In a lot of cases, the increase in cost would be minimal. Take larger print for user interface elements on an ereader, which would help people with limited vision. It would take additional development time, but the expense of that development time would be spread across the sale thousands to millions of units. Done properly, those software features can be used across multiple products and product lines to even further diminish the added expense. Even minor modifications to hardware (such as adding a headphone jack for TTS) would cost pennies per unit.

There are a lot of low hanging fruit out there that will make the life of disabled people easier, and even be useful for people without disabilities. I think that it is a bit selfish to pinch those pennies, nickles, and dimes in cases like that.

Now accessibility features that incur a major expense, are disruptive to the intended purpose of the device, and are highly specialized are a different issue. Designing an ereader with an integrated braille display or inputs for people with limited mobility would fit in that category, but surely developers could spend a bit of time incorporating features that interface with external assistive devices.