The dark-matter prediction is usually considered as one of the successes of low-energy supersymmetry. We argue that, after LEP constraints are taken into account, the correct prediction for the dark-matter density, at a quantitative level, is no longer a natural consequence of supersymmetry, but it requires special relations among parameters, highly sensitive to small variations. This is analogous to the problem of electroweak-symmetry breaking, where the correct value of the Z mass is obtained only with a certain degree of fine tuning. In the general parameter space of low-energy supersymmetry, one of the most plausible solution to reproduce the correct value of the dark-matter density is the well-tempered neutralino, which corresponds to the boundary between a pure Bino and a pure Higgsino or Wino. We study the properties of well-tempered neutralinos and we propose a simple limit of split supersymmetry that realizes this situation.

Latex2e, 29 pages, 5 figures, reference added, typo corrected, version to be published in NPB