How is AI improving the efficiency of smart electricity grids?

AI forecasting models predict renewable generation and demand spikes up to 72 hours ahead, enabling grid operators to pre-schedule storage discharge, reduce curtailment of wind and solar by up to 30%, and defer costly peaker plant starts.

What is a small modular reactor (SMR) and when will they power data centres?

An SMR is a nuclear reactor below 300 MWe built in a factory and shipped to site. Vendors like NuScale, Rolls-Royce, and X-energy target first commercial operation in 2028-2030. Microsoft, Google, and Amazon have signed offtake agreements for SMR output.

What is green hydrogen and how is AI accelerating its production?

Green hydrogen is produced by electrolyser stacks powered by surplus renewable electricity. AI optimises electrolyser scheduling, predicts renewable surplus windows, and manages storage logistics to reduce the levelised cost of hydrogen below $2/kg by 2030.

What are solid-state batteries and why do they matter for AI infrastructure?

Solid-state batteries replace liquid electrolyte with a solid material, enabling higher energy density (500 Wh/kg vs 250 for Li-ion), faster charge rates, and near-zero fire risk - making them ideal for data-centre UPS and EV fleets supporting mobile edge nodes.

How does AI help with carbon capture and net-zero targets?

AI models optimise solvent formulations, predict CO2 breakthrough in DAC units, and schedule capture operations around cheap renewable electricity windows, reducing the energy penalty of carbon capture from 30% to under 15% of plant output.

What is the carbon footprint of large AI models?

Training a frontier LLM consumes approximately 500-1,000 MWh of electricity, emitting 50-200 tonnes of CO2e depending on grid mix. Hardware efficiency gains (sparsity, distillation) and renewable procurement are key mitigations.