A significant difference between a conventional keyboard and steno is that a conventional keyboard is capable of producing all the of the text we normally type, and a steno machine is not. A necessary part of stenotype is a combination of pre-work (dictionary development for e.g. proper names) and post-work (editing the steno transcript to produce a "real English" document instead of one that may only have the generally correct sounds). Modern steno software helps a lot by partially automating these steps but steno is still inherently not capable of producing correct spelling without manual assistance - the basic architecture of steno is that you type the phonemes and (in modern usage) software guesses the correct spelling based on a dictionary. Much of the speed advantage of steno comes from the basic fact that it is a "lossy" process in the information-theoretic sense, that is, the "text" that you enter does not contain spelling information, only pronunciation... and even pronunciation is sometimes a simplified or substitute form as the American steno machine can't represent all of the phonemes that see use in English (mostly due to borrow words).

In the end, steno itself is probably not a lot harder to learn than QWERTY (although I think more frustrating because the "hunt and peck" option for steno is less intuitive and often slower). But it requires sort of a "supporting ecosystem" of skills and tools that is more complex and not amenable to use cases other than natural language. That makes it much less attractive for general use.