Models of the effective thermal conductivity lambda-e of packed granular beds can help in understanding why this parameter has such low values in hydride powder beds, as measured during the last 10 years. In particular, the Bauer and Schlunder (BS) model looks well adapted for hydride beds. Calculations of lambda-e for a reference bed show the significant influences of pressure and of grain diameter: the micrometric size of hydride grains is a major reason for low values of lambda-e. This suggests that aggregation of hydride grains with a metallic binder in centimetric pellets, as made by Bershadsky et al. (E. Bershadsky, Y. Josephy and M. Ron, J. Less-Common Met., 153 (1989) 65) is a promising method for improving lambda-e. Optimization taking into account every parameter still has to be performed. To this purpose the BS model yields values of lambda-e for the packed bed, but the "wall heat transfer coefficient" of hydride packed beds must now be investigated.