Subscription Pricing: A Tough Path to a Better Pricing Model

by Timothy Prickett Morgan

It has taken a little time--like a few decades--but the IT industry is finally coming around to one of the oldest ideas in the industry: that it is too hard to sell a big, expensive piece of technology, but it is relatively easy to get a company to subscribe to that same product and spread those acquisition costs over the long haul. After all, we are all used to paying monthly for most things: our cars, our homes, our bills. We're conditioned to pay this way by life itself.

The idea of subscribing to a piece of technology is not new. In fact, the idea of acquiring technology instead of renting it is the new idea. People didn't buy cash registers from National Cash Register in the late 1800s and early 1900s; they rented them. The punch-card equipment, perfected by IBM and Remington Rand, was rented, and it wasn't until IBM's second antitrust lawsuit was settled, in 1956, that the company (and soon other vendors) offered its new-fangled computer machines for sale, as well as for rent. And at IBM you paid a hefty premium to buy. Well, after 40 years of buying computer technology, IT vendors are sick of selling and customers are tired of hearing the sales pitch and of shelling out lots of cash for IT projects that, half the time, do not finish, and may not even be worth the money and effort when they do.

As it is being envisioned by companies like Sun Microsystems, subscription pricing is less sophisticated than the so-called capacity on demand and utility pricing (mostly for servers and storage) and application service provider (mostly for software) pricing models that IT vendors invented a few years ago as a rough draft for the kind of subscription pricing that will probably prevail in the future. Sun has two excellent examples of subscription pricing: one you can order today, and the other is still very much in the concept phase. First, there's the Java Enterprise System middleware stack, which can be licensed for $100 per year, per employee. You get everything but an operating system and a database, and can run it all on any machine in your company. Sun doesn't care. The other interesting one is the $1-per-CPU-per-hour pricing on server processing capacity, which Sun launched a few months ago. Sun wants partners to set up giant data centers and use Internet technology to connect users, mostly with non-OLTP workloads, to these centers and charge them by the CPU per hour.

Take a closer look at this server pricing model. To use a single processor for a year will cost $8,736. If a company built a utility data center using two-way servers, it would have the potential of $17,472 in revenue from a single box if it sold 100 percent of the hours in a year. This won't happen, of course, and Sun and its partners won't get anywhere near $1 per CPU per hour, either, because that price is too high. With a two-way X86 server costing around $3,500 with a reasonably hefty configuration, and another $2,000 or so a year to manage it, there is a ridiculous margin spread in this subscription approach, provided that the companies delivering servers on a subscription basis can keep utilization high and not discount too much. With these numbers, which are just provided for illustration, you'd have to set up logical partitions on that two-way server to isolate a processor, but you would be able to break even on costs with only 31 percent utilization over the course of a year. Assuming a 40 percent discount off that $1 per CPU per hour, you'd have to substantially lower the cost of supporting the server (perhaps $1,000 a year) and jack up utilization (perhaps to 60 percent) to get a decent margin of 28 percent.

That sounds like a very appealing pricing model, but customers come and go as they please in such pricing models. Sun appears to want to allow customers to subscribe to systems they keep on site and use as if they own them. And even if the pricing is very aggressive, such a dramatic change will have adverse effects on Sun's top and bottom lines as it makes the transition from a company that sells servers each quarter to one that distributes servers through a channel with partners and collects monthly, quarterly, or annual subscriptions to cover the costs of those servers. Ironically, if Sun prices this correctly, customers may all of a sudden stop buying servers and start subscribing to them, disrupting the company's short-term financial results, perhaps even more than the dot-com bubble bursting did.

This is particularly true if customers can cancel contracts at any time without substantial penalties. If you sell a company a server, you have essentially asked it to prepay for the use of the server over the course of 36 or 48 months, which is the economic life of a server. Having done that, companies are inclined to squeeze all of that money back out of the server. They buy it, and then they use it. With a monthly or quarterly subscription model, they can change their minds and acquire other products, just as we change telecom providers on a whim. Customer loyalty will have to come from service, not from the lock-in of a large investment.

If this subscription model flies, and it will be very appealing among the millions of small businesses that are always tight on cash and don't want to shell it out on big IT investments that lock up that cash and put iron on their books. And Sun is the first one to make the transition; it will have first-mover status and the momentum that this status often imparts. This is clearly something that Sun is counting on, and if it succeeds, you can bet other IT players will quickly adjust their own sales forces and reseller channels to go from product sales to subscriptions.