When Avnet bought Savoir Technology Group a few months ago, that left Arrow Electronics and Pioneer-Standard Electronics as the only other master resellers of AS/400 equipment in North America. This wholesale distribution channel, which accounted for about 80 percent of AS/400 deals and about 50 percent of AS/400 revenues in 1999, is now the main conduit for the thousands of AS/400 retail Business Partners to get the wares they sell to AS/400 shops. IBM itself accounts for the other half of AS/400 sales, of course, and it tends to keep tight control of the top AS/400 shops that probably bring in at least one-third of total AS/400 revenue each year. (That 50-50 split estimate comes from a high-level AS/400 BP who knows such things.) Since the wholesale AS/400 channel has probably consolidated about as far as it can—Avnet or Arrow or someone else could buy Pioneer-Standard at some point, but this is unlikely—there is another channel now emerging that could, with proper cultivation, account for a majority of AS/400 sales in the future. That channel is the collective application service provider (ASP) channel.
Strictly speaking, ASP is a service, not a channel. But as old stalwarts in the midrange enterprise resource planning (ERP) software market like Intentia International, Infinium Software, Lawson Software, and J.D. Edwards & Company—not to mention ERP powerhouses like SAP AG, PeopleSoft, and Oracle—all move toward distributing their core applications as services, rather than as software that customers install on their own servers and clients, the collective ASP channel could start accounting for a large percentage of the AS/400 aggregate processing power that gets shipped each year. This possibility—some say eventuality—has important ramifications for AS/400 customers.
First of all, the economics of volume purchasing will weigh heavily in favor of ASPs over mom-and-pop shops that buy their own AS/400 equipment and either buy new ERP software or maintain their own code on those AS/400s. I’ll do a little math on the back of the envelope to put this in perspective. Warning: Many of the numbers in this story are estimates. I am going out on a limb making estimates not because I think my guesses are about as good as yours but because the numbers are not the point. My main interest in doing this story is to show you a new way to think about the ASP trend, not to nail down exact numbers. Besides, only IBM has the exact numbers, and it will not divulge them.
Annually, the typical AS/400 shop buys somewhere between 100 and 500 Commercial Processing Workloads (CPWs) of AS/400 processing power, with a net increase between 50 and 250 CPWs. (CPW, as you may know, is the relative performance metric that IBM uses to gauge AS/400 horsepower.) In 1999, the average CPW bump in
the AS/400 base was probably about 200 CPWs, with a net increase of about 100 CPWs. With roughly 110,000 units sold last year, the number of CPWs shipped into the base totals about 22 million, with a net increase of about 11 million CPWs. (Some of these CPWs were priced at the relatively low client/server rates, and some were priced at the high 5250 interactive rates.) So, with $2.64 billion in sales last year, the average AS/400 customer paid about $240 per CPW for the net new capacity it bought.
That low dollar-per-CPW amount implies a few things. First, it indicates that most AS/400 customers were buying Model 170 servers at a fair discount rather than going with more expensive 7XX machines. Second, it implies that high-end AS/400 customers were buying bare-bones 7XXs rather than 7XXs with pricey interactive features. The list price of the 7XX models with base 5250 interactive features installed hovers around $100 per CPW. The price of machines with interactive features that allow the 5250 protocol to run increasingly unconstrained on the 7XX hardware ranges from hundreds to thousands of dollars per CPW. That price doesn’t include the cost of extra memory and disk capacity or other peripherals, which, I assume, is included in the $2.64 billion AS/400 hardware and operating system sales figure (this estimated figure is from analysts at Salomon Smith Barney, and it more or less jibes with IBM’s own vague financial disclosures). In a normal year, IBM sells about $3.3 billion dollars in AS/400 gear. I don’t think 2000 will be a normal year, so $3 billion—halfway between where 1999 was and where 2001 should
be—is probably a better reckoning. IBM has been forecasting that it will ship about 110,000 AS/400s this year. If the average CPW bump is again about 100, another net 11 million CPWs will be added into the base, and customers will spend an average of about $272 per CPW for the capacity they install. Roughly $1.5 billion of that $3 billion in AS/400 sales will go directly to IBM, and the remaining $1.5 billion (again, accounting for about 80 percent of the AS/400 deals) will be spread out across the Business Partner channel. BPs will, of course, mark up these machines, which they will probably buy for a 20 to 30 percent discount off IBM list price and sell for an average of 10 to 15 percent off list price. So, when you add in what the master resellers get (a few hundred million bucks) and what the retail distributors get (another few hundred million bucks), you come out with IBM raking in $1.5 billion in AS/400 sales and BPs in aggregate bringing in about $2 billion.
Let me stress again that all of these numbers are rough estimates I am making to illustrate my point. IBM has the real numbers, and those of you who are Business Partners should have a better shot at getting those real revenue figures than I ever will.
Now let me throw the ASP monkey wrench into the works. Say that, instead of having to go to their bosses to get AS/400 upgrades for their old V3R2 CISC or V3R7 RISC AS/400 servers, AS/400 MIS managers decide to move to modern AS/400 hardware and modern OS/400 software by hosting their own applications at an ASP’s data center. They can move their RPG and COBOL applications over, have someone else put a Web front-end on them, and pay a monthly fee to support those applications on an ASP’s iron. Or they can buy an Internet-ready suite of software from one of the big ERP vendors and throw out their software. Either way, the ASP just wants to make money by renting CPWs and networking services and by providing guaranteed application uptime. If ASPs get enough AS/400 customers to start thinking this way, ASPs could get hardware, software, and network bandwidth at such low prices that it would be less expensive for vintage AS/400 shops to go with an ASP than to run their own data centers.
I know that IBM has already signed up a bunch of ASPs for the AS/400. But what I am suggesting is a little different. The ASPs that IBM has inked deals with are themselves software providers or are supporting independent software vendor (ISV) applications as wholesale ASPs. What I am talking about is hosting AS/400 customers’ own applications under an ASP framework as well as doing the same with off-the-shelf applications. Most ASP business models are focused on hosting third-party apps. I’ll admit that ASP hosting of RPG and COBOL applications is an odd thought. But considering that AS/400 companies kept their AS/400s and RPG and COBOL apps while riding out the Y2K
transition rather than ditch them for third-party code, any successful ASP marketing model has to be built with the idea that companies do not want to get rid of their code, no matter how much IBM and ISVs want it to be otherwise. Certainly, some customers will want to move to third-party applications under an ASP model, but they would have moved to ISV applications even without the ASP offering. The ASP model will not drive software sales unless it can be positioned as a service either to improve what customers already have and then host those applications or to rehost them with the intent to improve them.
So how big could this get? It all depends on how IBM and its ASP partners play it. As Henry David Thoreau once said, “In the long run, you hit only what you aim at. Therefore, though you should fail immediately, you had better aim at something high.” Roughly 75 percent of the 550,000 AS/400s that are in operation in the world today are on OS/400 V3R2 or V3R7. They are located in shops that have avoided modernizing their equipment for over four to seven hardware generations and during the most impressive rollout of hardware and software enhancements that any midrange vendor has ever put together. Simply put, E04 or D50 or F80 customers aren’t going to move, because right now those AS/400s aren’t costing their companies anything. They typically own them outright, and they usually don’t pay maintenance. The main reason these companies bought AS/400s was that they liked the reliability and familiarity they had with the IBM midrange line, which is why they rarely upgrade. But if they can be convinced that those machines are costing them business opportunities and that the ASP model is less expensive and more flexible than buying new kit, they would probably sign up in a heartbeat. This speculation would be especially true given that the ASP model is a monthly service fee, not an acquisition, and, as such, gets the AS/400s off the capital-investment-and-depreciation side of the balance sheet and ASP service fees on the expenses side. This is the kind of thing that small and midsized private businesses are particularly keen on doing because every $100,000 that they don’t spend on capital equipment ends up in the owners’ wallets.
In aggregate, these vintage customers probably have machines that average somewhere between 20 and 30 CPWs in processing power. Many vintage AS/400 shops have big boxes, to be sure, but many more have smaller deskside servers that may have under 10 CPWs of power. Just for argument’s sake, say that 25 CPWs is the average across those 423,000 vintage AS/400s installed worldwide. That’s about—you guessed it—11 million CPWs of processing power: a whole year’s worth of AS/400 business just sitting there, not being business at all. If the average CPW horsepower in the vintage AS/400 base is a little higher, there may be as much as two years’ worth of AS/400 revenues sitting idle. If IBM and its ASP Business Partners could rig AS/400 prices so that it would be incredibly inexpensive for small and midsized businesses to go to an ASP model, it could bring a majority, maybe even a supermajority, of those vintage AS/400 customers into the modern age.
IBM could facilitate this modernization by delivering true rack-mounted AS/400 thin servers that cost $5,000 a pop or less and could do it now rather than sometime in
2001. Or IBM could extend logical partitioning on the 24-way I-Stars to include Vision Solutions’ CATSe virtual processing capability to create a machine that could have 192 virtual processors running on a single box, each one with about 80 CPWs of power by my estimate. I would guess that a reasonably configured 24-way I-Star will cost about $4 million, which works out to about $20,000 per virtual processor, or $225 per CPW per virtual processor. We want to mirror those virtual processors for high availability, so say that drives it up to $600 per CPW per customer. IBM and ASPs could carve up older 12- way Apache and Northstar processors to create mirrored virtual processors with 24 CPW and 47 CPW ratings, which would probably cost about the same as the I-Stars because IBM hasn’t really cut prices much in the AS/400 line since 1997. (Big Blue has improved AS/400 price/performance by increasing performance, but the cost per CPW has not changed all that much.) For simplicity’s sake, let’s stick with just I-Stars. The total cost to an ASP would be about $48,000 for that mirrored 80 CPW virtual I-Star slice. With appropriate ASP discounts from IBM, that could be dropped to as little as $360 per CPW,
or $29,000 per mirrored virtual processor. That $360 per CPW for the mirrored I-Star slice is essentially what it costs to install a small Model 170, but that purchased Model 170 would not be mirrored. And AS/400 shops would also still have to support it themselves.
Say that ASPs pass the cost of the dual I-Star slices right on to the AS/400 shops that rent them. Take $29,000 and divide by 36 months. You get a little more than $800 a month. The ASP customers have to pay another $500 per month or so for other support services. Networking costs would, obviously, be added on top of this cost. Say that, when everything but the ERP software costs are tallied up, it costs $2,000 per month for the whole shebang. (Back in the 1960s, a System/3, just for a bit of history, cost $3,000 a month to rent, including software and services.) If the entire vintage AS/400 customer base signed up, that would only come to $846 million per month in potential revenues for IBM, AS/400 BPs, and ASP services companies, with about $210 million per month going from ASPs to IBM to buy AS/400 hardware and about another $65 million per month for hardware and software support. That would be an extra $275 million a month, or $3.3 billion a year, for IBM’s AS/400 Division for harvesting 423,000 vintage machines. Even with an ASP model in place, AS/400 shops that did not want to outsource to an ASP would probably still buy about $2 billion a year in AS/400 hardware, so the AS/400 biz would be well over $5 billion a year. Talk about your server growth! How about double last year’s revenue?
Here’s the beauty: ASPs could run all of those vintage AS/400 workloads—get this—on fewer than 4,500 24-way I-Stars. A mix of 4,500 Apaches, Northstars, and I- Stars would probably be better, so customers would buy dual 24, 47, or 80 CPW slices. If IBM did this, the potential ASP-related revenues coming into Big Blue would probably be diminished to $2.5 billion or so. In any event, with a hardware investment between $1 billion and $2 billion to cover AS/400 servers and systems software, IBM and ASPs could create a market with a steady revenue stream of between $2.5 billion and $3.3 billion a year by replacing the vintage AS/400 base. ASPs and networking providers would get another $4 billion to $6 billion or so from that vintage installed base.
Of course, with prices that low, no one would actually buy an AS/400; only ASPs would. We would all be renting capacity. Odds are that $300 per CPW would do the trick, especially if ASPs offered 48- and 60-month contracts that spread the cost of those virtual AS/400 slices out so the monthly cost of mirrored I-Star slices were below $2,000 per month. How low IBM and ASPs would have to go on AS/400 prices under this scheme depends largely on the networking costs. If IBM could cut a deal with AT&T to offer dirt- cheap networking for AS/400 ASP customers, the total cost could approach $1,500 per month for the mirrored I-Star slices. Using Apache and Northstar servers would get the monthly cost for some configurations well below $1,000 a month. This would be considerably less expensive than acquiring and managing modern AS/400s and networks for these vintage AS/400 shops, and it would be a serious technological improvement for most of them, too. ASPs could give free AS/400-based ASP service for six months to any vintage AS/400 shop to get it started. IBM could stop spending billions of dollars a year on stock buy-backs and start spending that money on building an AS/400 infrastructure for a very large AS/400 information utility. All it would take is some courage and forward thinking.
This kind of talk probably scares the heck out of the AS/400 master resellers, but if IBM keeps Avnet, Arrow, and Pioneer in the ASP channel, they will all benefit from the business as long as ASPs are encouraged to eat up that vintage AS/400 installed base rather than convert it to someone else’s installed base. Yes, this is radical, so you had better think about what it means for you as a customer or as a Business Partner, whichever side of the bargaining table you are on. I am not saying that any of this will happen; I am just saying that the opportunity is there and that it is very large.