New Technologies Are Needed to Curb Data Center Energy Use, Says the IEA

Data centers and transmission networks have made advances in energy efficiency in recent years that have limited their energy consumption growth even as their capacity has increased rapidly.

However, with the rising adoption of data-intensive applications such as artificial intelligence (AI), cloud computing, cryptocurrency mining, and blockchain adoption,?the industry will need new technologies and solutions to achieve net zero emissions, according to a report by the International Energy Agency.

Data centers consumed an estimated 240-340?terawatt hours (TWh) of electricity globally in 2022, accounting for around 1-1.3% of final electricity demand. That excludes electricity consumed by cryptocurrency mining, which was estimated to be around 110TWh, or 0.4% of annual demand. Data transmission networks similarly consumed 260-360?TWh in 2022, or around 1-1.5% of global demand.

Since 2010, energy consumption from data centers (excluding cryptocurrency mining) – and, as a consequence, their carbon emissions – has increased moderately despite strong growth in demand for their services. This is in part because of improvements to the efficiency of IT hardware and cooling systems and a transition to hyper-scale facilities from small, inefficient enterprise data centers.

It is also because information and communications technology (ICT) companies are increasingly prioritizing purchasing renewable energy, as well as the broader decarbonization of electricity grids in many parts of the world.

Additionally, waste heat from data centers is increasingly being used to help supply heating to nearby commercial and residential buildings or industrial heat consumers, reducing their energy use.

However, governments and industry players need to invest in advancing energy efficiency, research and development, and the decarbonization of electricity supply and supply chains to curb energy demand and rapidly cut emissions over the coming decade. This investment is essential to get on track with plans to reach net zero carbon emissions by 2050.

To achieve net zero, emissions must halve by the end of this decade.

Large Data Center Energy Demand On the Rise

Despite the strong gains in energy efficiency, energy consumption from large data centers has climbed by 20-40% in recent years as their workloads have increased rapidly.

Amazon, Microsoft, Google, and Meta saw their combined electricity consumption more than double from 2017 to around 72TWh in 2021. Overall energy consumption at data centers, excluding cryptocurrency, is set to continue rising moderately over the coming years.

While total global power consumption at data centers has grown only slightly, some smaller countries where markets are expanding are seeing rapid growth. For example, Ireland’s data center power demand has more than tripled since 2015, accounting for 18% of the country’s total electricity consumption in 2022. Data centers and other large non-industrial electricity users could account for 28% of national demand by 2031.

Data center electricity demand in Denmark is projected to increase six times by 2030 to account for close to 15% of the country’s electricity consumption.

Global Trends in Data Center Use

New Technologies to Drive Energy Use

Emerging technologies such as streaming, cloud gaming, blockchain (particularly Proof-of-Work), artificial intelligence, machine learning (ML), and virtual reality (VR) are expected to increase demand for data services.

For example, Bitcoin – the most prominent Proof-of-Work (PoW) blockchain – consumed an estimated 110TWh of electricity in 2022, which was 20 times more than in 2016.

As the adoption of blockchain-based applications expands, managing the implications for energy consumption is likely to become increasingly important for the energy industry and policymakers.

The rapid emergence of AI and ML applications has potentially significant implications for data center electricity demand in the coming years. The amount of computing power required to train the largest ML models is growing rapidly, but it is unclear how quickly overall ML-related demand is rising.

Data from Meta and Google indicates that the training phase only accounts for around 20-40% of overall ML electricity use, with 60-70% used for the application of AI models and up to 10% for model development. ML is estimated to have accounted for 10-15% of Google’s total electricity consumption in 2019-2021, growing at a pace of 20-25% in line with its overall energy demand growth.

The combination of increasingly large AI models and higher computing demand is likely to outpace the improvements in energy efficiency, resulting in a net growth in total AI-related power demand over the coming years. AI applications can help optimize data centers to reduce their electricity use, but the rapid and mainstream adoption of AI chatbots such as OpenAI’s ChatGPT and Google Bard are likely to accelerate energy demand growth.

Demand for edge data centers is set to evolve over the next decade, with 5G telecom networks, the proliferation of the Internet of Things (IoT) devices, and metaverse applications expected to drive demand for low-latency computing.

A growing number of devices, such as smartphones, will contain ML accelerators, which will have an impact on overall electricity use.

Operators Turn to Clean Energy

ICT companies are increasingly investing in renewable energy to limit their environmental impact and improve their reputation.

Amazon, Microsoft, Meta, and Google are the four?largest buyers of corporate renewable power purchase agreements (PPAs). Together, they have contracted almost 50GW of electricity to date, which is equal to the generation capacity of Sweden. Data transmission network operators typically lag behind data center operators in renewable energy use, as they operate across multiple sites, unlike large data centers.

But the ambitious scope of net zero targets will require further efforts to maximize the use of renewable power capacity and reduce emissions.

Google and Microsoft have announced zero-carbon targets for 2030, and Iron Mountain has a 2040 target. A growing number of companies are working towards zero-emission electricity use, which could encourage the deployment of a wider range of flexible technologies required to achieve the net zero transition.

Policy Solutions

While there are broader policies and regulations targeting the decarbonization of electricity supply, there are few focused on reducing the energy consumption and carbon footprint of data centers and transmission networks.

And yet demand is set to grow rapidly from the acceleration of new technologies. Increased government requirements for transparency from various industries may focus attention on data center electricity demand as it continues to increase.

The IEA report recommends several actions from policymakers and the private sector to help the industry offset demand growth to reach its emissions targets.

Policies could include regulatory and voluntary schemes to improve the energy efficiency of components such as servers, data storage, heating, ventilation, and air conditioning.

Guidance for energy efficiency in buildings could also play a role.