Infrastructure-as-a-Service cloud computing has profoundly changed the way we procure, manage, and use compute and storage resources. Since it allows users to easily provision remote resources on-demand, it enabled whole communities to treat the acquisition of compute and storage resources as an operational consideration rather than capital acquisition. The emergence of this new model raises many questions. How will infrastructure clouds change our work patterns? Is it a viable solution for scientific applications? What challenges need to be overcome to enable and facilitate a transition to this new computing model? The Nimbus project has been promoting the use of infrastructure clouds in the scientific community for nearly a decade. In this talk, I will describe what attracted various communities to infrastructure cloud computing, give examples of how they integrated this new model into their work, and describe the challenges they encountered while doing so. I will briefly discuss how those challenges shaped the development of a variety of features in Nimbus Infrastructure as well as infrastructure cloud platform which provides an integrated set of tools enabling deployment, auto-scaling and high availability across multiple academic and commercial clouds. I will discuss scientific applications currently making use of such multi-cloud environment as well as factors shaping the future of this rapidly developing field.