"So you have to keep a large amount of that part in stock." That part in stock are bags of plactic chips, and spools of plastic. What's to stop the manufacturer from building something else for someone else. For example, Furniture, or Hand Tools?
Not necessarily. Half the time when they're talking about 3d printing like this they're referring to SLS, or printing metal. Your feed stock is this metal powder that these hundred thousand dollar (minimum) printers use.
Now back to your question. It's the Auto manufacturers that we're talking about. They make cars. Sure they could use those printers to make other things, but that's not their core competency. The big thing is that 3d printing is expensive and takes a while. Here's an example:
GM requires special part A. Every car has one, and only one of that part. No one but GM uses that part.
In a traditional world GM would have to go find a supplier and have them do "a run" of that part. This typically has months of lead time, and GM ends up with a warehouse full of part A. If they change lines or find a problem with that part they're F***ed. There only real options are to shut down the lines while waiting for new parts or use the old defective ones. This is probably what happened with GM's ignition switches, and why they didn't want to admit it was an issue.
Just In Time manufacturing says do away with the warehouse and have the factory deliver a truck load at a time. This is a super efficient way of doing business, and it tends to avoid some of the issues I mentioned above. The problem is that you must have guaranteed deliveries, either on a schedule or with a rapid turnaround time. A great quote from The Goal is, "the closer you come to a balanced plant, the closer you are to bankruptcy." So when you have a crappy supplier you're once again F***ed and have to shut the lines down, or pay someone else stupid amounts of money to do a rush order.
Now, you might ask, "why doesn't GM just produce the part themselves?" Well, GM is a car company, not a widget company. They don't want to spend the money on machines that don't make cars. They don't really have the expertise to make that part. If they do run into an issue with the part, it's going to cost them quite a bit of money to retool all those machines to fix the issue. In short, it's also a pita. Tesla Motors actually does this, and they've run into some quality issues because now one company is doing so much more than just make cars. It can easily turn into a Jack of All trades, master of none situation, and business is all about being the master of your specific trade.
3d printing presents a solution for some extremely specific parts that match the criteria above. These parts are needed in low quantity, be extremely risky/costly to store in bulk in a warehouse, and the suppliers can not be relied on to make regular deliveries. It's only worth it if 3d printing the part is less of a hassle than dealing with any of that. Furthermore, you still have a final assembly step, and not everything can be 3d printed. Ignition switches can be more complicated than you think, so they might or might not be something that it's worth to 3d print. If you get past all of that, it might make sense to 3d print some parts.