We show quadratic capacity scaling in wireless networks with randomly located 
source-destination pairs is possible if each node is allowed to have multiple antennas. 
We assume each node has a vanishing size with respect to the inter-nodal distance.
Under this scenario, we derive "pairing" upper bound, which limits the best scaling
to quadratic regardless of the number of antennas per node. Another upperbound 
on the performance follows from recently discovered limit on the maximum degrees 
of freedom in wireless networks by Franceschetti et al. By using a MIMO version 
of hierarchical cooperation we show the minimum of the two bounds is indeed 
achievable, i.e., quadratic scaling up to a certain node density.