GPU-based rigs give you quite a bit of flexibility as far as scalability is concerned. You can start from an extra graphics card or two in an existing desktop system, to a dedicated 6+ card rig, to multiple rigs around the home, or even renting out space to run many rigs.
ASICs are a different beast. Running a single ASIC in the home isn't as popular of a choice as picking up some extra graphics cards for small-time mining, since ASICs tend to be optimized for space-efficiency, and thus often have a compact body with one or more high RPM fans (which are NOISY).
Large ASIC-based mining farms can be more efficient with physical space, however.
Resale value is [arguably] more of a gamble than with GPU-based rigs, since it depends greatly on the price of the coins based on the mining algorithm the unit is capable of mining, and how much of an upgrade future ASIC chips are. (The value of the unit is solely tied to its ability to generate revenue for the user, depending on how much a potential buyer values the cryptos it can mine [usually, but not always, based on current or projected exchange rates to fiat currency]. Either that, or its ability to act as a paper weight or be scrapped for materials.)
Back in 2012, a lot of people jumped on the BFL bandwagon for preordering SHA256 ASICs, and many of us who didn't buy into it held the exact same concerns we have now: Why would a company sell the hardware instead of mining with it on their own? Are they setting some units aside to mine for themselves, and hedging their bets by selling the hardware for immediate gains? (Meanwhile shipping units out months late...) Do investors believe that mining would be unprofitable and just want to take advantage of the demand for dedicated miners?
On another note, as always, operating costs (primarily electricity, but also those associated with dedicated mining facilities, if applicable) are an important factor for profitability. Do research keeping your specific conditions in mind.
True, but betting against the profitability of directly mining gold was a safer bet than betting against cryptocurrency mining profitability, at least in my opinion. (And operating picks/shovels requires an additional investment in manpower, for each set of tools, while ASICs have relatively low operating costs per unit.)