Carbon emissions behind the expansion of artificial intelligence

AI can process data at lightning speed, and the latest version of ChatGPT can now talk to users and identify images.

  While the growth in AI and cloud computing services has been transformative in a variety of sectors, it is straining data centers and power grids due to the massive amounts of energy consumed.

Gulf countries

The United Arab Emirates and Saudi Arabia are among the Gulf countries that have adopted artificial intelligence and digitization in government, business and industry. Its advantage lies in modern infrastructure, unencumbered by legacy systems that could hinder its net-zero ambitions.

Carbon footprint

Regardless of discussions about the benefits and threats of game-changing technological advances, the carbon footprint of AI and cloud computing services must be managed. The challenge is to address this increase in electricity demand in a way that is sustainable and does not conflict with climate change goals. Developers like Microsoft, Google, and OpenAI are working on solutions, including using more renewable energy and more sustainable cooling systems.

While the global electricity system is transforming with the rapid growth in renewable energy, clean energy sources – solar, wind and hydropower – make up only 30 percent of global electricity generation. Oil, gas and coal still dominate, accounting for more than 80 percent of primary energy consumption.

The increasing carbon footprint of artificial intelligence

A number of studies suggest that the increase in demand from artificial intelligence and cloud computing has led to higher carbon emissions, undermining decarbonization efforts. A Columbia Climate School paper titled “The Growing Carbon Footprint of AI” estimates that greenhouse gas emissions from the energy consumed by data centers exceed those of the aviation industry.

Data centers have seen a greater volume of online activity since the 2020 coronavirus pandemic, when remote working habits and reduced travel meant more companies went virtual. Growth in the use of smartphone applications, Bitcoin mining, and cryptocurrency transactions have also contributed to increased demand for power and data.

Data processing

Data processing generates heat with every bit change, so data centers need to be cooled at all times, which increases the electricity load.

At a Petrodiplomacy conference organized by the Arab Gulf States Institute in Washington in June 2023, attendees heard that Bitcoin mining uses electricity equivalent to the diesel fuel consumption of the entire U.S. rail fleet. The conference also heard that the energy used to train the ChatGPT3 data app - which has already been replaced by a more advanced version - was equivalent to the energy use of 138 US homes in one year.

The Columbia report says that a single request in ChatGPT can consume 100 times more power than a single Google search. A single AI training unit could consume more electricity than 100 American homes annually, according to a Bloomberg report titled “AI Is Booming, and So Is Its Carbon Footprint.”

“The CO2 emissions from this exercise are equivalent to three round trips from New York City to San Francisco,” noted Sarah Vakhshouri, CEO of SVB Energy International.