What prevents attackers from faking stakes in proof of stake chains since they don't have to solve computational puzzles?

Question

In proof of work, an attacker has to find a nonce that when hashed will satisfy the "target" requirement. That means it takes computational power to create a fake bloc