Michael Turk has an interesting discussion on what people would pay for gigabit broadband access which raises more interesting questions because many of the costs of bandwidth vary so greatly.  The reason for the variation is because there are so many different ways to sell bandwidth and there are so many other variables than just the peak throughput.  For example, the degree of sharing (fractional ownership) and the limits on usage (duty cycle) play a very large role in the cost of bandwidth.

Digital Society hosts its servers with 100 Mbps access in a “colocation” data center facility at a cost of $50 per month.  But we have to subtract $10 for rack space, cooling, and electricity costs so we’ll call that $40 per month per 100 Mbps of bandwidth.  But to get the cost this low, the service caps us to 1000 GB of usage allowance which effectively locks us down to a 3% average duty cycle or 1/30th fractional ownership.  That means we can only be active at full 100 Mbps speed 3% of the time or we can be active 100% of the time at 3 Mbps or more likely some combination in between these two extreme states.  With the average page view size of 60 KB, 1000 GB usage allowance allows us to serve 16 million page views a month.  This is extremely valuable to a small nonprofit group like Digital Society because we value the ability to handle higher peak loads and the affordability of the service.

But some companies that may want the ability to have higher duty cycles so they might purchase a dedicated 100 Mbps connection.  I’ve priced this type of service in the United States to cost around $1100 per month, and $10000 per month in Tokyo Japan.  That means the 1/30th fractional ownership service that Digital Society uses is 1/27th the cost of a dedicated 100 Mbps circuit which sounds fairly reasonable.

Broadband is also shared between 20 to 40 homes for each unit of capacity sold which might lead one to wonder why there isn’t 100 Mbps broadband service for $40/month, but there are additional factors such as the last few miles of the access network that is needed to reach your home.  In other words, buying bandwidth delivered to your home costs a lot more than buying it from a data center where the provider only needs to provision an Ethernet port and some rack space.  In fact, it’s extremely expensive to build access networks which is why network infrastructure companies have to hire 10 times more workers than content and application companies.  Just a single broadband provider AT&T has to employ over 280,000 workers to build, operate, and support their network.

The reality is that $43 per month can get 15 Mbps of cable broadband service which is highly shared because the total neighborhood of ~150 homes with might only have 80 Mbps of total capacity.  If we assume an average of 15 Mbps per home with 150 homes, each unit of capacity is shared 28 times which means 1/28th fractional ownership.

In denser environments like some apartment complexes in San Francisco, tenants can get 100 Mbps for $30/month.  That sounds like a really great deal until we realize that this is even higher over subscription.  The unit might have 150 tenants on the service which means 10,000 Mbps of capacity is sold but the backhaul might be a $4400/month OC3 circuit with 155 Mbps of capacity.  This would mean that the $30 of revenue from 150 users barely covers the transit and backhaul cost of the dedicated OC3 circuit, and the apartment complex is merely selling the service as an “amenity” to keep these tenants paying their high rents.  Furthermore, the true cost of the broadband service may be obscured by the cost of the rent.  The “last mile” in this case is CAT-5 or CAT-6 Ethernet cabling inside the apartment complex and that access network infrastructure is covered by the rent.  At $2000 per month for an 800 square foot apartment, the $30/month 1/65th ownership 100 Mbps broadband service doesn’t seem like such a great deal any more.

It’s also to important to understand that this apartment setup requires a high degree of bandwidth sharing between 97 users or 1/97th fractional ownership.  That means the actual delivered performance of the network is substantially slower than the advertised speeds suggest, and the advertised performance is much more exaggerated than typical single family home broadband services.

Note: A person living in a rural region might have to pay higher bandwidth costs, but they could probably rent or own a 2000 square foot home for $800 per month.  The irony is that those living in the suburban/urban areas have to subsidize the communication costs of rural residents, but the rural residents don’t have to subsidize the cost of rent in the suburban/urban areas.  Perhaps next time someone complains about high rural bandwidth costs, I’ll offer to swap rent prices for bandwidth prices.

All too often are inflated and inaccurate broadband comparisons being floated around when the reality is that it’s not very useful to compare peak bandwidth and price.  There are many other variables like the degree of bandwidth sharing and the costs that are hidden in rent or home owner association fees.  Any valid broadband comparison must take all of these factors into consideration.