With Internet Service Providers (ISPs) in Canada and the United States planning to impose monthly bandwidth caps on us, and with extra fees being planned for people who exceed those caps (e.g. $1 per gigabyte), it brings up a great question – how much does it really cost to provide Internet access?
We can get a sense of that by looking at some of the most expensive bandwidth in the world – the bandwidth being provided to Africa. This article provides some interesting data:
A new undersea telecommunications cable to boost Africa’s Internet access will land this month, mobile service provider MTN said Thursday, calling it the continent’s largest data pipeline yet.
The 14,000-kilometer (8,700-mile) West Africa Cable System (WACS) fiber optic line is scheduled to reach South Africa’s Western Cape province on April 18, linking the continent’s Internet providers directly to the servers of Europe and boosting the bandwidth of the world’s least-connected region.
The cable starts in London and will connect 15 points along Africa’s western coast, said MTN, which has a $90-million (63-million-euro) stake in the $650-million system and is the project’s single largest investor.
From this article we learn several important facts:
- The cable costs $650 million
- The cable provides 5,120-gigabits/second
- The cable provides 15 access points along the coast
- The cable is incredibly long: 14,000-kilometer (8,700-mile). It could go from Los Angeles to New York City and back to LA and then back to NYC:
Now let’s make several of assumptions:
- The cable will need to be maintained. Let’s round the $650 million price tag up to $1 billion to cover that expense.
- The cable will eventually become obsolete. Let’s imagine that happening on an aggressive schedule of 10 years, given the pace of technology.
- The people who put this cable in place wish to make a profit on their investment. Let’s assume a 200% return on investment (20% per year). So, after investing $1 billion, the operators wish to receive $3 billion.
Given that cost structure, and the fact that the cable can handle (in round numbers) 5 terabits/second or 500 gigabytes/second, how much does a gigabyte cost in this system?
- Each year, the system costs $300 million
- Each year the system can transmit 500 gigabytes/second * 3,600 seconds/hour * 24 hours/day * 365 days/year = 15.7 billion gigabytes
- $300 million/15.7 billion gigabytes = 1.9 cents per gigabyte.
1.9 cents per gigabyte on a very expensive system. Remember that we already baked in a 200% profit margin. But even if you want to get greedy and mark that up an additional 100%, it is only 3.8 cents per gigabyte.
In other words, bandwidth by the gigabyte is incredibly inexpensive. Pennies per gigabyte. And that is on one of the most expensive systems we can imagine.
The question we have to ask, however, is this: does that number represent the true cost of bandwidth? And the answer to that question is no, because of congestion.
Note that people tend not to use bandwidth evenly throughout the day. If they did use it evenly, then the cost per gigabyte truly would be 1.9 cents per gigabyte. However, at 4AM this undersea cable is likely to be very underutilized. At other times during the day, demand is high.
The big uproar right now is the problem that Netflix and Hulu are causing. These online streaming services use a lot of bandwidth (relatively speaking), and usage tends to concentrate in the evening hours.
So let’s look at the very worst case scenario that Netflix/Hulu could cause. According to this article…
We know that the average encoding rate for video streamed to the Xbox 360 is about 2000Kbps. That means one person watching a two hour movie would transfer roughly 1.8GB of data. For high definition movies, the average encoding bitrate is around 3200Kbps and one user would transfer about 3GB of data.
Worst case, imagine that every customer wants to watch Netflix every night at exactly the same time. And let’s make it even worse – let’s imagine that a household has multiple people in it, and there are three simultaneous Netflix feeds going to each customer. Now we need approximately 10 megabits/second for each customer. During a three or four hour window every night, the demand caused by this Netflix loading caps the number of Internet accounts we can give out. The undersea cable can only handle 5 terabits/second, so we can have at most:
5 terabits per second / 10 megabits per second = 500,000 customers
So we have to spread the $3 billion cost of the cable over those 500,000 customers. What does that work out to per month? In 10 years there are 120 months, so:
$3 billion / 120 months / 500,000 customers = $50 per customer per month.
$50/month is the worst case scenario. We are assuming that every single customer will be watching 3 simultaneous high def Netflix movies (10 megabits/second) simultaneously. The pipeline can handle that kind of worst-case load, with a 200% profit margin, for $50/month. And that would be some of the most expensive bandwidth in the world.
What is the cost per gigabyte now? If you assume that each customer is allocated a true 10 megabits/second (call it 1 megabyte/second) pipe, and if you assume that each customer fully uses that allocation for 6 hours a day, each customer is pulling 21 gigabytes per day, or roughly 600 gigabytes per month. $50/600 = 8.3 cents per gigabyte.
In other words, Internet service can be provided profitably for pennies per gigabyte in the absolute worst case scenario.
Let’s imagine that we lived in a rational world, where ISPs were not trying to gouge customers and we were all trying to utilize resources efficiently. Netflix/Hulu really can increase the cost of bandwidth because they create congestion in finite pipelines. There are a number of things we could do, working together, to relieve that congestion. For example, bandwidth at low-usage times would be essentially free, and bandwidth at 8PM might cost a dime per gigabyte because many people are trying to use the pipe simultaneously.
If time-of-use pricing like that were put into place, what would happen? Many activities (like Internet backups, bit torrent, file uploads and downloads, etc.) would move to low-cost hours, spreading out the load. Places like Netflix and Hulu might offer customers services that pre-dowload what they want to watch tomorrow. People might voluntarily choose lower-bandwidth versions of shows they watch during peak hours. New applications might spring up to take advantage of essentially-free bandwidth in early morning hours. And so on. In other words, the marketplace would take actions to spread out bandwidth usage throughout the day to make better use of the resource, and the cost of bandwidth would go down.
We can dream…