and get it done. And if IBM can’t do it, then sell the business and let someone else who can do it, do it.
The System iWant, 2007 Edition
Don’t Wait Until 2008, Kick It Up to 11 in 2007
Bang for the Buck: User-Capped i5 520s Versus Windows X64 Servers