In this paper, we demonstrate that asymmetry in fluid distribution tree networks emerges from power requirement minimization, under global volume constraint. We have discovered several levels of asymmetry in optimal trees: different pipe lengths at the same level of branching, different mass flow rates at junctions or bifurcations, and different main branches to build the optimal dendrite. The emergence of asymmetry in optimal tree networks (man-made or natural) is a result of the optimization: it is not an assumption or a modeling feature. The constructal method that we used to discover asymmetry is predictive, and this distinguishes it from descriptive methods such as fractal geometry. © 2005 American Institute of Physics.