Less than $0.50/broadband customer/month in New York or London. Telegeography's respected survey reveals a price of $3.50/megabit in New York and  $3.13 in London, both down over 50% in the last year. That's in minimum quantities (1 gigabit, enough to support 10,000 broadband customers with minimal congestion.) I've reported from market makers like Hunter Newby they charge less than $1/megabit in the large quantities that would support a substantial deployment and Telegeography confirms that. They note that the U.S. price has been going down 26% compounded annually for the last 5 years. 

    Router and switch prices have been going down at a Moore's Law pace of 25-40% per annum, driving down the cost of transiting bits at a similar rate. Transit in major Western cities remains competitive, so the reduced costs are passed on to the broadband carriers. The result: Internet traffic has been expanding at what would seem ferocious rates, but the carrier's net cost has been generally flat to down.  

    Note that the bits need to be hauled from the Internet connection point to the local exchange. For carriers large enough to control their own fiber, that adds little to the cost and yields a total cost of less than $1/month/customer to carry the actual data. It's less than the typical marketing, support, or local loop cost, none of which are effected by how much data or speed the customer uses. There is only a trivial difference between the cost of a 3 meg or a gigabit service, or raising caps from 50 gig to 500 gig. In practice, the actual use goes up only modestly, perhaps by 30% in the typical upgraded connection. With bandwidth costs this low, we're talking dimes or at most a couple of dollars per month to handle any likely traffic flow.

   With Cisco seeing traffic growth falling towards 22% (U.S.) and 30% (worldwide, higher because of a greater increase in homes served), that means that the primary cost of Internet traffic per customer is falling. That growth is even smaller when adjusted for the continuing growth in subscribers. Kreifeldt predicts "global IP transit prices will continue to fall."

    Prices in Hong Kong are closer to $18/megabit and San Paulo almost $30. That's partly because of the costs to an Internet hub like NY or London but probably more because of limited local competition. That's still not enough to be more than 10% of the price of a broadband connection, especially because in Brazil a few large carriers getting much better prices dominate the market. (Telefonica, Telmex). Small carriers and rural ones, especially in the U.S., tend to be charged much more. 

   Erik Kreifeldt of Telegeography adds, 'Prices in central and much of eastern Europe resemble those of western Europe. You need to get out to Moscow, Istanbul, and Athens to see a categorical change in price. The median monthly lease price for a full GigE port in Moscow is $4.80 per Mbps, the lowest of these three cities. Much of the pricing is dependent on underlying transport back to the major IXs in Europe. So, once you're on that primary pan-European network footprint, the pricing is quite similar. Off that footprint, pricing reflects competition and costs to get access to that grid. This is also true of the global market in general. The Hong Kong price of $16 and São Paulo price of $27 are pretty good representations of the next two pricing bands. Next level might be Mumbai, with a median GigE monthly lease price of $38 per Mbps. Sydney should be a bit higher, but roughly in the same price category as Mumbai.'' I have confirming datapoints for Australia and Eastern Europe. Telegeography is generally right on target.

    Big carriers like Telstra in Australia or Reliance in India pay nothing like those high rates because they are large enough to own a share of the undersea fiber. Commsday just reported Telecom New Zealand books tens of millions of profit on their international fiber and their effective cost for bandwidth is much lower than competitors. This is crucial, because it gives the big incumbents a huge edge. As Shara Evans notes, the smaller Australian carriers are paying much more. The worldwide pattern is for all but the top three to five in a market to fade away. Smaller ISPs, especially in Britain, France and the U.S., have been mostly driven from the market because they can't capture these economies of scale. 2 carriers in most of the U.S. (local telco and cableco) typically have 90%+and get better every year at signalling each other to raise prices. New York, where Verizon's minimum for 1 megabit is $52, has some of the most expensive Internet in the world. In France, England, and Korea, 3-5 carriers combine for that 90+% coverage, and prices are often 20-50% less than the U.S. 

   Little guys have been falling by droves and competition reduced in nearly every market on Earth.

  I.D. Scales drew some interesting implications of this data at TelecomTV.

Here's the release.

IP transit price declines steepen

International