Does streaming really have a dirty secret?

BBC Three recently streamed a piece on how we access video content on demand, emphasising growth in data centre energy consumption and heavy reliance on fossil fuels. 

It is certainly true that we need to become more responsible digital citizens, and the programme had a valid purpose in helping people understand how the internet works:   Not everyone knows that our digital society is enabled by physical infrastructure that has an energy and carbon impact beyond our own devices.  Moreover, as we have frequently stated publicly (most recently in our Sector Energy Routemap, pictured) the freemium and advertorial business models that are so popular at the moment send no signal to the consumer of the energy cost of their online activity.

While we agree that it is important to keep a close eye on any potential negative consequences of the growth of digital services, this must be based on rigorous analysis, sound policy and informed citizens. It is therefore disappointing that a number of the claims made, or impressions given by the programme were misleading, or simply incorrect. Here we seek to correct them. We would have liked to see reference to peer-reviewed reports and industry data instead of alarmist claims and analogies that don’t bear too much close scrutiny.   We also wondered about the choice of a streamed medium – BBC iPlayer - for a programme about the negative impacts of, er, streaming.  And we wondered why the producers felt it necessary to fly a film crew to the USA when we have a world-class data centre sector here in London, which they are welcome to visit whenever they like.

We took particular issue with the following, which were stated or implied in the documentary:

 “There are hundreds of thousands of data centres like the large operations pictured.”  Yes, there are thousands of data centres, perhaps tens of thousands but to reach a total in the hundreds of thousands you would have to include things that are not like the data centres pictured; not at all.  You would have to include things like server rooms, closets and comms rooms, which are not data centres.  In the UK, a highly digital economy, there are between 400 and 450 recognisable data centre facilities, generally much smaller than those featured. 

“Data centres do nothing except allow us to upload videos of cats.”  Data centres underpin our economy. They transmit, receive, process, store and manage digital data and by doing so they support every conceivable part of our lives:  business processes, Government services, telecommunications, transport infrastructures and social networks. Data centres underpin an incredible range of activities across society and are now part of our critical national infrastructure. Our UK sector is a real success story, is globally important and provides the technical infrastructure for financial services, aerospace, transport, healthcare, retail and utilities.  Each new data centre contributes between £397m and £436m GVA per year to the UK economy[1] See our ten myths about data centres: https://www.techuk.org/insights/news/item/15255-ten-myths-about-data-centres

“Data centre energy is spiralling out of control.”  There is plenty of evidence to the contrary, most recently from Lawrence Berkeley Labs in the February edition of Science: https://science.sciencemag.org/content/367/6481/984

At a UK level our energy data from the climate change agreement shows an incremental increase in energy consumption from 2.57 TWh to 2.89TWh between 2016 and 2018.  Infrastructure efficiency in colocation sites has improved by 16% since 2014 (colocation providers only control the infrastructure, not the IT hardware). At the same time improvements in hardware, in software and in utilisation have massively increased productivity.  The energy needed to process a given amount of data has reduced by around 7 orders of magnitude over the last three decades.

“The ICT sector emits the same CO2 as the airline industry.”  This dates from 2007 when Gartner wanted to make the point that ICT was a significant energy consumer.  The emissions from burning fuel across the aviation industry were reported to be about 800Mt CO2 in 2015. Air travel, excluding air transport and military flying, forms about 80 percent of the total aviation sector footprint. So, in an unbalanced way, the ICT sector could be compared to the aviation sector (fuel only) – a similarity in terms of number. A more well balanced comparison of the two sectors’ overall carbon footprints could be carried out if figures for the manufacturing of airplanes, the operation of airports including ground vehicles, end-of-life-treatment and an accurate estimate of any other greenhouse gases (GHG) produced by the sector, including the high-altitude effects, were readily available. It’s also important to recognise that at the moment switching to renewables will cut the carbon footprint of the ICT sector by 80%. The aerospace industry has a much trickier path to decarbonisation.

“Streaming a video equates to 40g of carbon” and “watching this documentary is the same as driving 235m in a car.”  Both of these are wrong. The reasoning was influenced by a report from the Shift Project, which claimed that streaming was responsible for over 300m tonnes of CO2 in 2018. Perhaps because of the absence of other figures, this report was unduly influential, spawning a rash of similar coverage.  A review of the figures by an International Energy Agency analyst concludes that they are exaggerated between 25- and 53-fold.  This Carbon Brief provides the detailed analysis:

https://www.carbonbrief.org/factcheck-what-is-the-carbon-footprint-of-streaming-video-on-netflix

“Streaming of Despacito uses the same energy as the annual consumption of five African countries.”  This is incorrect, just like a previous claim about Gangnam Style and the energy consumption of Burundi.  Ericsson has demolished the former: “The Despacito example.. “Five billion downloads and streams clocked up by the song, Despacito (released in 2017), consumed as much electricity as Chad, Guinea‑Bissau, Somalia, Sierra Leone and the Central African Republic put together in a single year.” This was a statement made in several media articles around spring 2018.  The electricity consumption of the above countries was about 1TWh in 2017 (1TWh = 1 billion kWh). In comparison, let us look at YouTube, which contributes towards Google’s overall electricity consumption, being about 7.6TWh in 2017. It cannot be true that one song, even if streamed billions of times, consumes as much as one-seventh of Google’s electricity usage.  More accurately, 5 billion downloads of this song to a smartphone requires about 0.005TWh (a factor 200 less) including its share of networks and data centers. Typically, a download of one song requires 0.001kWh.”

“Growth of data results in a parallel growth in data centre energy consumption.”  Wrong: this is an old chestnut.  Multiple predictions of this type have been discredited simply by time, because they did not come true, but that hasn’t stopped fresh predictions emerging at regular intervals.  Simplistically, Moore’s Law, which means that the energy needed to process a given amount of data halves about every 18 months, together with technological developments such as virtualisation and infrastructure improvements, have enabled the sector to meet an explosion in demand for digital data with only an incremental increase in energy.  There are multiple evidence-based, peer-reviewed studies to back this up, the latest of which is from Lawrence Berkeley: https://science.sciencemag.org/content/367/6481/984

“Data centres are all buying dirty power.”  The ICT sector is world-leading in terms of its use of renewables and has recently committed to net zero emissions via a science based, sector specific decarbonisation pathway:

https://www.gsma.com/newsroom/press-release/ict-industry-agrees-landmark-science-based-pathway-to-reach-net-zero-emissions/

Within the data centre sector, 76.5% of the electricity purchased by UK commercial operators is already 100% certified renewable, and a further 10% is purchased according to customer requirement which increasingly means low carbon, taking that total up.  Google is the world’s largest purchaser of renewable power, meeting its requirements through power purchase agreements; direct contracts with renewable generators that create additional utility scale renewable capacity.   Operators are increasingly adopting this form of power procurement.  This is spelt out in more detail in our sector energy routemap: https://www.techuk.org/insights/reports/item/16263-data-centre-energy-routemap

“Data centre power consumption is a dirty secret”  It is not.  In fact, consolidating IT activity into purpose-built facilities improves both transparency and efficiency.  Enterprise operators report corporate energy and carbon through mandatory and voluntary regimes.  The UK commercial sector monitors and reports its energy consumption at sector level.  Commercial operators here currently consume around 3TWh of power a year, just under 1% of electricity production.  Energy consumption is measured, audited and publicly reported at regular intervals in our CCA reports: https://www.techuk.org/images/CCA_First_Target_Report_final.pdf and https://www.techuk.org/images/CCA_Second_Target_Report_04.pdf

“5G will stimulate an explosion in energy demand” While 5G will enable the internet of things and this will have an infrastructure requirement in the form of Edge data centres, it will not be an explosion. Although there will be an energy requirement, 5G is significantly more energy efficient than 4G, which was itself more efficient than previous generations. The business case for 5G applications like M2M, IoT, autonomous vehicles will depend on the transactions and activities being performed very efficiently. There is also scope for 5G to deliver economy wide efficiencies – reducing congestion through real time traffic management, for example.

In conclusion…

All this is not new:  Back in 2013 the New York Times ran a negative article about the impact of data centres and we clarified those claims (not all of which were wrong) point-by-point in a similar way in Data Centres and Power: Fact and Fiction.  What disappoints us is that, although the industry has made huge strides, the tone of reporting has not.  We would have liked to see a more objective, evidence-based approach and we expected a more reasoned discussion from a respected institution like the BBC.  The tenet of the programme is valid – that many people are not aware of the energy impact of their internet activities. Where it went wrong was in misrepresenting the sector and making inaccurate claims: a failure of due diligence.  It is a shame that an otherwise worthwhile  attempt to demystify this complex and important sector was so misleading.

Useful sources:

IEA report: Digitalisation and Energy which explains the trade-off between ICT and broader energy savings: https://www.iea.org/reports/digitalisation-and-energy

Carbon brief on impact of video streaming: https://www.carbonbrief.org/factcheck-what-is-the-carbon-footprint-of-streaming-video-on-netflix

Sector Energy Routemap 2019: https://www.techuk.org/insights/reports/item/16263-data-centre-energy-routemap

CCA reports 1 and 2 https://www.techuk.org/images/CCA_First_Target_Report_final.pdf and https://www.techuk.org/images/CCA_Second_Target_Report_04.pdf

Data Centres and Power: fact or fiction: https://www.techuk.org/insights/reports/item/275-data-centres-and-power-fact-or-fiction

Ten Myths about data centres: https://www.techuk.org/insights/news/item/15255-ten-myths-about-data-centres

[1]https://digitalrealty.box.com/s/bserfy44rne36jxupnnnirdcbwdcvp7f