What evidence do we have that significant quantities of molecular gas could be ``hiding'' in the ISM of many galaxies? Well, we do know that many galaxies contain dark matter, and at present there seems to be nothing wrong with assuming that at least some part of that dark matter is in the form of a disk of molecular gas. Low-density molecular gas could escape detection since the excitation of the CO molecules may be subthermal, but in that case it is not likely that much mass is involved. If gravitationally-significant amounts of H2 are present, say M(H2) M(HI), then the molecular gas would have to be more dense, and also generally cold in order to escape detection in the extensive CO(1-0) emission line surveys of galaxy disks. Such gas could exist in well-shielded parts of galaxies where the UV flux and cosmic ray density are both very low.
The inner disk of M31 is just such a ``cold'' environment, and it was here where an initial detection of faint CO emission emanating from large dark dust clouds was first reported (Allen & Lequeux 1993). Modelling of this data (Allen et al. 1995) confirmed that the ``cold gas'' interpretation was viable. Subsequent observations discovered similar emission in other dust clouds in the inner disk (Loinard, Allen, & Lequeux 1996), and also showed that such emission is ubiquitous in the inner disk of M31 (Loinard, Allen, & Lequeux 1995), covering more than half of the annular area between about 1 and 8 kpc in galactocentric radius. The distinction between cold, extensive gas and small, warm, beam-diluted condensations was made in favor of the former with a combination of JCMT CO(3-2) and OVRO CO(1-0) synthesis-imaging observations on the dust cloud D478 by Loinard & Allen (1998). The bulk of the molecular gas in this GMC-sized cloud appears as an absorbing screen at ~ 3.5 K. There is no reason to presume that D478 is peculiar, so we must conclude that cold molecular gas is indeed extensively present in the inner disk of M31.
How much mass could be in such a component? Loinard & Allen (1998) suggested that if the GMCs were at least marginally bound, then H2 mass surface densities of order ~ 100 M pc-2 would be present in the inner disk of M31. However, very recently Pringle, Allen, & Lubow (2000) have argued that GMCs are unbound, transient objects that are not formed from in situ cooling of HI, but from the agglomeration of the dense phase of the ISM, much of which is already molecular. How much mass is in this component is presently the big question, but it's a delicate one that needs some careful thought, and I am not prepared to address it here any further just yet.