The four most famous international exchanges:

Binance INTL	OKX INTL	Gate.io INTL	Huobi INTL
China Line　APP DL	China Line　APP DL	China Line　APP DL	China Line　APP DL

Note: The above exchange logo is the official website registration link, and the text is the APP download link.

According to The New Yorker, generative AI is consuming a lot of electricity while it is sweeping the world. ChatGPT, an AI chatbot developed by OpenAI, may consume more than 500,000 kWh of electricity every day to respond to about 200 million user requests.

Kilowatt-hour (kWh) is a unit of energy, also known as "degrees", which means that a 1,000-watt appliance consumes 3.6 million joules of energy after one hour of use. Kilowatt-hours are usually used as electricity billing units for power companies.

In comparison, the average American household uses about 29 kWh per day, which means that ChatGPT's daily electricity consumption is more than 17,000 times that of American households. If generative AI is used more, the power consumption may be even greater.

Alex de Vries, a data scientist at the Dutch National Bank, estimated in a paper that if Google integrates generative AI technology into every search, it will consume about 29 billion kWh of electricity per year, which is more than the annual electricity consumption of Kenya, Guatemala, Croatia and other countries.

Alex de Vries said: Artificial intelligence is very energy-intensive. Each AI server can consume the equivalent of a dozen British households, so the numbers accumulate very quickly.

How does it compare to Bitcoin mining? 

The New Yorker magazine also reported that Bitcoin mining consumes about 397 million kWh of electricity per day, which is 794 times the daily electricity consumption of ChatGPT. The annual electricity consumption (145 billion kWh) is more than the electricity consumption of the entire Netherlands.

Note: However, according to the Cambridge University Center for Emerging Finance, Bitcoin mining consumed 121 terawatt-hours of electricity last year, and the estimated amounts of the two platforms are very different.

It is difficult to estimate the power consumption of the AI industry

It is actually difficult to estimate how much electricity the AI industry consumes, because the operation of large AI models varies greatly, and large technology companies do not clearly disclose their energy usage. However, Alex de Vries made this rough calculation based on Nvidia data in his paper. Nvidia has a market share of about 95% of graphics cards.

Alex de Vries estimated in his paper that by 2027, the entire AI industry will consume 85 to 134 megawatt-hours (a billion times kilowatt-hours) per year, and by then, AI power consumption may account for 0.5% of global power consumption, which he pointed out is "a considerable number." This number is much higher than the world's highest power-consuming companies, such as Samsung, which consumes nearly 23 megawatt-hours per year, Google, more than 12 megawatt-hours, and Microsoft, which consumes more than 10 megawatt-hours to run its data centers, networks, and consumer devices.