The recession has coincided with a decline in the sale of servers masking the fact that selling servers is no longer the excellent business it once was. According to IDC, 2008 global server revenue fell 3.3 percent to $53.3 billion, although unit sales increased by 2 percent. This masks the declines of the fourth quarter which is when the recession began. In that quarter:

• High-end enterprise server sales fell 7.5 percent.
• Midrange enterprise server demand fell 14.5 percent
• Volume (i.e. commodity) systems revenue fell 16.8 percent year-on-year

This pattern will undoubtedly repeat in the next quarter (figures not yet in) and beyond. The simple fact is that Moore’s Law has finally caught up with the server business.

Virtualization Did It, In Collusion with Moore’s Law

You’ll notice that the decline in sales of high-end servers and midrange servers is less than the decline in sales of commodity servers, which is what you’d expect. In fact difference gives us a rough rule of thumb to judge what is happening. The decline in sales due to the recession is probably close to the 7.5% decline of high end servers. High end servers are much more efficient in their management of resource and the fall off in sales revenues there probably correspond to a decline in commercial activity. The decline in mid-range servers and commodity servers is too high to be explained in that way and is undoubtedly a gift from Moore’s Law delivered via virtualization.

The inefficiency of commodity servers is almost legendary. It just isn’t a surprise any more to hear of data centers where the average server utilization is 5-10%. Virtualization, using software from VMware or Citrix can increase that percentage significantly and when it does it eliminates the need to buy more servers especially commodity servers. Let’s do some math here:

Say you have a data center with 1000 servers and, in a consolidation exercise using virtualization you reduce the number of servers required to 750.

That’s normally achievable since, on average half the servers in a data center are running inefficiently. So now you have 250 spare servers. But on average you replenish data center servers, say, every 3 years. So you will still need to buy servers but instead of 333, you now only need about 250 and you may not need that many because you may be able to use some of the 250 redundant servers rather than sell them on eBay. So your server consumption has dropped by 30 percent.

This won’t happen all at once, because consolidation projects take time. There’s also the possibility that you’ll need new servers for new applications (like Voice over IP, maybe). There may be some underlying server growth – even in a recession.

Now let’s introduce the cloud into the equation and let’s suggest that, in the medium term, the cloud will pretty much eat up the underlying server growth in the data center. Then you’re left with the possibility, which I regard as a likelihood, that the server market in the corporation stagnates.

The Bottom Line

The more I think about this, the more I have to conclude that the corporate server market is in terminal decline. Indeed it looks like it’s caught between a rock and a hard place. Innovation will only make servers more efficient or have a neutral effect, because they offer better power management, say. The only thing that could save the situation would be compelling new applications, but they will almost certainly be delivered from the cloud.