A Wall Street Journal article, Spending Soars on Internet Plumbing on 7/20/2010 chronicles the growth in the investment in the servers that form the compute backbone of the Internet. The following examples were citied:
- ThePlanet.com is buying “thousands” of servers this year and now owns about 50,000 Dell Servers
- IBM stated that their sales of industry standard (Intel/AMD) servers jumped 30% this quarter after jumping 35% last quarter
- Google reported $486 million in capital spending in the most recent period – more than triple the amount in the same period a year ago
- Rackspace Hosting says it added 9152 servers in 2009 and an additional 3000 more in the first quarter of this year
- Savvis another hosting company said that it has purchased 80% more servers in the last twelve months
The article points out two drivers of these purchases. The first is the benefits of the new Intel Xeon chips which provide 8 cores per CPU and up to 4 CPU per server. These new 32 core servers provide dramatic increases in price/performance as well as greater density and energy efficiency.
The second driver is mentioned and that is the demand for these new systems on the part of vendors building out their cloud infrastructure. Notice that ThePlanet.com, Google, Rackspace and Savvis mentioned above are all cloud vendors. Terremark is also mentioned in the article as saying that they are experiencing 30% increases in sequential demand each quarter and that there is “zero chance of a bubble”.
It is this level of investment in the infrastructure for public clouds that very much reminds me of the enthusiasm and hype that surrounded the previous version of cloud, known as ASP’s (Application Service Providers). Now an ASP is really nothing more or less than a someone limited version of SaaS (most ASP offerings did not have the easy signup and billing that characterizes today’s SaaS offerings). There were also not robust IaaS and PaaS offerings back during the ASP era. It is also important to note that the ASP era ended badly with many failed companies and many more bankruptcies all part of the larger .com crash that occurred around 1999 and 2000.
So the important question is this. Are the cloud vendors investing ahead of the curve in order to make sure that they are there when the demand for their public cloud offerings shows up? Or are they as Terremark says merely responding to real increases in demand for their business and building out to meet that capacity? In the answer to this question lies the answer to a larger question. How real will public cloud computing be and how soon? It is probably a good guess that vendors like Microsoft and Google are investing ahead of the demand curve in the supply of their cloud infrastructure (a build it and they will come strategy). Where the other hosting/cloud vendors are in between Microsoft and Google at one end of the continuum and Savvis who is “responding to demand” is hard to know.
If history is any guide, the industry collectively will over-provision the capacity for public cloud computing sometime in the next couple of years. This will likely occur as once all of the applications that are “easy” to move to the cloud get moved, and we go through a period of figuring out what it will take to move the next set (leaving those that will likely never get moved). This time period is most likely to be a couple of years off, but if history is any guide we will have a cloud crash just like we had an ASP crash and a .com crash.