I had this debate about six months back; $3k for a 3D SLA printer, or $3k for a Chinese metal lathe? I rapidly determined I would be make far more useful objects, faster, using the lathe, and would be developing skills more fundamental than I would using a software/device that would be obsolete within a year.
3D printers themselves are presently in the '90's cell phone/home PC phase, where they are available, expensive but not prohibitively so, and more a curiosity for nerds than a practical tool in the vast majority of cases. They actually do work once set up properly.
3D printed parts are stuck somewhere even more primordial than that, however. Probably more like early alloy-steel or bakelite-era plastics. There are lots of possibilities on the horizon. The suitable applications for the materials available haven't been fully scoped out, yet, nor the full capabilities of the materials or unique additive-production techniques. The vast majority of items and parts are simply aping those designed for previous manufacture methods (injection molding, sheet metal, milling). It is still very common for designs/parts which are wholly incapable of fulfilling their task to be made by 3D printing users, simply due to quirks in the process/material rather than the design or promised specs.
I personally think the future is in 3D printing that utilizes long-chain cross-linking polymer matrix and oriented super-tensile fibers to 'weave' a fiber composite stronger and more optimized than any layup. Build up your internal form, then have the machine head crawl around it like a funnel-weaver spider, layering strategically-oriented strands of fiber composite until the part is built up. Boeing sort of has this with the enormous 'dream weaver' thing they use to build the Dreamliner fuselages from pre-preg CF tape, but a much smaller and more precise device laying a bead of material would be able to handle very diverse jobs.
Also, true closed-loop operation (self-sensing tooling) and AI-generated toolpaths will continue to reduce the difficulty and tedium of getting all these automated processes up and running. Eventually, you'll be able to use your solid model to directly initiate a part's construction with very little interference or tuning required, not unlike color inkjet printers in the last ten years, or so.
TCB