Energy Consumption of AI vs Cryptocurrencies
 
It is estimated that AI energy consumption will grow by approximately 25% annually until 2040. This represents an increase of 200% compared to 2022

Google and Microsoft each reported energy consumption of 24 TWh in 2023, in more than about 100 countries such as Jordan, Iceland, Ghana, the Dominican Republic, and Tunisia. This figure could double in 2026 to reach 1,000 TWh.

By 2026, the energy consumption of data centers critical to AI is anticipated to exceed 1,000 TWh, equivalent to the energy consumption of all of Japan.

 It is predicted that AI could account for up to 0.5% of global electricity consumption by 2027. 

 If Google's search engine became a ChatGPT-type AI tool, the 9 billion daily searches would require the same energy that Ireland consumes. 

Annual Consumption: According to the University of Cambridge Bitcoin Energy Consumption Index, Bitcoin energy consumption is estimated at approximately 137.78 TWh annually, although some more recent estimates place it between 70.4 TWh and 143 TWh. 

Although there are no exact figures for the power consumption of ChatGPT specifically, it is estimated that training advanced AI models such as GPT-3 and GPT-4 consumes between 100 and 300 MWh per training, and continued use of these models also requires a considerable amount of energy for inference.

 In 2022, AI and cryptocurrencies consumed almost 460 TWh of electricity, equivalent to 2% of total global production