It's not cheap, and you don't have to be Google for it to matter. Most of the time, the costs are just non-obvious to engineering. Here are some rough numbers.

It's not cheap, and you don't have to be Google for it to matter. Most of the time, the costs are just non-obvious to engineering. Here are some rough numbers. Power for a single 48U rack usually runs $200-$300/month for two 20amp circuits (possibly +2 for redundancy). A modern 1U server draws anywhere from (roughly) 1.66 to 2.5 amps, and you can only use 80% of a circuit's maximum supported load due to safety regulations, resulting in a total supported count of between 12 to 19 mid-range servers per rack. Higher-end servers have a significantly increased power draw, but potentially higher efficiency, and there can be a significant advantage to leveraging multi-core systems. In addition to rackspace and power, each server must obviously be connected to the network. The switches draw enough power to potentially decrease the number of supported servers. Each rack costs roughly $500/month (plus power), and often requires advanced provisioning to ensure that rackspace is available, and if possible, located in the same area. It's not unusual to buy an entire block of space well ahead of actual use to ensure that the business' growth can be accommodated. Additionally, one requires an operations staff to maintain the servers. Hard drives fail, backups must be done, hardware must be ordered, tested, and finally, deployed. Simple deployment of a single server, including the networking configuration, travel time, and physical work can easily consume 4-8 hours of operations staff time. When you add together the operational and capital expenditures, I'm not really convinced that the idea is so simple as "servers are cheap, programmers are expensive". Servers, electricity, physical space, and the labor necessary to install and maintain them are not particularly cheap. I'd propose a new mantra: programmers are expensive, and so are servers and IT, and so let's try to not be monomaniacally flippant about it. Software can be easy to write AND perform reasonably well, and to say otherwise simply appears to be a false dichotomy fronted by those who benefit from it most.