Cloud first, in-house later? The Zynga strategy

In a recent article, InformationWeek discussed the traditional use of the cloud:

The most frequently mentioned case for the hybrid approach of blending private and public clouds is "cloud bursting," where the in-house data center runs the bulk of the load, spots a spike building, and offloads that extra work to the public cloud.

But Zynga does the opposite, using the cloud first and then transitioning to an in-house data center:

Planning data center capacity for ... unpredictability is a slippery exercise. That's why Zynga instead launches games using Amazon's EC2 infrastructure as a service, so it pays only for the capacity it uses and is ready for spikes. But that's not the end of the story. Once a game hits a more predictable level, Zynga brings it in house, onto what it calls Z Cloud, servers it runs using a private cloud architecture similar to what Amazon runs.

And, of course, if a game tanks, Zynga doesn't move the game to Z Cloud. Instead, it cancels the Amazon service for that particular game, thus reducing its exposure.

Of course, this strategy only works if you have a massive in-house computing system in place already.