Surely if you have 40 contemporary calls, then you have 40 Erlangs, and you need 53 circuits if blocking is 0.01 (1%) and 60 circuits if it is 0.001 (0.1%). Using the Erlang-B Calculator on line on this site. Mind you if the peak number of contemporary calls is 90 to 100 then you need 90 to 100 circuits to carry them, as calls are not unifor but very peaky. Would be interesting to know what is behind the circuit that cause this call demand. Also Peak Busy Hour is actually 2 and a half hours, so engineering to the average will under size, though engineering to the peak of 100 could oversize unless the peak is measured over a half to one hour duration.
Assuming 100 Erlangs and 0.1 (10%) blocking then 97 circuits are needed, increases to 117 if blocking is 0.01 (1%). Interestingly, at 0.15 (15%) blocking, 90 circuits are needed, while at 0.413 (41%) only 60 circuits are needed. Thus telecom costs for this link could be increased by 25% to 50% depending on how the problem is approached and what service level is demanded! In answer to the original question, at least 53, probably 60, maybe 100+ depending on what the traffic is going to or coming from. If it is provided over a carrier circuit, you probably need to have 3-5xT1 (24 channels per T1) or 2-4xE1 (30 channels per E1) and it may depend on what the customer is prepared to pay for. If the end user application is a call centre, then the customer may not want to lose a call so 120 circuits are needed. On the other hand it may be a 200 extension office PABX where it seems every extension makes an outgoing call around 5PM and is otherwise fairly constant with less than 1 in 4 extensions on calls over this time. In this case even 60 circuits may be too much.
One needs to understand the reason for the peak calls and the inconvenience to the caller/called party if the calls are blocked.