We interviewed people in leadership positions from 11 infrastructure projects that weredetermined to have achieved long-term sustainability. The goal was to characterize theapproaches that lead to sustainability for each of these projects, and document if andwhen these organizations had difficulty. The work was originally an outgrowth of theEarthCube Project Council of Funded Projects. Protocol development, interviews, anddiscussion was conducted by a group of eight people with expertise on different aspects ofdata infrastructure. Almost all the studied projects started as federal funded research projects. Most projects(only one had not) had successfully made the transition from a project to an organization.This transition was a challenge for many projects, as it often involved changes inleadership and organizational funding model. The studied projects divided naturally intothree distinct types – Database, Framework, and Middleware – and some generalities forsuccess were noted for the different project types. The Framework and Database projectwere strongly supported by a disciplinary community, which provided the intellectualcommitment. Successful Databsase projects were critical for the workflow for one or moredisciplinary communities. Framework projects generally maintained active participation ofa community by utilizing bottom-up governance approaches. Middleware projectsapproached sustainability models in ways similar to software companies; success alsorequired maintaining strong scientific partnerships. The implications from our study include the following. First, science cyberinfrastructureneeds resources to develop and sustain itself; the mechanisms/guidelines/funding sourcesfor these supporting these activities are not worked out. At present, many resources areprovided through in-kind support from academics, researchers, and their institutes.Second, effort is necessary to find appropriate models that help sustain infrastructure forearth science over the long-term. Third, organizations – rather than projects – appear toprovide the best platform for long-term stability for earth science infrastructure. Umbrellaorganizations may be necessary for development of new Database projects forcommunities that are underserved in their digital needs.