What Does a Gig Cost?

The Montreal Gazette ran a major story over the weekend focused on the costs for ISPs to transport a gigabyte of data (picked up by others as well). As those following the usage based billing issue will know, the ISP overage costs – which run as high $10 per GB in Canada – have attracted the ire of customers and raised questions about the actual costs for ISPs.

Developing a better understanding of actual network costs was a big part of the paper I posted last week on UBB. This post features part of the discussion on costs, though the complicated appendix that uses Bell’s submission on network costs as part of the deferral account proceeding must be accessed from the original paper.

Costs related to Internet access pricing structures sit at the heart of the UBB debate, yet the most important data point remains shrouded in secrecy. The incumbent ISPs have long been reluctant to disclose their actual costs in maintaining their networks, arguing that the information is sensitive, confidential commercial data (though the CRTC has begun to push for greater disclosure from telcos and cable companies).  In recent months, owing to the fact that data caps and overage charges are typically based on gigabytes of data, the cost issue has crystallized around the question of the cost to transfer 1 GB of data.  

Reliable cost information would be extremely helpful in order to respond to at least two issues.  First, the claims regarding light users subsidizing heavy users would be a more informed discussion, since it would allow for a realistic assessment of the actual costs of servicing both light and heavy Internet users.  Second, reliable cost information would allow for analysis of the reasonableness of current overage charges. While retail Internet access pricing is unregulated, efforts to analogize Internet access to regulated utilities raises the specter of assessing the reasonableness of the markup for Internet access services.

i.    Internet facing data costs

Calculating the cost for an ISP to transfer 1 GB of data involves at least two cost components. The first component is the Internet side of the equation. The user’s traffic travels through three stages – the last mile, the ISPs internal network, and the public Internet. The public Internet cost is relatively easy to calculate since all ISPs depend upon peering and transit arrangements to carry data between networks.  Independent ISPs advise that the cost of a 1 Mbps (megabit per second) connection is roughly $5 per month (the cost is likely considerably less for large incumbent ISPs given their economies of scale as some estimates put the price at $3 per Mbps). 

With a 1 Mbps connection, an ISP is permitted to transfer 1 Mbps of data per second on a continuous basis. That is the equivalent of 7.5 MB (megabytes) per minute (60 megabits is equal to 7.5 megabytes), 450 MB per hour, and 10,800 MB per day (or approximately 10.5 GB).  Over a 30 day month, it adds up to 316.4 GB or roughly 1.58 cents per GB.  For larger ISPs able to negotiate lower costs, the $3 per month fee allows them to transfer a GB of data for under one cent.

Estimates of transit arrangement pricing typically point to between $3 and $5 per megabit per second per month. This figure is declining rapidly, with some estimates suggesting that it may drop to under $1 per megabit by 2014.  Transit arrangements are used when two ISPs do not have a direct connection and therefore rely on an intermediary network provider to complete the Internet connection. Some ISPs may directly connect with one another (known as a peering arrangement) that would further reduce costs since the data transfer travels directly from one ISPs network to another ISPs network with the ISPs agreeing to offset each other’s traffic. Yet another alternative for web-based traffic and video streaming is the use of Content Distribution Networks, such as Akamai. These companies are often used by content distributors to manage traffic more efficiently and provide ISPs with the advantage that the traffic may not run on their networks, thereby further reducing costs.

These numbers are consistent with other public numbers for large-scale data transfers.  For example, the Amazon EC2 service, a leading cloud computing provider, charges one cent per GB for all regional data transfers.  Regions include the U.S. East Coast, U.S. West Coast, and Europe.  For data transfers outside a single region, the Amazon EC2 service costs ten cents per GB for all in-bound data transfers in North America and Europe and as little as 8 cents per GB for out-bound data transfers.

ii.    Internal Network Costs

The far more difficult cost calculation involves the internal ISP network, namely the last mile and the internal network. ISPs zealously guard this information, demanding that virtually any potential disclosure remain confidential.  While the ISPs often disclose this information to the CRTC, most of the public record of CRTC proceedings has been redacted of information that would be useful to estimate the costs of increased network traffic on confidentiality grounds. 

There is one notable exception, however. As discussed at Appendix B of the paper, unredacted information about total costs and total subscriber numbers for Bell’s proposals for funding broadband expansion to underserved communities from the deferral account is available. Three separate sets of data were disclosed in 2010, representing different designs for the access network. Bell’s preferred option is based on High Speed Packet Access (HSPA) technology.  Other options investigated included a pure DSL technology option and a hybrid option, combining both technologies. 

It should be noted that Bell’s data describes a “worst case scenario” in the sense that Bell was making the case that such a network would not be economically viable and would therefore merit additional support through the deferral account process.  The Bell data disclosed the total cost of the project (over 15 years), the number of communities and premises where service would be available, the projected revenue, and the peak number of subscribers.

The detailed analysis of the Bell data, which admittedly involves a number of assumptions, suggests that the costs of building a new network would lead to estimates of roughly seven cents per GB.  The actual costs vary considerably between networks – capital expenditures, network usage, subscribers, and properties of the backhaul links all influence actual costs.  While the actual numbers may be higher or lower, the seven cents/GB estimate is fairly consistent in several network models. 

When combined with the Internet costs of roughly one cent per GB for larger ISPs, a high end estimate of the per gigabyte costs for large Canadian ISPs is approximately 8 cents per GB.   This assumes the creation of a new network and accounts for all aspects of the data transfer – the last mile, internal ISP network, public Internet transfers, and other associated expenditures. While this is higher than the 3 cents per GB that has been invoked in some discussions, it is far lower than overage costs imposed by some ISPs, which run as high as $10 per GB in Canada.


  1. Steven Scott says:

    Another Great Review
    Once again I have to applaud you for your detailed analysis of a common problem faced by all consumers. I do not believe that any rational person arguing against UBB suggests that internet service should be free, but the rates they are charging is crazy. Monopolies have allowed this, and while the big telecom/cable continues to control the infrastructure, average consumers will continue to get taken advantage of. It is a shame you are not on the CRTC or at least advising the government on these issues. Hopefully the next government is reading all of this and taking it to heart.

  2. Personally, for me, it’s not just the rates, but the stupidly low values that the big telecoms put on the lower limit of the UBB they chose. While I wouldn’t consider my wife and I heavy users, we use about two to three times Bell’s “most people only need this” 25GB cap.

  3. what competition?
    There’s only two things that keep a supplier from gouging his consumers: competition and regulation.

    Do we have (or will there ever be) competition in the wire-line Internet game? Nope and NOPE.

    The cost of running that last mile to everyone’s home is impossible to overcome given the 100 and 50 year monopolies on doing just that that the phone and cable have had. Those monopolies served two benefits: (a) it allowed those companies to go to that tremendous effort without having to deal with competition and (b) it funneled 100% of the consumer’s budget towards those companies to achieve it.

    Those conditions will *never*, *ever* exist again and the result is that there will never be a “third” cable plant running to everyone’s homes.

    So there is only one (well, two) option left and that’s to regulate the cost of and access to that cable plant so that it’s being used fairly to the benefit of the people who paid for it.

    The remaining option is radical. That option is for the government to declare that the public funding (i.e. you and I paying our bills to use them) that has gone into building those cable plants is being abused and to take the cable plants back, out of the hands of the profiteers, and to manage them as a non-profit, neutral access cable plant for (i.e. phone, television, Internet, etc.) providers to use “at cost”.

    As I said, that last option is radical. Perhaps it can be achieved through regulation, perhaps not.

    What is for sure is that there will never be another cable plant (and let’s not kid ourselves that wireless can and will replace the cable plants) and as long as the profiteers are in control of it, without competition or regulation, their customers will be victims of gouging.

  4. When browsing a Rogers blog, I saw a comment where someone asked rogers how much it cost to deliver 1 GB of data to a customer.

    Took a screen shot of it:

    I like how Rogers claims that they cannot tell us how much it costs due to competitive reasons, but what competition do they have? Rogers and Bell both charge about the same price for internet access, and both charge about the same price for overage fees, and they control 99% of the market!

  5. Bell calls “Uncle!”?
    Current Bell overage pricings is $5/40GB which is $0.125/GB for all Bell DSL and Fibe links. This corroborates that the real cost is in the range of $0.02 to $0.08/GB to allow for business overhead…

  6. @ Brian a “third” cable plant
    Would that “third” cable plant not be fiber? There are a fair number of places in the world that are actively installing fiber to the premise. Once you have a fiber network you no longer need either a phone or cable network. So this particular cable plant is not really the third. It is the last.

  7. Konrad Gigapipes says:

    The true costs of internet transit is confidential for most ISP’s to tell you due to their contracts. I know publically advertises $2/mbit, I can say some various big name providers are matching or beating this price if you buy multiple gigabit commits from them.

    The $5 quote you heard is grossly overpriced when dealing with multiple gigabit commits and the highly competitive environment of providers at 151 Front Street West. An unnamed website I’m involved with uses multiple gigabits and is paying under $2/mbit for a top tier provider, ie, not the Kraft Dinner providers such as Cogent or

  8. @ bwalzer: fiber? in Ontario? Canada even?
    Where is fibre in Ontario, or even Canada, being laid by a non-incumbent telecom provider?

    > There are a fair number of places in the world that are
    > actively installing fiber to the premise

    That’s completely and totally irrelevant to the issue at hand which is the lack of competition being the cause for gouging for Internet, here in _Canada_ and the reason for the lack of competition is the impossible task of third (independent) cable plant in Canada.

    Even a third cable plant is no panacea. Whatever provider provides that plant could just go the way WIND has gone in the mobile market and join the oligopoly (i.e. even with WIND, you cannot buy a data-only phone plan) to provide the same tired old services.

  9. Cost is irrelevant
    Without significant competition, cost is irrelevant because there is no drive for efficiency without competition. Internet in Canada is like pharmaceuticals: limited players and without caps on cost they players can make it cost as much as they need to to justify exorbitant pricing?

  10. I’ve said it before, and I’ll say it again.

    Prices are driven by what the customers are willing to pay, not what they want to pay. If internet access is so essential to you that you are willing to pay these rates, then you are encouraging them. It isn’t like we are talking about electricity or water or food here. What is the impact if you don’t have high speed internet service? I suspect for the vast majority of people their lives would be about the same; perhaps you can’t do as much online gaming, etc.

    If you think that they are gouging you, then examine your lifestyle and decide if you can do without. When enough people opt out because they aren’t willing to pay those prices for the service being provided, the carriers will drop their prices (the shareholders will insist), even without introducing new competition, just to draw back customers and therefore increase revenues.

    People will talk about telework, etc. You need to compare the extra costs of the ISP to what you save by having high speed access. For instance, if you would normally drive to work, you save the costs associated with travelling such as fuel, insurance, maintenance. If those saving are greater than the high speed costs, you are still on the plus side. For me, if I could telework I’d save about $300 per month in transportation costs. So, if high speed cost me an extra $100 per month for the overages, I’d still be $200 ahead each month. Would I like to keep more? Sure. But that $200 is better in my bank account than someone else’s.

  11. @Anon-K: “What is the impact if you don’t have high speed internet service?”

    How about *you* cancel your internet service for 1 year then come back here to tell us how it went?


  12. Overcharged
    Re: Anon-K:
    The internet is not a Luxury for the rich, it has traditionally been utilized by the poor, who cannot afford $100/mo for cable tv. I would at least appreciate it if the base data caps rose once in a while to match current average user consumption. Instead they try to charge far more for bandwidth with UBB. Unconscionable! That was the last straw, and the way to fight them is regulation, not… cancelling your internet. Due to lack of competitors, they have us cornered.

  13. Not quite right… at least for Internet Transit costs.
    Most ISP’s pay for bandwidth based on peak usage. Since your typical user doesn’t transfer 1Mbit/s of data continuously, using that as an assumption significantly reduces the cost per Gig.

    If a user uses that 1Mbit/s for only 6 hours, the cost to the ISP is still the same for having that 1Mbit/s of capacity for them (assuming during peak usage), but the cost per gig is now $0.06 per gig.

    The cost per gig for any given period should be the area under the average network utilization (gigs) over whatever has been paid for transit costs. It is in fact a fluid number.

  14. Netflix is the ‘B’ horror movie for the telecom industry …
    Slightly off topic but related … Shaw taking hard line w/Netflix: “B’casters offering service to Cdns must be Cdn owned/controlled and be licensed”.

    This is the problem with having the ISP and broadcasters being one and the same. With new delivery models the two pipelines of the internet/cable providers are competing with each other. Nothing scares the guys more than the Netflix model because it decreases their income form ‘cable cutters’ while driving up their costs on bandwidth delivery.

    There are two ways to deal with this threat, make it too expensive or kill it. The first is being attempted through UBB’s exorbitant overcharges. The second is by trying to tie it to Canadian ownership/content regulations, which is impossible in the global internet age.

    Finally all of this is to protect, among other things, the Shaw brothers $16,000 PER DAY! pensions 0_o … Your over-usage fees at work.

  15. NetFlix has a fair amount of Canadian content to be honest. More than I remembering seeing on Canadian TV when I had it (other than the CBC) and more than Rogers had in their Pay Per View.

  16. And the Netflix/Canadian Content note, if the CTRC regulates them out of Canada I’ll be very much annoyed.

  17. Estimate Seems Too Low
    While this estimate is a good start, there are already some obvious flaws. Your estimate for per/GB internet costs assumes that the internet bandwidth is 100% saturated, 100% of the time, which gives the lowest possible cost per GB. This is not a realistic scenario. In reality, an ISP is going choose a level of bandwidth that is saturated during peak usage, but would be overprovisioned the rest of the time.

    There’s another way of looking at this though. I personally use roughly 20GB per month on my Shaw connection (which has a 60GB cap). Your $0.08/GB cost would imply that Shaw’s total monthly costs are $1.60 for me. The price for my plan is $47/month, which implies that Shaw is making a profit margin of 97%. Gougers!

    However, when I look at Shaw’s 2010 financial statements, this estimate doesn’t make sense. With 1.8 million subscribers, this would imply that Shaw’s internet business has annual expenses of about $35 million, assuming I’m the typical household. According to Shawn’s financial statements, their entire operating expenses for 2010 were 1.958 billion (not including capital expenses). Are we really to believe that the cost of running Shaw’s internet business is 35 million and the cost of everything else is 1.923 billion? (over 50x as much). This doesn’t seem likely. Something tells me your analysis is missing something.

    I have no ax to grind here. I’d personally love cheap and plentiful internet access, but every time I read people ranting on message boards about how they’re getting screwed by the ISP’s, I get the impression that nobody has a clue what they’re talking about. Finding accurate information on the residential ISP’s business costs is difficult. This paper is an improvement, but the end conclusion doesn’t seem plausible.

  18. Or …
    @Chris “And the Netflix/Canadian Content note, if the CTRC regulates them out of Canada I’ll be very much annoyed.”

    … or you could subscribe to a VPN service and watch all the geo-blocked content you desire, complete with useless advertising from outside Canada and fees going to foreign business’s pockets.

    Great regulation and foresight … Way to go Canada!

  19. Once again …
    @SL “… every time I read people ranting on message boards about how they’re getting screwed by the ISP’s, I get the impression that nobody has a clue what they’re talking about. Finding accurate information on the residential ISP’s business costs is difficult.”

    Yes, I know this is still a fraction of Shaw’s 1.958 billion but it shows where their priorities may be a little out of whack.

  20. @SL
    Maybe if the big ISPs don’t want people guessing about how much is actually costs, they should release how much it actually costs. Personally, I don’t see a huge amount wrong with the given conclusion (that admittedly makes assumptions and he says so), but to prove that wrong there needs to be actual data to back it up. And the fact that smaller ISP tend to report cheaper prices doesn’t help their case that it costs more.

  21. Since Michael systematically avoids mentioning Sasktel:


  22. @Napalm: Sasktel
    Not sure where you are plugging that in here but in case it’s in rebuttal of my “third cable plant” it’s worth noting that that’s the incumbent telco, Sasktel that’s installing that.

    That does nothing for the competition problem.

  23. @Chris A.
    I’d be happy if the ISP’s would give us more information on their costs. In absence of that we have to go on the information we have. What I’m saying is that the estimate of $0.08/GB doesn’t match with other facts that we _do_ know (i.e. Shaw’s financial statements).

    If Shaw really was making 95+% profit margins in their ISP business, I would expect these obscene profits to be obvious from their financial statements. That is unless they’re losing money hand over fist on their TV business and are propping it up with their ISP business. That would imply they are charging too little for cable TV but too much for internet. Somehow that doesn’t seem likely.

    Internet prices in Canada are certainly very high, but Michael’s Geist’s numbers suggest I should be paying $2 per month. I’d love it if it were true, but this doesn’t seem plausible. Is there _any_ industrialized country in the world where you can get high speed internet service for $2/month? I suspect the real costs fall somewhere in between and that if we had healthy competition, prices would fall, but not to price-for-a-cup-of-coffee levels.

  24. To be sure, the executive pay at Shaw is like 3-4 times higher than the other highly paid execs at the other ISP if you check the public sites like and they give have amongst the highest dividends for any cable co. to share holder and they have highest profits per customer dollar spent according to recent news articles, However, this blog’s analysis does seem lacking as SL says. For example, tiny locations in the backwoods must have a much higher cost than high density urban areas like Toronto. It would make this a difficult problem to piece together from the bottom up. Therefore the costs should more reliably derived from their yearly expenses divided by the total number of customers and average usages.

    Shaw just announced layoffs too, the remaining employees must be really scratching their heads along with the consumers.

  25. working backwards
    Lots of discussion about Shaw’s “costs” here. Although we don’t know Shaw’s internet costs per GB, we can work backwards from their retail package prices to set some upper bounds.

    Shaw has 5 retail packages. Ideally they can be broken down into a “base cost”, which is the basic installation plus speed premium, and the “per GB” costs. As a first approximation, we can ignore the “base cost” and simply calculate a “per GB price” against the full price of the package (effectively assuming the base costs is zero).

    Lite – 1Mb/256Kb 15GB cap @ $37.00 = $2.40/GB
    High – 7.5Mb/512Kb 60GB cap @ $49.00 = $0.81/GB
    Extreme – 15Mb/1Mb 100GB cap @ $59.00 = $0.59/GB
    Warp – 50Mb/3Mb 175GB cap @ $107.00 = $0.61/GB
    Nitro – 100Mb/5Mb 350GB cap @ 160.00 = $0.45/GB

    So unless they are losing money on the Nitro package, we can safely assume their “all in” network “per GB costs” are less than $.45/GB, possibly much less.

    Trying to do some spreadsheet calculations and factoring some kind of “base cost plus speed premium” is very difficult. Not only are their 2 distinct types of installations across these packages (docsis3 and docsis3), their is some pretty obvious “sweet spot” packaging going on. If anybody is interested, it looks like the “Extreme” package is the best “bang for the buck” overall. But trying all kinds of combinations, I can reasonably assume that the “per GB” after profit pricing is somewhere between $.17/GB and $.33/GB.

  26. @Brian: “That does nothing for the competition problem. ”

    Except that we’re looking here at a company that has no caps or UBB, and is actively upgrading their network, without any competition.

    So what’s different here?

    (hint: Sasktel is a crown corporation)


  27. Ding .. Ding .. Ding .. Give the man a prize.
    So what’s different here? (hint: Sasktel is a crown corporation)

    When profits are not funneled to rich private pensions and shareholder dividends … then lo and behold the internet infrastructure is suddenly affordable.

    Internet access, no longer a luxury but necessity, should be a regulated service. Oh, but a radical extremist am I 😀

  28. Napalm, I’m not sure what special in that Sasktel link. All the big Canadian telcos and cable co’s announced FTTH about 1-2 years ago but other than a couple of exploratory projects none of them are expanding it in any significant way and neither is Sasktel from what I heard. Let’s not go crazy with Crown Corporations they are just plain wasteful and stifling – maybe a co-op like Access Communication is better but private companies should be able to work better. The Conservatives are the best ones to manage this as the Liberals and NDP will just go overboard like they always do. I don’t want iPod and MP3 taxes on top of everything else.

  29. Just say no …

    Give no advantage to the incumbents. They have failed to deliver cost and speed effective networks that our peers enjoy. Too much profits and not enough competition is the bane of the Canadian Telecom industry.

    I’m all for build out of faster internet to rural areas though, and agree that ANYONE bidding on new spectrum should be required to put that in their agreements.

    @Martin “Crown Corporations they are just plain wasteful and stifling” – Agreed. But rich corporations are no better. Your co-op suggestion sounds better.

  30. I can receive 3000 GB per month cable TV to my 3 televisions for $32/month for 48 channels (if all TV’s are on 24/7). Over the same cable lines, Internet data (identical in form and content) can travel to my home to be split through a cable modem to my computer for $45/month with a 26 GB cap and $1.50/GB for overage. The cable company techs admit that there is no difference in the data traveling down their lines to my household, except they can get away with billing me at different rates. They don’t like it either.

  31. Bruce, to the headend techs or the support techs they may not notice much difference since the headend tech only deal with layer 1 and 2 or the physical layer. But they are different.

    Essentially, TV is using broadcast and multicast while internet is mostly unicast and is treated differently and they have different costs involved.

    On another track, has Michael or anyone read Andrew Odlyzko’s recent paper on “Why Internet pricing is on ongoing tragi-comedy”?

  32. @Napalm
    I’ve mentioned here MANY times that high speed internet, while theoretically available at my house, in reality is not (it is wireless and the signal is unusable). I use dial-up at 26,400 bps. Costs me about $26 per month. The FASTEST I’ve ever seen on my phone line was once I saw 28,000 bps.

    My neighbour has wireless high speed (they are a little higher) but it is problematic during the summer due to the leaves on the trees interfering with the signal. During the winter it works. Oh, and by the way, she works from home for the most part.

    So no, I am proof that one does not require high speed internet at home. Nor am I contributing to the cause of your whining. Dialup is always an option.

  33. @SL
    That’s what the base costs would cover. How much it costs to send a GB really is immiterial to that. Dr. Geist’s posts addresses the overage charges that ISPs are charging, not their base costs. Using oldguy’s numbers, if the largest package allows you to send at $0.45/GB, why are they charging you $2-5 for going over?

  34. @Anon-K: “So no, I am proof that one does not require high speed internet at home.”

    I didn’t say “high speed”. Just basic internet access.

    If you look at it, internet has two aspects:

    1) communication tool for solving daily life items. Like online shopping, accessing info from government sites, ticket reservations, e-mail and such

    2) entertainment. Video streaming, gaming and so on.

    I agree that you can live without 2).

    However 1) is pretty much hard to avoid, just try to get an airline ticket reservation “in person” (no phone or internet allowed).

    Incidentally it is 2) that needs high speed high caps.

    So there’s an interesting effect to UBB: if you’re using Internet as a communication tool, you’re fine with the base plan. If you also use it as an entertainment source, you’ll go over your monthly cap and pay overage.

    So we can see two price plans here disguised under “UBB”.


  35. @Martin: “Crown Corporations they are just plain wasteful and stifling – maybe a co-op like Access Communication is better but private companies should be able to work better.”

    Yeah. We need some well managed, thrifty, friendly, working for the society’s progress companies. Like Enron, eh?


  36. Canadian Consumer says:

    Crown Corporations
    The main reason why crown corporations are mismanaged is because those who get the jobs are appointed based on their politics and the jobs given as kick backs.

    Kind of like the CRTC, look at them, no one knows what the hell they are doing.

  37. @Canadian:

    Large, public owned “private” corporations with gazillions of small stockholders are in no better position.

    But when it comes to fraud, the best you can hear from crown corporations is something like dancers paid with the corporate visa or kickbacks from vendors.

    Compare to Enron or the crop of banksters that recently drove US into the ground.


  38. @Napalm: Enron? is it still a company? uh, no.

    Name one good Crown Corp.

  39. @Martin: “Name one good Crown Corp”

    Heh. How about checking no more than Bell’s history:

    “The three prairie provinces, at separate times up to 1912, acquired Bell Canada operations and formed provincial utility services, investing to develop proper telephone services throughout those provinces; Bell Canada’s investment in the prairies had been scant or insufficient relative to growth. Having achieved a high level of development, Manitoba moved to privatize its telephone utility and Alberta privatized Alberta Government Telephones to create Telus in the 1990s. Saskatchewan continues to own SaskTel as a crown corporation. Edmonton was served by a city-owned utility that was sold to Telus of Alberta in 1995.”

    Why did they need to nationalize Bell???


  40. I said “good”, what made them any better than when they were privatized?