I'm not sure that my understanding about amortized analysis is correct. For example if we have a amortized cost that is O(1) and make m operations that would be a cost of O(m). Can we say that the worst-case is O(m)*m for m operations?
On a technicality only, since m × O(m) = O(m²) is an upper bound but not a lower bound.
By default, amortized algorithms start with no credit, so the actual total running time of those m operations is Θ(m), since credit cannot go negative.
Due to the amortization there are many possibilities for how this actual running time is divided between the operations. You could have one operation that takes Θ(m) actual time and the rest are Θ(1). You could have √m operations that each take Θ(√m) and the rest are Θ(1). They could all be Θ(1). What you can't have is total time ω(m).