Here’s my prediction: space in data centers is going to become almost free.
A convergence of trends is going to crash the price, and business model, of many Cloud providers in the next two years.
Those trends are four.
2> Blade servers and ARM-based blades in particular are off-the-hook available now. The hardware is maturing, which brings farm costs down, down, down. At the same time, projecting growing demand, cloud providers are pouring money into the emerging trend. Result: massive overbuild.
3> Desktop supercomputers are becoming a reality, through three main channels: GPU applications on fast graphics cards, cute desktop supercompuers and the drift towards multicore in general. Once programming languages mature to use multicore machines (golang, I’m looking at you here) huge-core machines will come soon after. Huge compute is coming home, not on the amazon clusters. It’s a lot more convenient to have the massive compute right in front of you than in a data center in many cases (moving terabyte datasets, particularly.)
4> Bittorrent is evolving to do Youtube’s job with a new P2P architecture and Dropbox’s job with a new sync app. Skype, Spotify etc. are already P2P apps. Bitcoin’s proven huge P2P for complex collaborative computing is a fact, and the swarm of DHT applications which bitcoin will inspire will soon-enough move all kinds of common “cloud” operations to the p2p networks. Expect a bunch of weird P2P hybrid physical devices like Space Monkey for backups and Freedombox for privacy. WebRTC makes it easy to push P2P right down into the browser, further enabling web apps without server farms.
Put it all together and the result is pretty clear: Cloud is going to die much faster than anybody expected, probably with a massive over-build of server farms which will sit idle, soon to be replaced by a new style of P2P apps and devices, desktop parallel supercomputers. A major industry shakeout will result.