Are there limits to growth in data trafﬁc?:
On time use, data generation and speed
Mike Hazas, Janine Morley, Oliver Bates, Adrian Friday
Bailrigg, LA1 4YW, United Kingdom
This discussion paper considers the nature of growth in data
traﬃc across the Internet, as a basis for asking whether and
how such growth might slow down or otherwise be limited.
Over the last decade, data growth has been dramatic, and
forecasts predict a similar ongoing pattern. Since this is
associated with increasing electricity consumption, such a
trend is signiﬁcant to global eﬀorts to reduce carbon emis-
sions. In this paper, we selectively explore aspects of data
growth that are linked to everyday practices and the way
they draw upon and generate Internet data. We suggest that
such growth does have some conceivable limits. However,
the nature of ‘Internet use’ is changing and forms of growth
are emerging that are more disconnected from human ac-
tivity and time-use. This suggests that although there may
well be limits, in principle, to some forms of growth, total
data traﬃc seems likely to continue growing. This calls for
careful attention to the nature of the trends involved, as a
basis for intentionally building limits into this system be-
fore levels of Internet electricity demand becomes directly
and more explicitly problematic.
•Social and professional topics →Sustainability; •Human-
centered computing →Interaction design;
Information infrastructures; social practice
In a recent article in Low-Tech Magazine, de Decker 
argues that “there are no limits to growth when it comes
to the Internet, except for the energy supply itself” and so,
uniquely, “the energy use of the Internet can only stop grow-
ing when energy sources run out, unless we impose self-
chosen limits”. Other energy-demanding systems, he sug-
gests, encounter inherent constraints which limit their growth:
Permission to make digital or hard copies of all or part of this work for personal or
classroom use is granted without fee provided that copies are not made or distributed
for proﬁt or commercial advantage and that copies bear this notice and the full cita-
tion on the ﬁrst page. Copyrights for components of this work owned by others than
ACM must be honored. Abstracting with credit is permitted. To copy otherwise, or re-
publish, to post on servers or to redistribute to lists, requires prior speciﬁc permission
and/or a fee. Request permissions from email@example.com.
LIMITS ’16, June 08-10, 2016, Irvine, CA, USA
2016 ACM. ISBN 978-1-4503-4260-5/16/06. . . $15.00
for example, the weight and size of cars can only increase
so much if they are still to remain compatible with infras-
tructures designed for them; infrastructures in which speed
limits are imposed for safety reasons. Similarly, he suggests
that some kind of ‘speed limit’ for the Internet is needed.
Putting to one side consideration of how such limits might
be achieved, it is intriguing to examine the claim that the
energy used by the Internet will continue to grow until the
availability of energy itself becomes problematic, that is, un-
less some other kind of checks or limits to growth are im-
posed ﬁrst. This is a rather radical, fascinating and, in so
far as it is plausible, troubling claim.
Currently, the direct energy used to power the Internet
and to produce, transport and dispose of its components
(embodied energy) are relatively small contributors to global
emissions in comparison to, say, global transport . Cur-
rent estimates suggest that operation of the Internet (power-
ing devices, networks and data centres) amounts to around
5% of global electricity use; yet this is growing faster (at 7%
per year) than total global electricity consumption (3% per
year) . In other words, the Internet is consuming an in-
creasing portion of global electricity supply. In the context
of changing energy systems that include greater renewable
sources, and new forms of electricity demand such as electric
vehicles, the growing portion of global electricity required
to run the Internet may become increasingly signiﬁcant in
eﬀorts to balance supply and demand, reduce carbon emis-
sions and, as such, become potentially subject to energy-
related limits as de Decker suggests. Yet given the radically
distributed, and largely ‘invisible’ nature of this energy con-
sumption, how large a portion of global electricity could this
represent before such limits might be imposed? Some pre-
dictions suggest that production and use of information and
communication technologies might grow to around 20% of
global supply by 2030, or as much as 50% in a worst case
In this short discussion paper, we consider trajectories of
growth in the energy that powers the Internet, not through
modelling, but through a more conceptual exploration of
some of the trends currently and historically associated with
such growth. Speciﬁcally, we ask whether or not these trends
may ultimately be limited in themselves, and thereby may
help to slow and limit growth in the future, at least in prin-
ciple. This is a complex question, so we frame and focus our
exploration in two ways.
First, we focus on volumes data traﬃc as an indicator of
operational energy use. This is by no means a direct or
simple relationship but in broad terms “the very substantial
gains from energy eﬃciency improvements have been more
than oﬀset by increased consumption of services”[12, p. 583].
Thus, as ﬂows of data traﬃc have increased so too has energy
use, albeit at a lesser rate.
Second, we focus on the link between the use and gener-
ation of data and activities in which people are more and
less directly involved. We explore the idea that, to date,
human time and attention has been related to the growth
in data traﬃc, in various ways. We ask whether this might
represent some limits to ongoing growth, and whether other
forms of growth are emerging and taking over.
We start by characterising the rate and composition of
data growth. We then consider aspects of this growth re-
lated to activities and time spent ‘online’, and ask whether
such trends will continue indeﬁnitely or, at some point, rep-
resent some kind of check to growing data traﬃc. Whilst
growing levels of access to the Internet, a greater range of
‘online’ activities and services migrating and emerging, and
increases in time spent online continue to be relevant to the
growth in data, there is, in principle, a limit to these forms
of growth: the global population is ﬁnite (though clearly
growing), and the hours in the day available to each person
are also ﬁnite. We then turn to consider how the dynamics
of data growth are changing: the relationship to attention
and time-use is becoming less direct as data intensities of
certain services increase and as ‘background’ connectivity
and connected things become more prevalent.
2. GROWING DATA, GROWING ENERGY
Measures of actual Internet traﬃc volume and compo-
sition tend to be partial, limited to certain countries or
narrow periods of time. Based on a “representative cross-
section” of service providers, who volunteer to allow their
aggregate data to be reported, Sandvine  report that
the average monthly per-subscriber traﬃc volume on ﬁxed
lines like broadband and ﬁbre increased by about 50% in
North America, and 170% in Europe over 2013 and 2014
(Figure 1). According to Ofcom, the UK telecommunica-
tions regulator, UK home broadband data volumes also grew
markedly: monthly average traﬃc volume rose from 17 GB
in 2011, to 82 GB in 2015 .
Taking just one survey point, public aggregate input/output
data statistics published for January each year 2002–2016
from the Amsterdam Internet Exchange (Figure 2), we see
a clear pattern of year on year growth of between 20–140%
(20–40% each year in the last 5 years). Crudely, this data
demand has an equivalent direct energy cost as data is trans-
mitted and processed. Despite step changes in energy eﬃ-
ciency as new technology is introduced, this could arguably
be oﬀset by innovations in the marketplace such as increased
expectations around high-deﬁnition video (e.g. 1080p cam-
eras on smartphones, 3DTV and 4K video). Holistically, this
pattern of growth also causes increases in embodied energy
(emergy ) of the Internet as equipment needs to be re-
placed to cope with the continued increase in data demand,
and its associated demands for processing and storage.
For mobile access, data volumes are generally smaller, but
have consistently fast growth, doubling every few years ac-
cording to Sandvine (Figure 1). In fact, Ericsson  report
a 60% growth in global mobile cellular data in recent years,
and the forecasts are for 45% or more annual growth through
Figure 1: Growth in ﬁxed and mobile traﬃc vol-
umes. Data source: Sandvine reports 2012–2014.
2021 [4, 6].
Several forms of growth are reﬂected in the overall in-
creases in mobile traﬃc volume: a) per-subscriber demand,
which is b) compounded by the increasing number of tablet
and smartphone handsets connected to mobile cellular data
services, which is c) related to, but not entirely explained
by, the proportion of the global population who have some
form of Internet access.
On the face of it, the latter form of growth has some lim-
its, even though these may be many years away: broad-
band services currently only reach around 30% of homes in
most developed countries, and only 43% of the world popu-
lation is using the Internet . There are biophysical limits
to human perception and attention, which ostensibly could
place an upper bound on the ﬁdelity of digital media we
deem ‘suﬃcient’. Some limits are indeed already visible: in
economically developed nations, mobile phones are reaching
saturation. For example, the number of mobile subscrip-
tions already exceeds the population in the US, Scandinavia
and Australia. And while by 2020 over 3.1 billion mobile
subscribers will have access to LTE and 4G communications
networks worldwide, this falls far short of world population.
3. CURRENT DATA GROWTH: DEVICES
PRACTICES AND TIME-USE
Alongside questions of population and ‘saturation’, it is
important to consider what all this Internet traﬃc is for -
what activities it seems to be intertwined with. As Røpke
points out, “People are practitioners who indirectly, through
the performance of various practices, draw on resources” .
Figure 2: Monthly data from Amsterdam Internet
Exchange aggregated for each January 2002-2016.
Communications enabled devices increasingly augment ev-
eryday practices that previously would have not have re-
quired or even beneﬁted from, Internet connectivity. In our
recent study of the communication impacts of applications
of mobile devices in everyday life , we found practices such
as exercise regimes to be augmented by social media appli-
cations; and migration of once oﬄine or broadcast viewing
and listening of TV and radio to media rich communication
and streaming media platforms.
The personal ownership of these devices, coupled with the
convenience of accessing the Internet and their mobility, en-
able communication to ﬁll and even expand small pockets of
‘dead time’. Along with increased multitasking supported
by digital technologies “enabled by the partial decoupling
of many practices from previous time and space constraints
through the use of ICT, contribute to a more densely packed
everyday life” [16, p. 356].
In other words, because Internet connectivity is being in-
tegrated into a huge variety of practices which may take
place throughout the day, time ‘spent online’ attending to
digital services is growing. Indeed, in 2005 in the UK an
average of 9.9 hours was spent online in a “typical week”
across home, work and elsewhere. By the end of 2014 this
average had grown to 20.5 hours per week [10, p. 28].
Yet, not all ‘online’ activities (those that require an In-
ternet connection) are equal. Firstly, signiﬁcant diﬀerences
emerge between traﬃc types on domestic and mobile Inter-
net networks. Drawing on Sandvine, the category of “real-
time entertainment”, which is primarily video, accounts for a
large portion of total Internet traﬃc, at over a third of peak
period traﬃc. This is followed by web browsing, which might
support a wide variety of activities. Social networking is also
signiﬁcant, particularly on mobile networks. Secondly, these
activities ‘take-up’ time in diﬀerent ways, with some clearly
taking more time than others.
Some of the growth in time which is reportedly spent on-
line can be attributed to a daily increase in the number of
Internet users who watch TV or ﬁlms online (10%–27% be-
tween 2007–2014) . The number of Internet users who
watch short video content (i.e. video clips) has almost dou-
bled in the same amount of time, growing from 21% to 39%
Figure 3: Composition of peak period traﬃc in Eu-
rope in the second half of 2014. Redrawn from Sand-
Ericsson’s February 2016 report sums up why mobile video
is traﬃc is growing at such a rate, due to: larger screens en-
abling better quality streams; video content increasingly ap-
pearing as part of other applications (e.g. social networking,
news, advertising); growth of video streaming (50–70% of
video traﬃc for some mobile networks is YouTube); growth
in uptake of video on demand; changes in where and where
video is consumed and faster infrastructure.
3.2 Other online services
Whilst the number of people who are using online services
to consume media content is growing, it is worth noting
that the amount of time spent using these services is also
growing. In the US, the average time spent per day paying
attention to digital activities (including digital video, social
networks, digital radio, Facebook, Pandora) has increased
from 226–364 minutes per day between 2011–2015 . The
time spent online watching digital video has grown from 39–
115 minutes, with growth also observed for social networks
(71–104 minutes), and digital radio (53–65 minutes) .
3.3 Concurrent uses of the Internet
In 2014 additional connected devices are being used to
go online compared to 2009, including Smart TVs, e-book
readers and wearable devices, with all technologies (com-
puters, smartphones, tablets, games consoles) other than
“portable media players” increasing in their use for going
online [10, Fig. 34]. It’s worth mentioning that multitask-
ing that takes place whilst watching TV is performed by
53% of UK adults [14, p. xi]. This multitasking is en-
couraged by “living room connected devices” (e.g. devices in
the living room that are connected to the Internet) seen to
“blur the line between passive and active entertainment”[14,
p. 4]. Whilst we are currently unaware of the overlap be-
tween watching on living room connected devices and pri-
mary viewing devices (e.g. Smart TVs, laptops, consoles) it
is worth acknowledging that living room connected devices
are likely to increase time spent online as they are predicted
to increase from 114 million to 267 million units shipped
worldwide by 2017 [14, p. 5].
4. FUTURE DATA DEMAND: INTENSITY,
BACKGROUND AND NON-HUMANS
So far we have talked about ‘attention-connected data de-
mand’, that is, exchanges in data that occur as the (more
or less) direct result of what people do by paying attention
to online services, reﬂecting the presence and signiﬁcance
of media rich communication in everyday practices. As can
be appreciated from Figure 3, much of the data traﬃc dur-
ing peak times, currently seems to be associated with such
activities (e.g. entertainment, web browsing, social network-
ing). Although such traﬃc may continue to grow as a result
of increases in the data-intensity of these activities, such as
through higher deﬁnition video content or through new en-
tertainment services requiring complex forms of cloud com-
puting such as virtual reality and cloud gaming, there is
some limit to the time that is available to ‘invest’ in them,
and thus some ‘check’ to growth in these types of data de-
Yet not all traﬃc is so directly associated with attention-
linked services, and thus with patterns of time-use. One al-
ready automated form of demand on the Internet is through
software updates. These are currently about 6% of download
traﬃc, or perhaps up to 10% if computer game downloads
and updates are included (on marketplaces such as Steam,
the PlayStation Store, and Xbox Live) .
In addition, activities are at times accompanied by ‘unin-
tentional’ or ‘background’ data demand. In recent studies
of mobile device use, we found unexpectedly high levels of
communication between apps and the cloud when speciﬁc
applications were not in active use, for instance ensuring
that applications and operating system services were up to
date (900Mb or 5% of their overall traﬃc for one partici-
pant’s iPad), backup of application data and digital photos
to the cloud (2.25Gb/week or 71% of data demand cloud
syncing of photos and videos in one case). These levels
of data traﬃc, were both unobserved, uncoordinated, and
largely unmanaged by or even unmanageable for most of our
participants [3, ch. 6], . While these studies are arguably
with small and isolated populations, the very unremarkable
nature of these ﬁndings coupled with the vast numbers of
similar devices in common use, suggests a very signiﬁcant
data demand that is diﬃcult to manage and limit at scale.
As the dominant paradigm for software development, tool
chains, and indeed the very business models that drive the
mobile eco-system push the design of software as thin clients
to powerful cloud backend services, we will doubtless witness
these kinds of data patterns repeated many-fold. Most sig-
niﬁcant perhaps is the ongoing development of machine to
machine communication enabled by the so called ‘Internet
of Things’ (IoT). What about the data and energy impacts
of smart ‘things’, in homes, workplaces and civic infrastruc-
tures? This introduces another form of growth that is more
dissociated from the limits associated growth in direct forms
of Internet ‘use’ that have to date been so signiﬁcant. Con-
necting ‘things’ or ‘machines’ to the Internet changes the
connection between what people do and exchanges of data.
This communication will occur transparently, without obser-
vation or interaction, and potentially without limit. At the
time of writing, the existing 6.4bn connected IoT devices is
only slightly less than world population (86%), but market
predictions suggest this will reach 21bn devices by 2020—
roughly three times world population estimates. Some pre-
dictions put machine to machine communication as 45% of
the whole Internet traﬃc by 2022 , with additional re-
liance on data upload  and processing in the cloud .
While IoT devices are presumably assumed to be low
communications footprint devices such as smart meters, or
smart home thermostats that communicate sporadic values
to backend services; others like self-driving cars, and re-
motely monitored wireless cameras and wearable medical
devices will be highly data intensive, and require match-
ing facilities in communications networks and data centres
to oﬀer timely and responsive communication, computation
and storage. Some see IoT as a potentially ‘net negative’
contributor to energy demand due to the eﬃciency gains
made through better measurement and control. We merely
observe, that these new facilities have the potential to be
adopted into everyday practice, to raise expectations and
lower the burden of creating additional data and electricity
demand and have an initial and ongoing embodied energy
cost, by way of rebound eﬀect—in exactly the same way
that smart devices, media and broadcast are already doing.
What seems clear, is that the impact of IoT is unknown or
unknowable at this point. Algorithms that create/consume
data can carry on doing so, regardless of who is paying at-
tention to them, and even less clear are the limits and brak-
ing functions that will ensure that such systems will operate
within sustainable limits.
This returns us to the question of whether there is an in-
herent limit to Internet traﬃc growth (other than energy)?
On the one hand, yes. Some of the reasons for current and
past growth will themselves be limited at some point in the
future; there is after all a ﬁnite, if growing, population on
the planet and only a certain number of devices that one
might explicitly use to access Internet services. Even if we
assume that the current access patterns in the most ‘con-
nected’ societies are replicated worldwide; the full panoply
of second screen, higher deﬁnition, and on-demand over 4G
and higher, there is plausibly a limit to the hours in the
day and content that can be actively engaged with. On the
other hand, ‘the data under the hood’ is growing unwatched
and unabated. The automated updates, cloud syncing, of-
ﬂoading of storage and computation to the cloud, that are
an increasing feature of the design of applications, and en-
demic to the tools and pervading technological culture that
is bringing these about.
Further, the Internet of Things is set to trigger a whirl-
wind of investment and connected infrastructure growth that
has the massive potential to grow operational electricity use
and emergy of the Internet. Despite sometimes question-
able beneﬁts and motivations, the IoT is currently under
construction, in many diﬀerent ways. This raises key ques-
tions as to whether, and what kinds, of limits there may
be to potentially self-generating cycles of data generation,
processing and circulation within such an Internet. If such
cycles are largely automated and operate at remove from
the time-limits associated with human activity then, poten-
tially, any ‘inherent’ limits to the growth in Internet traﬃc
will fall a long way short in the future.
Yet this Internet of Things is still in-the-making, and such
limits (or lack of them) are not yet ‘inherent’. Thus, as the
Internet continues to develop, de Decker’s proposal of some
kind of speed limit, that might be built into this system,
shaping its development for years to come, is an important
proposition to consider; especially in comparison to an alter-
native prospective of making retrospective reductions in In-
ternet traﬃc in the future. It is far from clear how such lim-
its could and should be formulated and enforced: should ‘un-
limited’ data tariﬀs be replaced with volume quotas or dif-
ferential pricing for services of various ‘importance’ ? Should
micro-payment schemes incentivise use of more bandwidth
frugal and oﬄine media?
It is clear, however, that the dynamics of Internet traﬃc
growth are changing. In many ways the technical capac-
ities of data infrastructures, such as broadband networks,
continue to limit to the data ﬂows and services that they
also make possible. Speciﬁcally, many broadband and mo-
bile contracts are limited and traﬃc is managed. In many
parts of the world, including areas in developed countries,
these infrastructural capacities are increasingly experienced
as tangible limitations to accessing a range of services that
are taken for granted elsewhere. To the extent that such lim-
its are experienced directly as limits to activity, then there
are important issues of social equity that cannot be ignored.
But where diﬀerences in data demand are less ‘visible’ and
less connected to explicit forms of ‘using the Internet’ other
possibilities for limiting data intensities might be explored
in ways that do not impact and consolidate existing inequal-
ities of access.
Our community is well placed to help shape these debates
and possible futures. By measuring and understanding the
holistic value and impacts of such systems, and understand-
ing the varied ways in which they change, can we help to
bring about a future where these systems operate within
similar and suﬃcient limits?
We would like to give our thanks to our colleague Kelly
Widdicks, for her compilation and analysis of the Sandvine
reports from 2012 to 2015.
 A. Al-Fuqaha, M. Guizani, M. Mohammadi,
M. Aledhari, and M. Ayyash. Internet of things: A
survey on enabling technologies, protocols, and
applications. Communications Surveys & Tutorials,
IEEE, 17(4):2347–2376, 2015.
 A. S. Andrae and T. Edler. On global electricity usage
of communication technology: trends to 2030.
Challenges, 6(1):117–157, 2015.
 O. E. G. Bates. The impacts of domestic media and
ICT: a study of digital technology, energy
consumption, data demand and everyday practice.
PhD thesis, 2015.
 Cisco. Cisco visual networking index: Forecast and
methodology, 2012–2017. Technical report, May 29,
 K. de Decker. Why we need a speed limit for the
internet. Technical report, Low-tech Magazine, 2015.
 Ericsson. Ericsson mobility report: On the pulse of the
networked society (mwc edition). Technical report,
Ericsson, Feb 2015.
 J. Gantz and D. Reinsel. The digital universe in 2020:
Big data, bigger digital shadows, and biggest growth
in the far east. IDC iView: IDC Analyze the future,
 ITU. The world in 2015. Technical report,
International Telecommunications Union, 2015.
 C. Lord, M. Hazas, A. K. Clear, O. Bates, J. Morley,
and A. Friday. Demand in my pocket: mobile devices
and the data connectivity marshalled in support of
everyday practice. In Proc. CHI, 2015.
 Ofcom. Adults’ media use and attitudes report.
Technical report, Ofcom, 2014.
 Ofcom. Communications market report. Technical
report, Ofcom, 2014.
 C. Preist and P. Shabajee. Energy use in the media
cloud: Behaviour change, or technoﬁx? In Cloud
Computing Technology and Science (CloudCom), 2010
IEEE Second International Conference on, pages
581–586. IEEE, 2010.
 B. Raghavan and J. Ma. The energy and emergy of
the internet. In Proceedings of the 10th ACM
Workshop on Hot Topics in Networks, HotNets-X,
pages 9:1–9:6, New York, NY, USA, 2011. ACM.
 N. Robinson, J. Freeman, J. Gaspers, V. Horvath,
T. Hellgren, and A. Hull. Living room connected
 I. Røpke. Theories of practice —new inspiration for
ecological economic studies on consumption. Ecological
Economics, 68(10):2490–2497, 8 2009.
 I. Røpke and T. H. Christensen. Energy impacts of
ICT – insights from an everyday life perspective.
Telematics and Informatics, 29(4), 2012.
 Sandvine. Global internet phenomenon report.
Technical report, Sandvine Incorporated ULC, 1h
 M. Z. Shaﬁq, L. Ji, A. X. Liu, J. Pang, and J. Wang.
A ﬁrst look at cellular machine-to-machine traﬃc:
large scale measurement and characterization. ACM
SIGMETRICS Performance Evaluation Review,
 W. Van Heddeghem, S. Lambert, B. Lannoo, D. Colle,
M. Pickavet, and P. Demeester. Trends in worldwide
ict electricity consumption from 2007 to 2012.
Computer Communications, 50:64–76, 2014.
 A. Walgrove. The explosive growth of online video, in
5 charts. Technical report, 2015.