Font size:
Print
Global Energy Demand from Data Centres
IEA Report the Need for AI Reshaping Global Energy Demand
Context: The International Energy Agency (IEA) has released a critical new report detailing how artificial intelligence (AI) and data centres are reshaping global energy demand.
Energy Consumption Projections
- By 2030, global data centres could consume around 945 terawatt-hours (TWh) of electricity.
- This would exceed the current electricity consumption of Japan, a major global economy.
- As of 2024, data centres use approximately 415 TWh, accounting for 1.5% of global electricity consumption.
- Electricity use in data centres has grown at 12% annually over the past five years.
AI: The Major Driver
- Artificial Intelligence is significantly driving the surge in energy demand.
- “AI is one of the biggest stories in the energy world today,” said Fatih Birol, Executive Director, IEA.
- Data centres play a pivotal role in training and deploying AI models, requiring significant infrastructure and energy.
- The United States alone could account for nearly 50% of the projected growth in electricity demand from data centres over the next five years.
Data Centres: Powering the Digital Age
- Modern data centres — the backbone of AI and cloud services — house vast arrays of computing infrastructure, including:
Current Energy Sources (2024)
- Coal: 30% of electricity for data centres (largest share).
- Renewables: 27%.
- Natural Gas: 26%.
- Nuclear Energy: 15%.
- Servers (which handle processing and storage): ~60% of energy consumption
- Storage systems (centralised data storage and backup): ~5%
- Networking equipment (routers, switches): ~5%
- Cooling and environmental control systems: 7–30%, depending on energy efficiency
- A conventional data centre typically uses 10–25 megawatts (MW) of electricity, while hyperscale, AI-focused facilities can exceed 100 MW, consuming as much energy annually as 100,000 households. These hyperscale centres are expanding rapidly to accommodate the rising demand for training and deploying larger AI models.
Future Energy Mix
- By 2030, renewables could supply nearly 50% of new demand.
- Followed by natural gas and coal.
- Nuclear energy, especially Small Modular Reactors (SMRs), could play a greater role post-2030.
SMRs: A Game-Changer?
- SMRs are advanced nuclear reactors with capacities up to 300 MW(e) per unit.
- The first SMRs may be commissioned in the US after 2030.
- Tech companies have planned over 20 GW of SMR capacity so far.
- Maria Basso, Head, AI Applications & Impact at the WEF, notes:
- SMRs offer efficient energy generation and can reduce reliance on fossil fuels.
- They are effective during intermittency in renewables like solar and wind.
- Challenges remain around cost, safety, and regulations.
Risks and Concerns
- Cybersecurity: Increased digitalisation and electrification raise risks.
- Energy security: Data centres need critical minerals — many are sourced from limited global suppliers.
- Gallium, a key mineral, sees:
- 95% of refining in China.
- 10% of supply may be needed for data centres by 2030.
- High risk of supply shocks due to weather, accidents, trade, or geopolitics.
The AI Energy Paradox
- While AI drives high electricity demand, it can optimise grid operations and reduce emissions.
- AI may: Enhance efficiency in the energy sector. Accelerate innovation in batteries and solar technology.
Way Forward
- Environmental considerations must be embedded in AI development from the start.
- Focus on valuable AI applications, avoiding energy-intensive uses with minimal societal benefit.
