Report: AI, crypto drive spike in data-center energy demands

The rapid growth of emerging technology such as artificial intelligence (AI) and cryptocurrency is driving higher electricity consumption from data centers, which could see consumption double globally by 2026, according to a new report.

Energy-intensive data centers are “an important new source of higher electricity consumption” and play a significant role in driving the growth of electricity demand in many regions, according to International Energy Agency’s (IEA) Electricity 2024 report. The report offers an analysis of recent policies and market developments in all things electricity, providing forecasts through 2026 for electricity demand, supply and carbon dioxide emissions.

According to IEA, global data-center energy consumption could get so high that it equals that of the entire country of Japan, or more than 1,000 terawatt-hours (TWh), by 2026. This is an increase from consumption of an estimated 460 TWh in 2022. Moreover, in the US alone, data-center consumption is expected to account for more than one-third of all energy demands by 2026.

This overall surge in consumption by computing power – which is necessary for technologies like AI and cryptocurrency to exist and evolve – will require updated regulations and technological improvements regarding efficiency to moderate the effects of such a dramatic increase, according to the report.

Meanwhile, strong growth in emerging economies paired with an anticipated recovery in industry and ongoing electrification of the residential and transportation sectors in many global regions “will be the mainstays of increasing electricity over the next two years,” the IEA report found.

Balancing tech benefits and energy needs

Technology providers are aware of the strain that emerging technologies are putting on energy consumption, as they require ever-more compute power driven by large clusters of machines that require electricity as well as cooling to maintain always-on availability.

Generative AI (GenAI) in particular – with its requirement to quickly search and process massive data stores and deliver results instantaneously – will have a significant global environmental impact that the industry must consider as it continues to build and evolve the technology.

“The data processing and modeling of GenAI has an extensive resource consumption and carbon emission effect, which must be balanced against the potential outcomes of the learning models,” according to a recent Gartner Group report, “9 Environmental Implications of Generative AI.”

While AI will provide numerous benefits to helping global companies optimize emission reduction through climate-change modeling and other risk-aware solutions that its output can deliver, it has a fundamental “consumption problem,” according to Gartner. Organizations need to understand from the outset that it requires enormous amounts of energy and water for compute processing and cooling, and make smart energy decisions going forward to meet these demands.

To reduce GenAI’s energy footprint, technology and business leaders both providing and using the technology can consider several strategies going forward to get ahead of the problem. One that Gartner recommends is to monitor energy consumption during machine learning and then halt its training as soon as improvements flatten out.

Organizations also should only run AI in the right place at the right time, considering local energy consumption needs and other factors and optimizing the use of resources when running the technology.

“Best practice is to use energy-aware job scheduling for generative AI, along with carbon tracking and forecasting services to reduce related emissions,” according to another Gartner report, “Balance the Environmental Perils and Promises of Generative AI.”