Subscribe
The energy appetite of the world's data centres could reach around 945 terawatt-hours (TWh) by 2030, according to a new report from the International Energy Agency (IEA). This will be slightly greater than the energy requirement of the entire country of Japan.
Currently, the electricity generation to power data centres is estimated to be around 415 terawatt hours (TWh), which amounts to about 1.5 per cent of the global electricity consumption in 2024. It has grown at 12 per cent per year over the last five years.
The increase is driven by artificial intelligence (AI). “AI is one of the biggest stories in the energy world today — but until now, policymakers and markets lacked the tools to fully understand the wide-ranging impacts,” IEA Executive Director Fatih Birol said in a statement.
Birol added that the global electricity demand from data centres is set to more than double over the next five years, concentrated in some countries. For instance, data centres in the United States could account for almost half of the growth in electricity demand.
Data centres provide infrastructure for training and deploying AI models. These facilities house servers (computers that process and store data and account for 60 per cent of the energy demand), storage systems (devices for centralised storage and backup, account for around 5 per cent of electricity consumption), networking equipment (consists of switches to connect devices, routers to direct traffic, which also make up 5 per cent) and associated components like equipment needed to regulate temperature and humidity to keep IT equipment running at optimal conditions. The electricity consumption ranges from 7-30 per cent, depending on the efficiency levels.
A conventional data centre consumes around 10-25 megawatts (MW) of electricity, while a hyperscale, AI-focused data centre can have a capacity of 100 MW or more, consuming as much electricity annually as 100,000 households, the report stated. Data centres that are AI-focused are increasing in size to accommodate larger models that come with the growing demand for AI services.
Currently, coal makes up the largest source of electricity, with a share of about 30 per cent, followed by renewables, contributing 27 per cent of the electricity consumed by data centres globally. The third and the fourth largest sources are natural gas and nuclear energy, meeting 26 per cent and 15 per cent of the demand.
Over the next few years, the landscape is likely to change, with renewables meeting nearly half of the additional demand, followed by natural gas and coal. Nuclear could begin playing an increasingly important role towards the end of this decade. Small Modular Reactors (SMR) could enter the energy mix post 2030. SMRs are advanced nuclear reactors that have a power capacity of up to 300 megawatt-equivalent MW(e) per unit.
The first SMRs are expected to be commissioned in the US after 2030. Technology companies have plans to finance more than 20 gigawatts (GW) of SMRs to date.
SMRs are likely to increase efficiency. “While SMRs may not directly reduce energy demand, they can contribute to lowering overall system energy use through more efficient energy generation and distribution,“ Maria Basso, head, AI Applications and Impact, C4IR Digital Technologies, World Economic Forum, had told Down To Earth.
They can also reduce the need for fossil fuel-based backup power. “For example, when renewables like wind or solar experience intermittency or low-generation periods, SMRs can provide consistent energy without defaulting to inefficient fossil fuel options,” Basso explained, adding that SMRs still remain a developing technology, and their cost, safety and regulatory challenges are being assessed.
The report also warned of cybersecurity concerns with increased electrification, digitalisation and connectiveness of the energy sector.
There are also concerns about energy security as building data centres require critical minerals, which are concentrated in the hands of a few.
China accounts for 95 per cent of gallium refining, leading to significant vulnerabilities to supply shocks when supply from large producers is disrupted due to extreme weather events, industrial accidents, trade disruptions or geopolitics. This is worrying, given that data centre demand for gallium could reach up to 10 per cent of the present-day supply by 2030.
AI has also created an energy paradox. While the technology has a huge energy demand, it could also be used to reduce emissions by optimising the operation of electricity grids and thus making the energy sector more efficient.
AI could potentially offset emissions reductions if adoption of the technology is widespread, according to the report. It could speed up innovation in energy technologies such as batteries and solar energy.
“AI is a tool, potentially an incredibly powerful one, but it is up to us — our societies, governments and companies — how we use it,” Birol said.
The key, Basso added, is to integrate environmental considerations into AI development from the start — while minimising AI applications that add no value to our economy, industries and society.
