AI and Energy

AI and energy refers to the two-way relationship between artificial intelligence and the electricity system. On one side sits demand: training and running large models requires dense clusters of accelerators that draw large, continuous loads, making data centres one of the fastest-growing sources of electricity consumption in the world. On the other sits optimisation: AI techniques are increasingly used to forecast demand, integrate renewable generation and operate grids more efficiently.

Demand and Projections

The International Energy Agency's landmark Energy and AI report, published in April 2025, estimated that data centres consumed about 415 terawatt-hours (TWh) of electricity in 2024 — roughly 1.5% of global consumption — and projected in its base case that the figure would double to around 945 TWh by 2030, approaching 3%.^[1][2] A follow-up assessment found that data centre demand grew 17% in 2025 to about 485 TWh, with electricity use at AI-focused facilities rising 50%; the updated outlook again points to roughly 950 TWh by 2030.^[3]

Growth is highly concentrated: the United States and China account for nearly 80% of the projected increase to 2030, and AI data centres are geographically clustered, so their local impact can far exceed their global share.^[1] The IEA notes that a single AI-focused facility can draw as much power as a power-intensive factory such as an aluminium smelter. Academic analysis published in 2026 estimated that the six largest data centre operators could consume between 239 and 295 TWh by 2030, equivalent to about 1% of projected world electricity demand.^[6]

Technology and Drivers

AI workloads differ from conventional computing in their power profile. Training runs load accelerators at high utilisation for weeks at a time, while inference creates steep, variable demand, and modern AI racks draw tens of kilowatts each, pushing operators toward liquid cooling and reshaping data centre design.^[3] Energy per query has fallen dramatically as models and hardware improve, but total consumption keeps rising as usage grows — a rebound effect that makes absolute demand, not efficiency alone, the central planning question.^[3] Meeting the load involves long-term renewable power purchase agreements, nuclear restarts and small modular reactor proposals, and new gas generation; the IEA has estimated that backing variable data centre load with on-site gas can require overbuilding generation infrastructure by 30–70%.^[3]

Impact and Policy

The surge is stressing grids that were planned for slower growth, and in several markets regulators have introduced special tariffs, connection requirements and efficiency metrics. Analysts warn that focusing only on ratios such as power usage effectiveness can amount to greenwashing if absolute consumption still rises.^[5] At the same time, AI is being deployed for weather forecasting, grid optimisation and industrial efficiency — among the clearest examples of the technology's potential to reduce energy use elsewhere.

>See Also

• AI Data Centres in Malaysia
• Green AI
• Sovereign AI
• GPU Cluster
• National AI Roadmap Malaysia