This short note outlines our top conclusions about the energy consumption of the internet, which now comprises c2% of global electricity and 0.7% of global CO2. In the next decade, remarkably, the CO2 footprint of powering the internet could surpass that of producing oil or gas.
[restrict]
Energy consumed by the internet?
c500TWH of electricity is likely consumed to power the internet in 2019 (chart below), which is almost 2% of the world’s 27,000 TWH electricity market, and 0.8% of the world’s total global energy. To illustrate the point, consider the ‘top five facts’ we have found on the digital world’s energy footprint…
(1) A single google search activates data centers in 6-8 locations globally, consuming 0.3 watt-hours of energy and emitting 0.2g of CO2.
(2) Yet 80% of the internet’s data volumes are for video. For example, the 6.5bn cumulative YouTube views for the song, ‘Despacito’, have consumed as much energy as 50,000 US homes do annually, around 0.55TWH.
(3) A smartphone may weigh 100-200g, directly consuming 2kWh of energy per year, but the underlying data footprint is equivalent to a fridge, weighing 100-200kg, consuming 500-1,000kWh of energy per year.
(4) BitCoin mining is estimated to consume 40-60TWH of energy in 2019: about the same as the total grid in Greece, Peru or Singapore.
(5) Data centers. A $1bn Apple data centre planned for Athenry, Galway, expects to eventually use 300MW of electricity, 8% of Ireland’s national capacity, potentially more than Dublin. 40% of the energy use in data-centers is for cooling, hence one IBM data centre in Switzerland is used to warm a nearby swimming pool.
A meaningful contributor to global CO2.
0.8% of the world’s energy is a small percentage. Nevertheless, it is meaningful, because global CO2 emissions are extremely fragmented (chart below). The largest single contributor is passenger cars, but this is just c14% of the total. 30 individual sub-industries in our data-file each comprise <3% of the world’s total emissions, but cumulatively account for over 50% of the CO2e.
Internet energy use is also growing
Future internet energy use is uncertain, finely balanced between 20-40% growth CAGRs and 20-30% annual efficiency gains. But a sensible base case forecast for 2025 foresees growth to around 700TWH (chart below).
On the one hand, Cisco thinks global internet traffic will increase 3x in the next 5-years. 4.2 zetabytes of data will be transmitted in 2022, doubling from 2 zetabytes in 2019 and quardupling from 1.1 zetabytes in 2017. A rough doubling is suggested every 2-3 years. Similarly, Gartner estimated the number of internet connected devices will rise from 8bn in 2017 to 20bn in 2020. (Our CO2e estimates also exclude the emissions required to manufacture these devices and their battery materials).
On the other hand, chips continue getting more efficient, as we saw when screening the Majors’ Super-Computers (chart below). 60TWH of traditional data centers (c50% energy efficiency) are being phased out and replaced with hyperscale data centers (80-90% efficiency). The industry is also phasing out “zombie servers” which consume 25% of the sector’s energy but do no useful work.
Combining these trends, the IEA sees the internet’s electricity demand peaking at c4% of the global total. More pessimistic estimates from Huawei fear the internet could rise to 8-20% of global electricity use by 2030. But let us grant the benefit of the doubt to efficiency gains for now.
Internet emissions versus oil and gas emissions?
What is most striking about our numbers is that CO2-emissions to power the internet’s 700-1,000 TWH of energy needs in 2025 could emit 0.5-0.7 bn tons of CO2e. This could surpass the CO2-equivalent emissions from producing oil (if oil flaring can be eliminated) or from producing gas (if methane emissions can be eliminated).
Interesting considerations are raised for the way ESG investors allocate capital. It is doubtful that there will be mass campaigning for “divestment” out of internet stocks. Instead, we argue carbon-conscious investors, who are aiming to drive an energy transition, must look industry by industry, company by company, in order to prioritise leaders and avoid laggards.
[/restrict]