Depends.
If a researcher decides that they don't need or want the things a publisher offers - some mixture of peer review, editing, branding, marketing, hosting, printing and distribution (if dead tree versions are still being produced) - then they don't need to use a publisher, no. They can write whatever they like and stick it on their own blog, which their institution probably pays the costs of.
And a number of academics do in fact do this. Though it probably helps if they're sufficiently advanced in their career that they've already established a reputation / brand. How seriously would you take a random blog written by someone you've never heard of? - not very, not least because you'd probably never come across it in the first place.
If a researcher *does* want those things that a publisher does then someone has to pay for them. Whether the publisher is for profit or not is just the standard public / private debate - sometimes public sector is cheaper, and sometimes it isn't.