Digital Carbon YOLO Calculator
As the world moves to cleaner energy, the Internet's carbon footprint shrinks. At the same time, AI and video consumption are vastly increasing energy use and therefore total emissions.
Methodology
While usage patterns may vary, research has concluded that it is best to measure average impact on a time-based basis. Depending on the scope of an evaluation, data center energy use is also a point of consideration as is the full lifecycle of a device (around 80% of the lifecycle carbon emissions of a mobile device are in the production phase). Our calculations are based on data provided by greenspector and they measured them on a Samsung S7 smartphone (Android 8). We also use data provided by Our World in Data to factor in the actual average energy mix of the selected country.
Energy consumption per AI request varies significantly between models — a large reasoning model can use 10× or more energy than a smaller, optimised one. Our per-model energy values (Wh per request) are sourced from Epoch AI (2024), Luccioni et al. (2023), and IEA (2024). Values for closed, proprietary models are estimates — exact figures are not publicly disclosed by providers. CO₂Eq is calculated using the actual carbon intensity of your country's electricity grid (gCO₂/kWh) from Our World in Data, which accounts for the full generation mix rather than just the renewable share.
Our estimate of ~9 Wh per hour of HD streaming covers the data centre, network transmission, and device combined. This is based on IEA (2020) and Carbon Brief / Kamiya (2020). The widely-cited figure of 1.58 kWh/hour from The Shift Project (2020) was retracted and is not used here. CO₂Eq is calculated using the actual carbon intensity of your country's electricity grid.
*Even though the actual CO₂Eq impact of renewables may not be zero, we assume this for now to reduce complexity.
GHG emissions (CO₂Eq) by country
Global greenhouse gas emissions (CO₂Eq)
The internet accounts for about 4% of global emissions (CO₂Eq), while the airline industry´s share is about 3%. So the Internet´s share is not only larger, we think it´s also easier and faster to reduce. +Analytics Pro is our first solution to this.
onEco® +Analytics ProEnergy use and carbon impact
- Data Centers
-
Data centers and data transmission networks alone required up to 660 TWh of electricity in 2021. There´s already a lot of renewable energy in use, but it´s not enough to reach net-zero in the near future.
- Crypto
-
Crypto mining accounted for up to 140 TWh of energy in 2021, which is roughly half of data center energy use. Most of this is Bitcoin mining, far less is for Ethereum. As Ethereum transitioned to proof-of-stake mechanism in September 2022, energy use for Ethereum is expected to be reduced by 99.95%.
- Social Media
-
Carbon impact of TikTok is at 2.63 gEqCO₂/min of newsfeed use on a mobile device. That´s more than all other social networks. Have a look at our YOLO Calculator
- Artificial Intelligence
-
AI is a major driver of energy consumption and therefore a major driver of carbon impact. The energy consumption of AI is expected to increase by 300% by 2030.