Despite the huge number of contributions dealing with the evaluation of cellular networks performance, tackling with more and more complex systems including multi-tier networks or MIMO systems, the fundamental limits in terms of capacity in an information theory sense is not known for these networks. Stochastic geometry helped doing a step forward, relying on Palm theory and providing coverage statistic at the network scale. However, this statistic is not sufficient to establish a fundamental limit, namely to characterise a Shannon capacity region of the network. In this paper, we propose a new approach exploiting the cell capacity of the Spatial Continuum Broadcast Channel (SCBC) recently introduced for an isolated cell. The network capacity is linked to the cells' geometry statistics in a Voronoi tessellation. The fundamental limit is characterised by the minimal average cell power required in a network modelled as a Point Process (PP) to achieve a desired rate distribution. A direct relation is established between this minimum average power and the partial area statistics of the cells geometry, which constitute a sufficient statistic. Our approach is validated through Monte-Carlo simulations.