You must be looking at very expensive GPS modules. The cheapest GPS solutions are so cheap that they are integrated into cameras and sometimes even toys.

The products I work with normally contains both GPS and GSM (or CDMA, WCDMA, ...) and the GPS modules are not (!) more expensive. Besides that, the GPS modules do not require any subscription fee, unless used with an RTK service where there is costs involved to get reference correction data allowing the GPS to produce positions good to within a fraction of a meter.

Ignoring GSM modules, and instead looking at GSM phones complete with a subscription, you can get the GSM phone for free, i.e. zero initial cost. The reason is that it is sold bound to a specific subscription. Do you really want to set out embedded equpment bound to a subscription? What would the total subscription cost be for a 10-year period? What service would the phone give if the user happens to forget to pay to prolong the subscription? What would be the administrative costs of keeping track of 1 or 10 or 100 or maybe even 1000 subscriptions?

What would the customer think if they do believe that their equipment is supplying a specific service, but it has silently stopped working because the GSM subscription have ended?

What if the operator ends selling the service? A new visit to all installation sites, to install a new SIM card with a subscription with another operator... And a certain percentage of SIM cards will fail, requiring replacement. If replaced, the technician needs either to deactivate the PIN code before installing the card (making the SIM card valuable to a thief) or needs to have an interface to teach the equipment the new PIN code. If used with GPRS/Edge/... for network communication (i.e. a data connection for access to a NTP server), the technician must also make sure that the installation knows the relevant APN and possibly PAP/CHAP settings.

If not using a network interface (which may add extra traffic costs), the device may need to make use of a proprietary API to request time from the phone network. Will that API work in case the customer has to change to a subscription with another operator?

Are you really, really sure that the GSM route is cheaper?

If this is a one-off installation - would the customer pay enough for the installation that it is meaningful with all the extra development time to handle that GSM solution instead of just extracting the time strings from a GPS module and have a local table for daylight savings and a setting for current time zone?