Microsoft More Than Doubles Electricity Consumption in the Last Four Years

Training artificial intelligence consumes a significant amount of energy. Chatbots and many other AI models count hundreds of millions of weekly users worldwide who use them to perform their tasks more easily, whether it’s writing an email or coding. All of this brings substantial bills for technology companies.

Visual Capitalist reported that training OpenAI’s chatbot GPT-4 consumed up to 62,000 megawatt hours, equivalent to the energy needs of a thousand households over five to six years. This can be confirmed when looking at Microsoft’s energy consumption in terawatt hours and the associated carbon emissions in millions of metric tons of CO2 over the last four years, according to the company’s 2024 Sustainability Report.

In just four years, Microsoft’s electricity consumption has more than doubled from 11 to 24 terawatt hours. For context, the entire country of Jordan, with a population of 11 million, consumes 20 terawatt hours of electricity in a year. This surge in electricity consumption has been accompanied by a 42 percent increase in total carbon emissions, indicating a relatively growing share of renewable energy sources at Microsoft.

Both trends coincide with the use of Microsoft’s Azure platform, which enables clients to perform cloud computing for training and running AI models, of which OpenAI’s ChatGPT is the most prominent. In fact, Microsoft has spent ‘hundreds of millions of dollars’ developing supercomputers solely for ChatGPT.

Training individual artificial intelligence models requires many resources and advanced data centers that consume much more energy than data centers not used for developing artificial intelligence or any other form of advanced technology. The construction of Microsoft’s data centers alone is responsible for 30 percent of the increase in emissions between 2020 and 2023.

This increase in emissions comes after Microsoft announced ambitions to become carbon neutral by 2030. Meanwhile, Google is in a similar predicament as they announced carbon neutrality within the same timeframe as Microsoft, but this now seems unfeasible as their emissions have risen by 48 percent in the last five years.

Earlier this year, consulting firm Grid Strategies reported that the U.S. will need approximately 34 new nuclear power plants or an additional 38 gigawatt hours of electricity in the next five years to meet the demand of data centers, factories, and electric vehicles.

In 2022, data centers, cryptocurrencies, and artificial intelligence consumed about two percent of electricity globally, according to an analysis by the International Energy Agency (IEA), and that share is expected to grow in the future.