Observations of extragalactic molecular clouds using CO to trace H2 reveal a surprisingly uniform population of clouds once biases for sensitivity and resolution are taken into account. This is unexpected because environment, especially metallicity and the local interstellar radiation field, is expected to have a strong influence on the structure of molecular clouds. On the other hand, when other tracers - especially far-IR dust emission - are used to model the distribution of H2 in the nearest galaxies, dramatic differences appear. I will present results from both methods of studying H2 in nearby galaxies and discuss how they can be reconciled.