Job,
A few clarifications as below:
The setup is all Google backed from the Gigabit uplink (zero cost to the ISP here)
The cost of the uplink (refreshing) is not catered for by Google. So about 30% of the traffic served by the cache consumes the uplink of the service provider(can be international/undersea capacity or Google peer).
All the ISP does is to peer with it ensuring all google related traffic stays within the network
True, the ISP can peer with Google...but in Mombasa. The main reason ISPs peer with Google is to have redundancy but the cost of picking it from Mombasa vs using undersea may not be too different. Our terrestrial connectivity costs are expensive and the traffic needs to be carried from Mombasa to Nairobi ...maybe at a cost of about 1,000/Mbps
other CDNs are doing the same
So far all other CDNs are also anchored in ISP networks and the big challenge still is who will absorb the cost of refreshing
My own opinion on this is lack of will to work together. For instance, if KIXP members came together and agreed to absorb the 30% cost of refreshing the caches and content delivery in Nairobi then it would be possible for say Google, Akamai, Cloudflare, etc to be pulled from KIXP in Nairobi and refreshed using the contributions by the members - or those who are pulling content from the CDNs. Nothing is at zero cost, there's a cost but the problem is evryone else wants it for free.
For the end user cost, our main hindrance in the country is the cost of acquisition and maintenance of the access network (prohibitive costs of licenses, permits, 'facilitation', etc) and this is the bulk of the cost for delivery of Internet services. Until the unit cost of access networks and terrestrial in-country networks reduce substantially the cost of delivering the content may not significantly drop.
But, I could be wrong.
Kennedy