The Cost of Bandwidth in Internet Video Distribution

Why doesn't a broadband television system exist today that viably competes with the likes of cable and satellite? In other words, why can't we get our TV over broadband from any number of providers like we access websites with our browsers today? There are, in essence, two factors holding us back from this future. One is streaming technology over broadband. The other is the cost of bandwidth. We believe that if both of these issues are addressed, a single broadband TV tuner "browser" will also become predominant on the Internet -- just like our compatible NetScape, Internet Explorer, Safari, Opera, and Mozilla browsers did.

H.264 (MPEG-4), VC-1, and other codec technologies have dramatically reduced bandwidth requirements over the past couple of years. The following visual should help put his into perspective:

Premium Quality Video Requirements

In other words, on our 1.5, 4 and even 10 Mbps broadband connections, it is technically possible to begin delivering television services if they didn't have a best-effort delivery model. Unfortunately, as it is today, wholesale IP bandwidth can only be purchased on that best-effort basis. This is about to change, and InterStream is in the middle enabling a highly scalable model for premium guaranteed bandwidth. IP Transit is the Internet's current wholesale method of selling bandwidth. By enabling transit suppliers to offer streaming transit services, InterStream will enable a new form of wholesale bandwidth.

Cost on the other hand is another issue. Let's look at 'typical' transit pricing over the past six years:

Transit Pricing

These figures come from attending peering coordinator and NANOG meetings over this period. The pricing is representative of large "buys" or commitments by the purchaser. As we can see, prices seem to be declining by approximately 60% per year over this time.

Unfortunately, even at $17 per 'meg', transit prices cannot economically support "television streaming". Let's see how this plays out by looking at the numbers.

Transit is sold on a per megabit per second per month basis. In other words, a fixed capacity on the network is sold each month to a buyer. Therefore, there are different efficiencies by which that buyer uses the bandwidth. One way to express the efficiency is via a peak-to-mean (P:M). We divide the peak usage which corresponds to how much transit must be purchased by the average, or mean, consumption each month. Higher P:M correlates to greater inefficiency. Buyers who efficiently use the bandwidth by having many subscribers consuming content through the day and week, effectively lower their costs. It is reasonable to assume that a very efficient 'streamcaster' could have a P:M of 7:1.

With the figures in mind, we can calculate the average cost of streaming one hour of video at the last figure we have for transit service, $17 per meg. Let's assume a DVD (or standard definition) video can be streamed at 1 Mbps.

For a one meg of transit price, the cost of one hour of streaming can be amortized over the month:

24 hours/day * 30 days/month (on average) =

720 streaming hours per month

With each streaming minute costing a little more than 2 cents/hour at %100 efficiency, we can use our P:M of 7 to calculate our average efficiency. We multiply our cost by the P:M.


$0.0236/hour * 7/1 = ~17 cents per hour

If we consider Spencer Wang's view on Internet video distribution, we quickly discover that in our simple distribution model, costs (17 cents per streaming hour) currently overwhelm any revenue (57 cents per hour) that could be gained from offering TV over a new distribution medium. This analysis also does not take into account CAPEX, operational costs, or the premium that will be charged for a 'streaming' transit service.

In other words, from the television industry's point of view, the Internet would be a lower margin distribution medium than satellite or cable. This appears to be true, at least for standard defintition video -- not the low resolution video on a video iPod. For enhanced definition, and high definition, the numbers are probably about 4.5 and 2.5 times higher, respectively. As we know, however, wholesale IP bandwidth costs have been dropping by approximately 60% a year. In addition, technologies such as multi-cast can be used to make the network much more efficient at carrying TV. We're on the cusp a revolution as these bandwidth costs continue to decrease and new technologies are deployed to support video.

For a simple Excel spreadsheet to do your own calculations, click here.

Jeff Turner


Post new comment