Something to talk about: ChatGPT’s enormous energy drain revealed

By Dr. Tim Sandle
DIGITAL JOURNAL
September 6, 2024

Image: — © AFP

Artificial intelligence’s energy consumption is skyrocketing, driven by the expansion of large language models and other forms of AI. To keep the data insights turning, carbon needs to be burnt. A new survey reveals just how much energy is required for ChatGPT to handle user queries throughout the year.

Since its launch nearly two years ago, OpenAI’s ChatGPT has quickly captured the public’s attention, amassing an estimated 100 million monthly users within just two months. However, while its capabilities, from drafting essays to solving complex math problems and writing code, are impressive, they come with a hefty energy cost. Each ChatGPT query reportedly consumes about 0.0029 kilowatt-hours of electricity, nearly ten times the 0.0003 kilowatt-hours required for a standard Google search.

To put the scale of ChatGPT’s energy consumption into perspective, the firm BestBrokers has calculated its total annual electricity usage for generating command responses alone. The results are concerning: ChatGPT consumes approximately 226.82 million kilowatt-hours each year just to process user queries.

According to the calculations, this amount of electricity is enough to fully charge about 3.13 million electric vehicles, each with an average battery capacity of 72.4 kWh. In fact, this represents nearly 95 percent of the 3.3 million electric cars on U.S. roads by the end of 2023.

In terms of individual contributions,ChatGPT uses 621,429 kWh of electricity every day to handle over 200 million user queries. To put that in perspective, it consumes more than 21 thousand times the daily energy of an average U.S. household, which uses about 29 kWh.

Over a year, ChatGPT consumes 226,821,615 kWh to manage more than 78 billion prompts. This adds up to about $29.71 million in energy costs, based on the U.S. average commercial electricity rate of $0.131 per kWh as of June 2024. Each query costs $0.00038 in electricity.

What does this energy drain mean?

As it stands, the power used by ChatGPT to generate responses in a year could fully charge 3,133,371 electric vehicles, each with an average battery capacity of 72.4 kWh. That’s nearly 95 percent of the 3.3 million electric cars on U.S. roads by the end of 2023.

What could be done with the energy?

ChatGPT’s yearly energy consumption to handle requests could also power 21,602 U.S. homes for a full year. While this is only 0.02 percent of the 131 million households in the country, it’s still a significant amount of energy, especially considering the U.S. ranks third globally in household numbers.

The energy ChatGPT uses in a year to answer questions could also charge 47.9 million iPhones 15 every day for an entire year, each with a battery capacity of 12.98 Wh. Additionally, the chatbot consumes as much energy to process user queries in one hour as it would take to stream video for 137,728 hours in Europe.

Image: — © AFP

What does this mean geographically?

ChatGPT’s yearly energy consumption for handling prompts exceeds the annual electricity usage of twelve small countries and territories, including Gibraltar, Grenada, Dominica and Samoa. It could also power all of Finland or Belgium for a day, or keep Ireland running for over two days.

The future

The future looks like recording a great energy drain for training the GPT-4 model, with over 1 trillion parameters, required 62.3 million kWh of electricity over 100 days, 48 times more energy than it took to train GPT-3, which consumed nearly 1.3 million kWh in 34 days.