I've seen several conferences go this route recently, and both costs and the fees charged have varied wildly, but for good reasons.
The key components driving cost are, in decreasing order:
- Salaried staff (especially if the conference is a major support for a society or similar organization): tens or hundreds of thousands of dollars
- Publication fees (e.g., for archival proceedings of an IEEE conference): thousands of dollars
- Organizational costs (e.g., servers, corporate fees): thousands of dollars
- Teleconferencing software: hundreds of dollars
The conference fees thus shift accordingly. The models that I have seen so far are:
- The large conference supported by full-time employees didn't change its fees much at all.
- The mid-sized conference associated with IEEE charged authors about half the normal price (since their papers were the dominant remaining cost), and everybody else a pittance.
- The mid-sized independent conference charged PI and industry attendants a couple hundred dollars to cover their fixed costs, while students were free.
- The small interest group meeting made everything free.
If (as it sounds) you are running a conference for just a few hundred people using all volunteer labor and don't have fixed costs, then something like a monthly Zoom (or equivalent) webinar subscription would likely be sufficient to manage all the online aspects (assuming you let connectivity be the problem of the attendees). That will cost only a few hundred dollars, and you can likely cover it with a low registration fee --- or maybe even just use an institutional account from one of the organizers.