

You need to know which basis the sender use to collapse and measure in the same basis. Then you need to sample a statistical distribution and the desired information will be the average of the distribution. This is very well proven in the Bells inequality experiment and can definitely be used to gain information.
It is clearly not very efficient in the sense a lot of transported bits are wasted to convey less information. But the advantages of instantaneous and secure communication will be worth it in some use cases.
That is, of course, if the engineering issues such as quantum repeaters (a sort of range extender) and high fidelity storage are properly solved. It is a few years ago since I did any quantum information in uni, so I don’t know what the current state of things are.
Why not both?