Cosmological observations (supernovae at high redshift, cosmic microwave background, ...) provide compelling evidence for a component of dark energy. I will argue that this statement relies on several untested assumptions of the Λ cold dark matter model, especially regarding the properties of dark matter. Giving up the assumption that cold dark matter is a dissipationless fluid allows us to describe the observations without dark energy and suggests a link between the phenomenon of dark energy and the formation of non-linear structures. The key to understanding ''dark energy'' may be found from a solution of the averaging (backreaction) problem.