Increasing use of artificial intelligence (AI) is boosting the electricity demand of data centres. AI may have much more widespread impacts on the energy system over time, however, affecting both energy supply and also economy-wide energy demand
Energy demands of AI
Growing power demand from data centres, driven in large part by rising use of AI, is likely to provide a material boost to electricity demand, albeit to differing degrees in different regions and countries.
In Current Trajectory, growth in data centres’ use of electricity accounts for around a tenth of global power demand growth out to 2035. But that impact differs substantially across regions: for example, rising data centre demand accounts for 40% of overall US power demand growth over the next decade (see Power sector).
Any such projections of the power demand of data centres are highly uncertain, however. In part their demand will depend on the evolution and rate of adoption of AI. But it will also depend crucially on the energy efficiency of data centres, which has risen dramatically over recent years: digital data traffic rose more than 25-fold between 2010 and 2024, but the energy use of data centres rose only two-fold over that period.
While some of the drivers of those past efficiency improvements – for example the widespread shifts from on-premises data centres to the cloud – may not have much further to run, other innovations, including continuing advances in chip design and AI programming, are likely to drive more improvements in the energy efficiency of data centres in the coming years.
Impacts of AI on the energy sector
The demand for electricity of data centres is only one narrow aspect of the likely impacts of AI on the global energy sector, however. AI may over time have significant impacts on the supply of energy, too.
AI is already being widely used in the oil and gas industry, for example improving and accelerating exploration through better analysis of geological structures. It is also being used to plan and design new oil and gas wells, improving the subsequent operation and efficiency of those facilities.
AI also has the potential to accelerate innovation in low carbon energy, for example through the development of new materials for solar panels or carbon capture, of new battery chemistries, or through improving the design and efficiency of low carbon hydrogen systems. More radically, continued advances in AI could unlock major technological breakthroughs in low carbon energy supply, such as through the development of new advanced biofuels, or even of usable nuclear fusion.
Accelerating use of AI could also improve the efficiency with which energy systems are operated. It has the potential to enable more efficient planning and operation of electricity grids, helping to forecast demand spikes and optimize battery storage deployment. It could help to make grids ‘smarter’, enabling more efficient aggregation of large quantities of small assets such as EVs, rooftop solar and smart thermostats. And it is already being used in both fossil fuel facilities and power systems to better predict faults or failures, improving safety and reducing downtime or the need for additional backup capacity.
Wider impacts of AI on the global economy and the energy system
Such effects of AI on energy supply are, however, only energy sector-specific examples of the wider potential impacts of the technology on the global economy, with broader-still implications for the energy system.
AI could materially boost the growth of the global economy if it leads to faster productivity growth. That could occur in a wide range of ways, including the automation of an increasing range of tasks, more efficient use of physical assets, and the acceleration of innovation and new discoveries. There is currently an extremely wide range of estimates of the likely size of these effects. One recent OECD survey1 reported estimates varying from only moderate impacts – boosting the level of US GDP by only around 1% over the next decade – to effects as large as an additional 2.5pps per year on productivity growth in the US.
AI-driven improvements in productivity growth could have very significant implications for energy demand. The average of the estimates from the OECD survey – higher productivity growth of around 1.2% per year – would, if manifested at a global level, boost total energy demand by around 15% by 2035, assuming that global energy efficiency continued to improve at around its past average rate. That is twenty times as large as the increase in data centres’ power demand in Current Trajectory.
It is probably unrealistic to assume that the energy efficiency of the global economy would itself be unaffected by major advances in AI, however. AI could help to optimise manufacturing processes and accelerate the development of more energy-efficient products, leading to substantial improvements in industrial efficiency. In the transport sector, AI may be used to better manage traffic and optimise routes, and it could also bring about substantial reductions in the use of energy for heating and cooling in buildings. Recent IEA analysis2 suggests that wide application of AI to improve economy-wide energy efficiency could have very substantial impacts on global energy demand.
Assumptions about the effects of AI in Current Trajectory and Below 2°
Current Trajectory and Below 2° both assume only a moderate boost to productivity and GDP growth from the use of AI over the outlook. Moreover, the scenarios do not explicitly incorporate significant AI-driven technological breakthroughs in energy supply. But given the rapid pace of developments in the design and use of AI applications, any estimates of these AI effects are, at present, enormously uncertain, and their impacts could be far larger. The uncertainties around their eventual size dwarf those solely around the future power needs of data centres.