The sticker price of releasing a player is not his salary. It's the unamortized signing bonus that accelerates onto the cap the moment you cut him.
Here's how it works. Signing bonuses get prorated across every year of the contract, up to five years. While the player is on the roster, that prorated charge hits the cap annually, spread out, manageable. The day you release him, every remaining year of that prorated amount collapses into a single dead-cap charge — due immediately, against a team that no longer has the player. That's not a salary expense. That's duration risk. You issued long-duration cap debt when you signed him, and you cannot refinance it. You can only hold it or take the accelerated hit.
Front-loaded deals with large signing bonuses are exactly this: a bet that the player stays productive longer than the dead-cap cliff lasts. A $25M signing bonus on a four-year deal means roughly $5M prorated per year, but if you cut him after year two you're absorbing $10M in dead cap against a roster that moved on. Teams do this deal because the AAV looks clean in March and the cliff is someone else's problem in year three. Except the GM who structured it is usually still there in year three.
Every time a talking head says a team 'saved money' by releasing a player, check the dead-cap figure first. The savings are frequently negative. The math does not care about the press conference.