QGIS is a very powerful GIS environment. More and more features have kept coming in the 2.0 branch, thanks to a growing number of users, developers and funders. But it is occasionally time to look up and envision the future to make sure this growth of energy is used at its full potential, especially to make sure new features are not added to a base that will become hard to maintain or evolve. Discussions and active work have already been done about how to transition away from the obsolescence of Python 2 and Qt 4. Some API breaks will have to occur and this is an opportunity to include major changes, both for users and for developers. This talk will present some of the changes that are planned or wished for the 3.0 version of QGIS and will detail challenges that remain to see them exist, from a technical, organisational or economical point of view.