ABSTRACT
We investigate the power system and climate policy implications of the anticipated expansion in electricity demand from datacentres. We perform a joint optimization of Generation and Transmission Expansion Planning for the island of Ireland considering uncertainty in future datacentre growth under various climate policies. Datacentre expansion imposes significant extra costs on the power system. A renewable energy target is more costly than a technology-neutral carbon reduction policy due to the restriction to invest solely in renewable energy, and the divergence in costs increases non-linearly in electricity demand. A carbon reduction policy is more robust to uncertainties in projected demand than a renewable policy. The increased demand from datacentres, which is relatively constant, is best matched by a power supply with constant output, generally fossil-fuelled power plants, potentially in combination with carbon capture and sequestration technology, than variable supply. Datacentre demand increases power system costs by 6% under a carbon reduction policy, and by 9% to 15% under renewable energy targets. However, the abatement cost of carbon under renewable targets decreases at higher levels of demand, while the opposite is true for the abatement cost under technology-neutral targets. High renewable targets crowd out other low-carbon options, namely CCS, leading to a higher cost solution for the same reduction in emissions, although this result is sensitive to increases in fossil fuel prices. The results suggest that climate-related energy policy should focus on technology-neutral carbon reduction policies. The climate impacts of datacentres on energy demand and thus supply requirements should also feature in climate and energy policy discussions.
Key policy insights
Climate policy that focuses on renewable energy only is more expensive than a technology-neutral climate policy
The costs of a renewable-only focused climate policy increase non-linearly as demand increases
Disclosure statement
No potential conflict of interest was reported by the author(s).
Notes
1 Note that none of our scenarios considers the extreme case of the highest datacentre demand projected by EirGrid occurring with 100% probability. We essentially consider the probability of this eventuality to be so low as to render its examination unnecessary, based on the historical ratio of planned to actual investment by large industrial electricity users. However, should this level of datacentre investment take place, the power system impacts will be even greater than those modelled here.
2 We assume intermediate 2025 targets of 47% in the first case and 55% in all other cases