Overestimating mobile data growth

Cisco has recently updated their forecast of mobile data growth.  The good news is their forecast compound annual growth rate (CAGR) is down, from last year's forecast of 131% to their current forecast of 108% growth per year.

 
 

Unfortunately, when you look over the past 15 years of Internet traffic growth, the idea that any specific sector could exceed 100% growth for more than 1-2 years just isn't credible. 

In the early days of the Internet (prior to 1994) traffic did approximately double each year. And, as Andrew Odlyzko points out in his classic study, there was a period of perhaps 18 months in 1995-1996 when US traffic grew much more rapidly, however growth reverted to 100% per year in 1997.  And since then, Internet traffic growth has been slowing.

The most comprehensive data I am aware of is from Minnesota Internet Traffic Studies (MINTS) which shows the rate of Internet backbone traffic growth slowing so that in recent years (2008-2009) it has been in the range 40%-50%.  For a good discussion of long term trends see Internet Growth Trends & Moore's Law.

Mobile network traffic growth

It's interesting to look back on the growth of mobile networks so far.  Here are some numbers on cell sites, subscribers and voice minutes of use, extracted from CTIA data by Andrew Odlyzko.

                             1991              2006

subscribers            6.38 M          219.7 M

cell sites                6,685           197,576

MOUs                    5.2 B           858 B
 

Over 15 years, voice traffic grew 165x which is 41% CAGR.  

As an aside, it's interesting is to separate the growth in capacity due to more cell sites from the growth due to better wireless technology.  Over the 15 year period 1991-2006, we got a 30x growth in the number of cell sites, or 25% CAGR.  That suggests improvements in wireless technology (the transition from 1G AMPS to 2G and 3G systems) provided for only 5.58x of the increased traffic, or 12% CAGR.  This surprised me.  It can't be due to uncounted data traffic as data traffic was a small percentage of the mix in 2006. If true, it's rather embarrassing for the 2G and 3G equipment providers.  Any suggestions?

There's a different view in this study of available mobile data download speeds done by Novarum Inc.

This shows available download data rates doubling in a little less than 24 months, or ~45% CAGR. However, this is measuring available bandwidth, not actual traffic.  None-the-less, it's suggestive of capacity as these measurements were made on live networks during business hours.

A More Reasonable Projection

Given both Internet and wireless history, I suggest mobile Internet traffic growth over the next 5 years is more likely to be in the range 50% to 100% CAGR.  There may be a 12-18 month period when it grows more than 100% per year, but only for 12-18 months and only if some alternate technology enters the mix (for example, widespread use of Wi-Fi hotspots as I suggested a few weeks ago).

Comments

[Russ]

I'm afraid I've got to challenge your logic on this one Brough.

I think there are two problems with your approach to questioning mobile data growth.

The first is making the parallel to Internet traffic growth. If mobile data were ONLY Internet traffic, that might not be a bad approach. However, most observers believe that the number of wirelessly connected devices that aren't phones or computers is exploding, at much higher than 100% per year (e.g. http://www.cellular-news.com/story/41276.php ). Many of these devices are low-bandwidth, but with the advent of 4G services, increasingly high bandwidth devices will also be connected. This is all part of the "mobility revolution." I believe that, just as the microprocessor revolution has led to microprocessors being embedded in virtually any product that has a power supply, in the not-too-distant future (perhaps 10-15 years), virtually any product that has a microprocessor will also have wireless connectivity embedded. To limit your view of mobile data traffic growth to simply be Internet traffic is a mistake I'm surprised to see you make.

But then, the second flaw in your argument is in believing that, simply because Internet traffic growth has dropped to 40-50%, no single sector can grow significantly more than that. Mobile as a percent of total Internet traffic is tiny. There's huge growth opportunity simply from Internet use/traffic shifting from fixed to mobile. Using the MINTS data that you referenced, if we assume 2009 Internet traffic is 10EB (roughly in the middle of their range) and 40% annual growth (at the low end of their range), Internet traffic will grow to about 54EB by 2014. Simply plugging in Cisco's numbers, that means that mobile (even if it were all Internet traffic) would increase from about 1% of Internet traffic to a bit less than 7% of Internet traffic by 2014. That sounds very reasonable to me.

Another way to think about it is to ask the question - what is going to continue to drive Internet growth over the next 5 years? Some of it will be the increasing size of fixed pipes and the increasing adoption of applications like video. But you've got to assume increasing mobile Internet use, especially with increasingly fat pipes, will also be a key driver. Again, simply using the Cisco and MINTS numbers, mobile data growth (if it were all Internet) will only be 8% of that total Internet traffic growth. Again, a very reasonable assumption.

(I won't even get into the flaws in your logic of using mobile voice traffic growth as a proxy for potential growth in mobile data traffic. Where would using voice traffic growth over the previous 15 years to project fixed Internet traffic growth in 1995, or even 1999 have gotten you? Or the flaws in attributing the voice traffic growth on a 1:1 basis to the number of cell sites - don't you think the cell sites that were in place in 1991 were in the heaviest use places? And how many cell splits are included in the number of number of new cell sites? Oh, and I've always admired your technical knowledge, so maybe you can explain how new voice technology would translate into either an increase in subs or an increase in MOUs and how that correlates to new (faster, higher capacity) data technologies translating into increasing data traffic...)

Come on Brough, try not to be so dogmatically opposed to optimistic views of mobile traffic growth. :)

(As always, my comments do not necessarily reflect the opinions of my employer.)

[brough]

Russ, I am optimistic about mobile traffic growth and I agree that mobile data is starting from an extremely small base so growth percentages could be quite high for a while. If you look at slide 10 of my January talk at 4G Wireless Evolution ( http://su.pr/2J7Y2J ) you'll see me projecting high growth from a small base.

My concern is people expecting such high growth rates to go on for 5-6 years or more. There are many examples of specific markets segments with dramatic growth, from a small base, for a short time. For example, the US burst of >200% Internet traffic growth that lasted for ~18 months in 1995-96 and a comparable growth in Japanese traffic (caused by xDSL and FTTH deployment) that also lasted for perhaps 2 years, or the Hong Kong spurt that peaked in 2000. There are many others, but in each case, the duration was 12-24 months before returning to sub-100% CAGRs.

I think we're both optimistic. I just want to avoid a repeat of bubble thinking we saw in the US in the late 1990s when, as late as 2000, two different authors, in a Fortune magazine special (November 27, 2000), claimed Internet traffic doubles every 100 days. That is something that might have been true for a carefully selected 100 day period in 1996, but not for a 5-6 year period.

The other problem with mobile data traffic growth is 3G and 4G technology can't accommodate the demand. The reason I brought up growth in voice minutes was not because there is any relationship in voice and data applications, but because of what it shows for the growth in the underlying networks. During the period 1991-2006, voice traffic is a near perfect proxy for growth in mobile capacity. That growth comes from technology improvements plus more cell sites - both dependent upon capital investment. Once the needed technology exists, spurts of growth can come very rapidly, limited only by capital - think of the pace at which 2G mobile voice was rolled out in Pakistan ( http://tinyurl.com/yen98mo ). But there is no evidence that 3G or 4G can accommodate >100% mobile data traffic growth given the available capital and the available technology.

I believe the answer will be wholesale offload to Wi-Fi, using existing ( http://su.pr/2L14tQ ) and new Wi-Fi hotspots. I do expect an enormous surge in mobile data traffic - a surge that could easily go over 200% CAGR at least briefly, but only over 100% for 18-24 months, as capacity catches up with demand. And, I expect most of this growth to be on Wi-Fi, not 3G/4G, as Wi-Fi is the affordable way to accommodate the demand ( http://blogs.broughturner.com/2010/02/wifi-offload-not-femtocells.html ).

We’re both optimistic in our own ways. Let’s agree to check back in 2013-2014 for the details of what has actually happened. :)

[Russ]

Sounds good. Thanks Brough.

[Mobile phone tracker]

I use my mobile phone to surf the internet everyday, does that count?