Why do we charge for data transfer out but not for in?

In the early days of the Internet (think 25 years ago), there were effectively three kinds of networks:

1. Content providers / Hosters: Would host content to be accessed, e.g. websites
2. Content consumers (eyeball networks): Home or company networks were end-users would reside that consume this content.
3. ISPs (today called transit provider) connecting the two above together.

Keep in mind that at this time content would primarily flow into one direction. This pre-dates concepts like Peer-to-Peer or platforms like Youtube, where users upload masses of content every minute. But it is also the time, when ADSL was invented. Keep in mind that the "A" in ADSL stands for "Asymmetric" because the target group of that product (home users and small businesses) was expected to download more than they would upload.

The rest of the links - and especially the ones connecting content providers to other networks - were symmetric. While these links were heavily used by content providers in the outbound direction (to send out all this content), they were barely utilized in the inbound direction. ISPs would charge per bidirectional capacity (at 95th percentile) and therefore it would make sense to charge content providers only for outgoing traffic.

This principle just stuck throughout the industry since then and never changed, despite so many other things around the Internet changing.

A good summary of this can also be found here: https://medium.com/google-cloud/why-ingress-traffic-to-the-cloud-is-free-79dc217b916