As Reliance on AI Grows, So Do Expected Hikes in Energy Costs, Demand

The artificial intelligence revolution has done more than just provide students a shortcut to demystify a physics formula or translate a phrase for Spanish class. It is transforming how entire industries operate. But the efficiency and ease borne by AI and its chatbots comes at a significant cost.

The increasing use of AI requires a massive amount of electricity, which is severely taxing our energy grid and natural resources. As AI grows more sophisticated, so too does the demand and the cost of building more data centers – along with the electricity and water necessary to power and cool those facilities.

According to a Goldman Sachs Research report released in May, the demand for electricity to power data centers will increase 160% by 2030. The report’s authors said one query to ChatGPT, OpenAI’s ubiquitous chatbot, requires nearly 10 times as much electricity to process as one Google search. That portends a coming “sea change” in how the United States, Europe and the rest of the world consume power and how much that power will cost.

For reference, when OpenAI launched ChatGPT in November 2022, it gained 100 million monthly active users within two months. By January 2023, it was receiving roughly 13 million queries a day.

The technology sector already consumes a substantial amount of water to cool power-hungry data centers, which are integral for meeting the growing demand for online services and generative AI products. Company executives and board directors investing in AI would be wise to keep the water footprint of data centers in mind, wrote Cindy Gordon, CEO of Toronto-based AI platform Sales Choice, in Forbes. “Rising water use in data centers is very concerning due to the incredible global freshwater scarcity we have,” she said.

Data center carbon dioxide emissions are also expected to skyrocket at a “social cost” of upwards of $140 billion over the next six years, the Goldman Sachs report found. Although most tech companies working on AI do not disclose their emissions, Google recently released a sustainability report that mentioned its greenhouse gas emissions have increased 48% since 2019 as a result of data center energy consumption and supply chain emissions. “As we further integrate AI into our products, reducing emissions may be challenging,” the report said.

Data center owners and policymakers alike are mulling potential solutions to the tech industry’s natural resources drain. Countries such as Ireland and Singapore, for instance, have put stops on data center construction to avoid overwhelming their electrical grids. In a report issued in May, the nonprofit Tony Blair Institute for Global Change touted the prospect of data center developers investing in renewable energy sources. Factoring in the effects of new data centers coming online could also help state legislators write more effective laws.

And perhaps technological innovation can play a role in relieving the strain on the power grid that it has caused. As an example, Nvidia claims its chips used for machine learning now operate 45,000 times more efficiently than their predecessors. Moreover, AI applications may eventually enable users to identify and address more wasteful and inefficient uses of power in our everyday lives.

If AI can tell us how to set air conditioning units for optimal performance or help develop strategies to enhance the fuel efficiency of our cars, that could turn out to be a net gain for the power grid.