Did you know that streaming videos will negatively impact your carbon footprint? Despite recent headlines, it doesn’t cause as much damage as we are led to believe, especially in comparison to other activities and sectors.
There has not been a year go by that we cannot say we have all seen a drastic improvement to our internet; whether the connection or the way we use online services. Even before we were confined to our home in a global pandemic, which led to a rise in streaming all over the world – Of course it did, and by leaving little room for other forms of entertainment, the likes of Netflix and Hulu are seeing increased traffic than ever before.
In this piece of writing we are looking to consider the carbon footprint of streaming services and the energy use and carbon emissions associated with devices, network infrastructure and data centres.
By diving deep into IEA’s Data Centres and Transmission Networks analysis alongside other sources, we look to deconstruct the results of misleading reports on the results of watching just 30 minutes on a site like Netflix, and instead highlight that a considerable amount of statistics reported may have exaggerated the actual climate impact, times 90.
The primitive reason tat streaming video today creates a lower impact on our climate than these reports claim is due to the fast improvements in the energy efficiency of data centres, networks and devices, but by showing efficiency gains, rebound effects and new demands from emerging technologies, such as artificial intelligence and blockchain, we may have reason for concern about this sector’s environmental impact over the following decades.
As of December last year, the energy intensity figures for data centres and transmission networks were updated to reflect more accurate research. This resulted in the central IEA estimate for one hour of video streaming I drop from 82gCO2 in 2019 to just 36gCO2 in February of the following year. This was brought together by corrected values by French experts The Shift Project in June of that year, alongside updated charts and comparisons and other estimates quoted in the media.
Fact, or false? 30 minutes of Netflix uses the same amount of CO2 emissions as driving almost four miles.
Well, articles from the likes of the New York Post, CBC, Yahoo, DW, Gizmodo and Physical.org (to name a few) have retreated the claim that 1.6kg of CO2, the amount generated by watching 30 minutes of Netflix, is about the same as driving almost 4 miles. They took these figures from a report made in July of 2019 by the Shift Project, focusing on the “unsustainable and growing impact” of online video.
This report claims that streaming was responsible for over 300million tonnes of CO2 (MtCO2) in 2018, the equivalent to emissions from France. Following this report two years later, the project corrected their statistics as a byte conversion error, revising the original data down 8-fold to only 0.2kg per half hour of streaming.
The original, incorrect statistic is around eighty times higher than any peer review study focusing on the energy and emissions of video streaming, whilst the new estimate of 0.4kg per hour is closer to the 2014 study linked above; this study found that video streaming sites in the United States in 2011 emitted only 0.42kg CO2e per hour on a lifecycle basis, including “embodied” emissions from manufacture and disposal of infrastructure devices. From operations, emissions were comparable in scope to The Shift Project analysis, at only 0.36kg CO2e per hour.
Still, with data centres and networks improving at a rapid rate, the energy use and emissions from streaming today should be substantially lower than the estimates made over 12 months ago, with their success doubling every couple of years.
Fact, or false? Watching a YouTube video uses over 16000W of electricity, which makes up the equivalent of using 15 big screen TVs.
Considering electricity consumption alone, the origin Shift Project figures suggest that one hour of Netflix brings our consumption to 6.1 kilowatt hours of electricity. 6.1 kilowatt hours of electricity would:
- Allow us to drive more than 30km in a Tesla Model S
- Constantly power an LED lightbulb for one month
- Boil a kettle once a day for nearly three years
Netflix estimates an average of two hours of streaming per day by each of its 167 million subscribers. This means that The Shift Project’s corrected figures would lead Netflix to consume around 94 terawatt hours per year, 200 times larger than the figures reported by Netflix themselves in 2019.
Channel 4 Dispatches also became part of the conversation after commenting that 2017’s hit song “Despcito ft. Justin Beiber” by Luis Fonsi and Daddy Kankee could consume 900 gigs watt hours of electricity, or 1.66 kWh per viewing hour. With over 1billion viewing hours per day, YouTube has the potential to use over 600 TWg a year, which is 2.5% of the globe’s electricity use. This would be more than the electricity used by all of the data centres and transmission networks in the world, at 200-250 TWh.
It is obvious that these statistics are too high, but how can we calculate just how much? With inaccurate assumptions made on the consumption made by these networks and data centres, each statistic has become more exaggerated over time, whether or not this is intentional.
Despite the Shift Project’s assumed statistics being reviewed and corrected both in 2019 and then again in the following year, together, these flaws can seriously exaggerate the electricity consumed by streaming video, and with the media reporting on these incorrect statistics, this false information is more common than the real thing; The original “1.6kg per 30 minutes” claim was too generous to bitrate (the amount of data transferred each second during streaming), which somehow moulded itself to a figure of 24 megabytes per second (Mbps), the equivalent of 10.8gigabytes (GB) per hour.
The 10.8GB calculation is six times higher than the global average bitrate for Netflix in 2019, at around 4.1Mbps or 1.9GB per hour, excluding cellphone networks, and over triple the transfer rate of HD (3GB/hr).
Other typical transfer rates are:
- 7GB per hour for ultra HD (UHD/4K)
- 0.4GB per hour for standard definition (SD)
- 0.25GB per hour for mobile
The easiest way for you to test this directly could be by using the likes of the 4G allowance on your phone, if you have the ability to track your data usage at any time.
The visible difference was formed with a stated assumption of 3Mbps apparently being converted in error to 3MB er second, with each byte equivalent to eight bits. Although The Shift Project fixed this error in this June 2020 update, they did not correct their other inaccurate statistics.
The above charts brought together by The Shift Project and IEA statistics shows three separate ways in which the Switch Project overestimated the electricity use necessary for streaming video. This is shown in bitrate, as well as where they overestimated the energy intensity of data centres and content delivery networks, known as CDNs. CDNS serve subscribers with streaming videos around 35-fold, which is comparable to the figure taken from Netflix electricity consumption data and subscriber usage data, both collected in 2019.
The analysis above shows that the Shift Project also exaggerates the energy intensity of data transmission networks by around 50-fold. This is calculated based on the average bitrates for streaming video, and is a result of using high and outdated energy-use assumptions for various access modes. For example, comparing the likes of 0.9kWh/GB for mobile compared to the most recent peer reviewed estimates of 0.1-0.2kWh/GB for 4G mobile in 2019.
The February 2020 analysis linked above shares that the Project assumptions for data transmission energy intensity were up by a long shot in comparison to most recent estimates, from 0.15-0.88kWh to 0.025-0.23kWh/GB. In contrast to this, though, the latest research shows that the data-based intensity values are not appropriate for making these estimates, explained at 41min 44sec of this YouTube video. Experts suggest that using time based energy intensity values, and so IEA updated their findings under these values.
Another flaw in the Shift Project report is how it assumes that viewing only occurs on smartphones and laptops, both equally at 50% of the statistics. Instead, it is true that 70% of viewing occurs on TVs, 15% on laptops and 10% on tablets, with only 5% of viewing on smartphones, according to Netflix. This suggests that the Shift Project is underestimating 4-fold.
Fact, or false? One hour of streaming video only uses around 0.08kWh typically, but actual consumption depends on the device, network connection and resolution.
Well, this one is certainly true. Together, an updated analysis claims that streaming via Netflix in 209 would typically consume around 0.077 kWh of electricity per hour, which is 80 times less than the original figure, and ten times less than the corrected figure of 0.78. Although, due to the fact that these specifics are dependent on the device used and the network connection, the below chart is highly sensitive.
CHART: Electricity per hour of streaming video, Shift Project compared to different use cases, 2019
To further define this, a 50-inch LED television consumes at least 100 times more electricity than a smart phone and 5 times more than a laptop. Because your mobile is designed to be efficient, data transmission accounts for more than 80% of the electricity consumption when streaming. If you were to watch an hour-long SD video trough your phone, on WiFi, you would use only 0.037kWh, which is 170 times less than The Shift Project’s statistic.
Based on your average viewing habits, it can be assumed that viewing deceives account for 72% of energy use, with data transmission following behind with 23%, and data centres at 5%.
In opposition, the Shift Project claim that viewing devices count for less than 2% of total energy use, with both data transmission and energy use being overestimated by 35% or more.
CHART: Share of streaming energy use from devices, data transmission and data centres
Overall, we can confidently say that the carbon footprint of streaming video is still relatively small, especially in countries with low carbon electricity. Our footprint when streaming firstly depends on the electricity usage set out above, and secondly to the CO2 emissions associated with each unit of electricity generation.
Alongside the likes of electric vehicles, this means that the overall footprint depends most heavily on how the electricity is generated. If this is powered by the global average electricity mix, 30 minutes of a show on Netflix in 2019 should release around 0.018kg of CO2e (18 grams)- This is 90 times less than the figure we were give by The Shift Project, and 11 times less than the corrected figure of 0.02kg.
The IEA estimate is also considerably lower than the estimates quoted in media; it is 22 times lower than the disparity claim cited on Channel 4, the BBC and Al Jazeera, as well as11 times lower than the claim by Save on Energy, who sited that 80million views of Birdbox emitted 66ktCO2. IEA’s estimate of 36g CO2 per hour is over 1200 times lower than what Marks et al. (2020) estimated, at 35 hours of HD video equalling 2.68t CO2 or 77kgCO2 per hour.
CHART: CO2 emissions associated with a half-hour show on Netflix, 2019
To help you put this into context, a half hour of watching Netflix, rather than a allowing us to drive 30km in a Tesla Model 5, is more likely to show the equivalent of driving 100 meters in a conventional vehicle. Still though, this figure is deeply pendant on the mix of generated energy in the country we are diagnosing. In France, around 90% of their electricity is generated by low carbon sources, meaning emissions would be around 2gCO2e, making up to 10 meters of driving.
CHART: Breakdown of CO2 emissions from 1 hour of streaming, World (2019)
Using the average emission factors for the specific country in question still has the potential to develop flaws in calculating emissions, particularly from data centres. Technology firms operating large data centres are often at the head of corporate clean energy procurement, accounting for about half of renewable power purchases agreements in the last number of years.
Our electricity mix is also quickly decarbonising in many parts of the globe, for example, the United Kingdom’s emissions intensity of electricity fell by nearly 60% from 2008 to 2018. In comparison to that of 2019, emissions intensity of electricity falls by around a quarter by 2030 in the IEA stated policies scenario, and by half in the sustainable development scenario.
While the carbon footprint created by streaming your favourite shows may remain low, it is still reasonable to expect the statistics shared in this document to rise through the coming years, given exponential increases in usage. Energy efficiencies of digital technologies has improved rapidly, and there have been a number of improvements in the likes of computing, which is known as “Koomey’s Law’. Koomeys Law describes trends in the energy efficiency of computer usage, which has doubled roughly every 1.6 years since the 1940s, and every 2.7 years since 2000. Another trend has been observed in data transmission networks, with energy intensity halving every two years since 2000. Even a change from CRT to LCD screens on your devices have the potential to make a difference.
With the short lifespan of your mobile phone or tablet, the officially of overall stock devices, data centres and networks is improving rapidly. For example, you can see in increasingly efficient IT hardware and a major shift to what is known as hyper scale data centres, help has been given to eep electricity demand flat since 2015. Today, worldwide data centres only consume around 1% of global electricity use, even with the rise in internet traffic since that year.
CHART: Global trends in internet traffic, data centre workloads and data centre energy use, 2015-2021
AND
Global data centre energy demand by data centre type, 2015-2021
It is important to consider, though, whether the efficiency and growth in demand have the potential to eventually cause problems. Well, we can say that streaming media is growing rapidly, with Netflix subscriptions growing by 20% last year, and electricity consumption rising by 84%. Whether or not our situations are subject to change, these statistics project what we may see in the future. In recent months, we have seen the launch of new streaming sites and gamin services, mobile networks growing video traffic by over 55% every year and phones already accounting for more than 70% of how we stream sites like YouTube today.
It is so easy to access media streaming online today, but we are yet to consider what types of streaming, for example e-books and online shopping, may quantify the net environmental impacts by taking away the other forms of consumption, for example by driving or physical copies of books that may lead to deforestation. It is clear that the rise in digital technologies like 5G and VR really are likely to further the demand for data centres and network services, with researches already looking into the potential impacts of these technologies as they develop.
It is becoming increasingly likely that the gains of current technologies may not have the ability to keep peace with the growing demand for data, and so many continue to decarbonise the electricity supply with the intention of efficient next-generation computing and technologies needed.
In closing, you can see here that streaming video is a safe way to say that you are being conscious today, but rather than relying on media coverage, we still must consider the impacts of further growth on streaming and data usage over the coming decade.
While all sectors and technologies are needed to help achieve the goals of the Paris Agreement, around 1.5% of all global carbon emissions are currently coming from digital technologies, but this is not necessarily daunting. In fact, the likes of artificial intelligence could help accelerate climate action and even has the potential to make oil extraction cheaper, and extend the lifetime of coal and plants.
On completing this document, graphs to work from can be found here: https://www.iea.org/commentaries/the-carbon-footprint-of-streaming-video-fact-checking-the-headlines