Marc Garner Global President, Cloud &
Service Providers Schneider Electric
Q. HOW ARE GRIDS EXPANDING TO KEEP PACE WITH RAPID AI-DRIVEN DEVELOPMENT?
ยป The rise of vertical AI is likely to see reduced energy intensity in some instances as the models are expected to be smaller than current LLMs, but it is also expected that vertical AI is likely to be more distributed, leveraging more edge deployments, potentially creating different energy hotspots.
Energy demand globally will likely rise beyond 30,000TWh by 2030, with a CAGR of 3.6 %. Different regions and countries are experimenting with new models to cope with demand growth, which can sometimes be highly localised. Spain has conducted grid connection auctions to gauge interest and generate revenue, while France has carried out extensive anticipatory planning, leading to a new national strategy to 2035. The UK has designated AI growth zones, giving more control and targeted development.
Large energy users, including data centres, are experimenting with a combination of storage and renewables, serving as a useful reference for grids doing similar on a larger scale. This means that the future is not renewables versus core generative capacity, such as that from natural gas or nuclear, but a hybrid system of all these elements that ensures reliable baseload performance under all conditions. There have already been instances of hyperscalers reviving dormant or underutilised nuclear plants, such as Amazon with Talen Energy in Pennsylvania, and Microsoft with Constellation Energy and the Three Mile Island plant.
datacentremagazine. com 71