If you want to really split hairs, at the very least they affect the compilation time, since the optimizer has more indexes to consider. They also consume time to load and memory to store their metadata. It may seem a trivial amount, but consider the extreme case of an index with 16 columns over 1000 partitions: it requires 16000 metadata objects to read and store in memory, it may add up to quite a bit. And by simply existing, they add storage space which translates into larger/slower full backups, more space needed for ops, slower restores etc.
But, realistically, a redundant index does not affect query performance. I know of some examples when certain indexes may fool the optimizer into generating a bad (slow) plan, but that is a separate topic.