Simple: they don't!
Like, I made my own from the proverbial box of scraps; but the firmware is proprietary, and it's just a shitty serial link. I don't expect it'll run with anything else (though it should run, I think, with a few user-created tools; I just haven't bothered to try yet), and I don't have anything else I have to use it with (i.e. VISA drivers, Excel plugins, etc...), so it's fine with me.
Classic adapters, are either needed for compatibility with existing software and hardware (replacing broken components on old installations, where updating hard/software is impossible for a variety of reasons), or because they Just Work(TM) and have the best drivers/compatibility/support.
There also just isn't much product movement (assumption), so the price is highly variable. GPIB is old, mostly obsolete; a lot of home lab users just don't need it much, or at all. Most equipment comes with USB/Ethernet these days, with mfg-supplied drivers, or operated directly from a browser.
Put another way: how would you solve the apparent problem? If they're as expensive as you perceive, surely there is profit to be made. How much time will it take you to create your own clone? And, can you do so without infringing whatever IP exists? (NI might not appreciate your cloning their USB PID/etc., for example.) Will you sell enough, at the lower price point, to make back your investment?
The market seems to agree: no, you won't. Granted, real markets are very, very far from the omniscient optimizers certain people (a LOT of people, actually..) want to think they are. Maybe there is indeed an unseen opportunity here. You can always try!
I think a better way to view it would be: to a certain extent, the well known, classic, brand name devices remain expensive, because they are sought after -- whether merely as collectibles (unlikely, but probably a few?!), or for a purpose. Either way, they're moving out of circulation, at least at a given price point, so these are functionally similar outcomes. Meanwhile, for those who are more flexible (tolerant of poorer quality tools, no support), there are myriad cheaper options, including "free". Put another way: the confusion seems to revolve around highly visible, brand-name items; rather than comparing them to themselves, compare to the myriad poorly visible, no-name or even free items/solutions. Which, being poorly visible and numerous, are... hard to see, hence the confusion.
Which, to solve that, try searching relevant keywords: clones (especially of moderately-well-known adapters e.g. Prologix); Github for GPIB adapter sources; Tindie, OSH Park, etc. for proprietary or free designs; etc.
Examples:
OSH Park, GPIBTim