'Dumb' Computers

I had the unique opportunity to chat with Steven A. Ballmer '77, senior vice president of Microsoft, at an event sponsored by the Harvard Computer Society two weeks ago. The talk focused mostly on Microsoft's new Internet strategy, but towards the end of the discussion, Ballmer posed an interesting question:

"Would you rather buy a dumb computer for $500 or a decent computer for $1000?"

Balliner was challenging Oracle CEO Larry Ellison's recent announcement that the growth of the network meant the death of the personal computer. Ellison and others argue that the utility of computers lies in the information they provide, and that a stripped-down computer--essentially a monitor and a keyboard--connected to the Internet was just as useful (if not moreso) than a very powerful computer sitting on one's desktop.

Several companies have adopted this philosophy, including Oracle and IBM, and are now developing so-called "dumb PC's." Although Ballmer questioned this philosophy at the meeting, he said that Microsoft would develop software for whatever machine people wanted to use.

So the question remains, would you rather have a dumb computer connected to a network or a smart one? I'd rather have a smart computer connected to a network than a dumb one.


Consider first the fundamental question: Where does the utility of the computer lie? Ellison is correct in claiming that computers are useful as a means of accessing information. However, what's at stake is not the information itself, but how the computer accesses it and what it can do with it once it has it.

What can a $500 dumb computer do? Not much. A decent 17-inch color monitor costs about $700. Most people want to see their information in color on a nice, big screen; how will IBM and Oracle overcome this difficulty?

Powerful applications require lots of Random Access Memory (RAM). Minimum requirements for most operating systems plus applications are at least 16 megabytes of RAM. Presumably, a dumb computer on a network using modularized applications will mean the demise of RAM-hungry applications. Assuming this is true and that a dumb PC only requires about 8 megabytes of RAM, this is still about $250.

A networked computer supposedly won't need a lot of hard disk space, and hard disks are cheap these days anyway. A 500 megabyte hard drive costs a little over $100. A good keyboard costs about $70. A cheap processor costs around $100. A decent Ethernet card costs around $100.

Assuming that somehow, companies will reduce their costs by packaging all of this together, you have a mediocre to poor, monitor-less computer selling for about $500. Considering you can buy a fairly powerful machine these days for about $1,000 (with prices still dropping), why would anyone buy these dumb computers?

Do Oracle, IBM and other companies know something we don't? I doubt it. These companies have made it their goal to break the so-called "Wintel" dominance over the computer market (the large number of users using Microsoft products on Intel-based machines). While this announcement certainly raises the question over the future effectiveness of the personal computer model, realistically, Ellison and his peers are spewing hype rather than predicting the future.

Eugene E. Kim '96 is former president of the Harvard Computer Society. He may be reached on-line at ""