Federated learning - share of ROI

I am reading about federated learning and have a quick question

1) I know in federated learning, the model updates are shared to a central server

2) All the parties involved in FL can generate benefits because their model has seen more variation in data (due to different parties involved)

But my question is,

Let's say Site A contributes/has 80% (more data points) of the data and site B has only 20% (less data points).

So we know that in this case, Site B will benefit more from the model because with little data he gets to have a better model whereas Site A even with more data contribution has only very little to gain.

For ex in a commercial setting, let's say they want to build a model to detect fraud/anomalies.

In this who will get the most monetary returns? Is it Site A or Site B?

Can you please direct me to research articles, opinion articles and tutorial regarding this please?