Abstract
It has been said that streaming 30 minutes of Netflix releases as much carbon as a short car ride, but this isn't true, says Michael Le Page
EVERY time you search the internet or stream a video, a computer in a data centre somewhere in the world spins into action. With rising internet use, there have been fears that this is driving a big increase in energy consumption, undermining efforts to limit climate change.
The good news is that, according to the most detailed study to date, the energy use of data centres rose by just 6 per cent between 2010 and 2018, despite a 550 per cent hike in demand. Dramatic efficiency improvements have almost cancelled out the big increase in use, according to Arman Shehabi at the Lawrence Berkeley National Laboratory in California and his colleagues (Science, doi.org/ggm3sk).
That isn't all. Video streaming in particular has been singled out as a major driver of rising energy use, with several media outlets reporting that the emissions generated by watching 30 minutes of Netflix are the same as driving almost 4 miles.
This claim is based on figures for global streaming in a 2019 report by French think tank the Shift Project. But according to an analysis by George Kamiya at the International Energy Agency published by CarbonBrief last week, errors in those figures mean emissions from viewing Netflix were overestimated by 30 to 60 times.
So does this mean we can stop worrying about the climate impact of internet traffic and, in particular, streaming videos? Not quite.
The main conclusion of the Shift Project report was that transmitting and viewing online videos generates 300 million tonnes of carbon dioxide a year (mtCO2), or nearly 1 per cent of global emissions, as New Scientist reported (although we didn't mention the unlikely driving claim).
Kamiya's analysis applies only to Netflix, but he thinks the Shift Project did overestimate the overall emissions from streaming video. “I haven't done any estimates of the global emissions from all streaming video,” he says. “This would require quite a bit more analysis.”
One author of the Shift Project report, Maxime Efoui-Hess, says his team welcomes an open discussion and will contact Kamiya to talk about the differences in numbers. Kamiya makes the same recommendations for reducing emissions as the report did, Efoui-Hess points out, so these won't change.
An independent study by Chris Preist at the University of Bristol, UK, estimated that YouTube alone generates 10 mtCO2 a year, and he says nothing in Shehabi's or Kamiya's work undermines this conclusion.
According to the Shift Project, YouTube accounts for less than a fifth of total streaming emissions. If that is correct, overall emissions due to streaming based on Preist's research would still be more than 50 mtCO2 a year – a lot less than 300 mtCO2, but still substantial.
While big efficiency gains appear to have limited the rise in energy use by data centres, we can't assume this will continue, says Shehabi. The centres may have already implemented all the easy wins.
Meanwhile, Anders Andrae, an efficiency researcher for Chinese technology firm Huawei, thinks Shehabi's team has got it wrong. He is sticking with his forecasts of huge rises in electricity demand due to growing internet and computer use.
Resolving these conflicting numbers won't be easy. We have limited information on the data industry, as companies disclose little. All studies are essentially educated guesstimates.
Another complication is that rising energy use in this sector might sometimes lower demand in others. For instance, demand for streaming video might shoot up this year if millions of people are forced to stay at home because of the coronavirus outbreak. But if they are doing this instead of driving or flying to places, overall carbon emissions will fall.