Intangible goods are always hard to price, and software is no exception. Advancing technology is making it harder.

Server software typically had a per-CPU price. This provided a very rough proxy for value—companies running the software on a single CPU were presumably using the software less (and receiving less value, perhaps) than companies running on multiple CPUs. The CPUs could be in different computers, or in powerful multiprocessor machines. Software vendors also typically differentiated based on the processor type, charging more for software that ran on more expensive UNIX based computers from the likes of Sun and IBM than on cheaper x86 hardware running Microsoft Windows or Linux. This was another way to use a proxy for value. The more expensive machines ran in larger companies on larger, more mission critical tasks.

Three trends have muddied the waters.

First, x86 based systems may now rival or even surpass the performance of more expensive UNIX systems in many situations (Google’s infrastructure is x86-based).

Second (and the real focus of this post), is that new chips will have “dual cores” meaning that they essentially have 2 chips in one. Microsoft prominently announced that they would treat dual cores as a single chip for licensing purposes. Microsoft probably has little choice in the matter, since many chips will be dual cores by the time Longhorn ships, and Microsoft is desperate for people to upgrade en mass. Doubling prices is beyond the pricing power of even the Redmond giant. Oracle announced that it, too would price dual cores as one. IBM had held out on the position that two cores meant two chips and twice the price. Last week, they partly changed course, announcing that x86 dual cores would count as a single processor, but the company’’s own Power processors will have a per core price. IBM’s approach partly reflects technology. Customers will probably see a bigger performance boost from the additional core on the IBM chips. And partly this reflects the old value proxy. Companies are more likely to run their most mission critical applications on the pricier chips.

The problem is only going to get worse as chipmakers offer chips with 4 or 8 cores.

Which brings us to our third trend: software that runs entirely over the web from a central server that the customer does not even own is catching on. This is a great topic in its own right and will get its own post.

Comments are closed.