Its possible to 3D print something as a prototype (a number of shops use stereo lithography for that) for a couple of hundred dollars. But that doesn't insure that it can actually be injection molded. You need a shape you can represent as a convex hull.
The process of developing the molds (aka putting them into a CAD system, slicing them, and then cutting the mold itself. Is called 'tooling.' Any tool and die shop can give you a quote. Back in '99 when the company I was with got quotes for tooling for a desktop gateway, a tool and die shop in Milpitas quoted us $36,000 out the door, and then $4,800 for each additional set of dies. A set of dies was good for between 100,000 and 500,000 uses. I don't doubt they were not the cheapest way to go, shops in China were willing to spread the tooling cost across an order of 100,000 units.
So are you trying to see if your design "fits"? If so get a stereo lithography shop to print one for you. If you are trying to see what it will cost to make take your printed part to a tool and die shop and ask for a quote. You will probably specify which type of machine you are going to use for the molding (dies are specific to machine models).
Toolmaking is not cheap. You're starting with a large block of steel or aluminum, and that's usually a custom alloy with extra hardness if you plan to make more than a few thousand. The molds wear out over time. So that's money right there.
Then it's a couple of weeks of milling, drilling, polishing, welding on support channels, etc. There's a lot of CNC time involved, and that's runtime on million-dollar machines with high maintenance and deprecation costs. If you want it done by hand by experienced machinists, that's even more money.
The Aeropress has some pretty tight tolerances on it's design, so there's more money right there. Size also kills. You want something the size of an Xbox housing? That's gonna be even more.
Protomold[1] are probably that niche, although I'd treat it more as proof of concept than as scalable as proper production molds. If you've validated your prototype that way, you'll probably end up with DFM[2] changes you want to make before dropping 10's of K (or more) on full production stuff.
