Sometimes they do. Regardless, if it's often coming from funding that implies the authors have to have grants or come from institutions that are willing to pay, which disadvantages individuals without resources (they often have discounts for those without resources but in my personal experience the discount is absurdly small, like 15%, which is a lot for someone who just doesn't have external resources to publish under that model, if they're asking for 2k+ per paper).
Either way it creates a backward incentive structure, of pay to access.