With the advent of generative artificial intelligence ChatGPT, various companies and institutions have also launched their own large-model artificial intelligence AI.
The application of this technology has undoubtedly brought many conveniences and many untapped opportunities to mankind, for example, it is now easier for students to solve complex problems, medical advances and so on.
However, these conveniences are not without a price, and they are not small. Google, for example, uses 10 to 15 percent of the company's total electricity consumption, or about 2.3 billion kWh a year, which is equivalent to the annual electricity consumption of all households in a state capital.
Even after training, these AIs still require a lot of computing power to run, and therefore a lot of energy, OpenAI spends up to $700,000 a day on ChatGPT.
It is assumed that interest in AI technology continues to grow, coupled with an abundant supply of chips. Then generative artificial intelligence (ChatGPT, Bing Chat, etc.) will generate huge energy consumption, which may even exceed the electricity needs of some countries.
The study notes that AI-related energy consumption is likely to increase significantly in the coming years. By 2027, the energy consumed by generative AI could power a country the size of the Netherlands for a year, equivalent to about 85-134 terawatt hours (TWh).
OpenAI trained GPT-3 to consume 1.287 GWh, which is roughly equivalent to 120 U.S. households using electricity for 1 year. And this is only the upfront power used to train the AI model, accounting for only 40% of the electricity consumed by the model when it is actually used. In January 2023, OpenAI consumed the potential equivalent of 175,000 Danish households in an annual electricity consumption in just one month. Google AI consumes 2.3 terawatt hours of electricity per year, which is equivalent to the electricity consumption of all households in Atlanta for 1 year.
At present, the problem of energy consumption has attracted widespread attention. The best way to minimize costs is to build data centers in areas with good renewable energy while also improving energy efficiency. For example, data centers in Iceland and Norway run on green energy and are not affected by global electricity costs.
In addition, Microsoft is purchasing renewable energy and taking other steps to meet its goal of becoming carbon negative by 2030 and further committed to providing 100% renewable energy to all facilities by 2025. Amazon's AWS cloud division is also shifting to renewable energy, shifting from diesel to hydroprocessed vegetable oil (HVO) to fuel backup generators in its European data centers.
But as technology continues to advance, more efficient and environmentally friendly generative AI solutions are expected in the future, and the power consumption problem should be solved by then.