Views: 301
Data centers, grid operators and national decision-makers are aware of ever-growing electricity demand for the overwhelming needs and trends in digital and AI transitions. Understanding the electricity consumption of data centres is critical to draft national digital and AI-impacting energy demand. However, there are no comprehensive global assessment on the electricity consumption of data centres: the only viable source is the IEA’s global model providing some basic estimates.
Background
Contemporary, most AI models are too large to be trained and run on end-user devices like laptops and phones; hence, specific data centres provide the needed computational power for digital transition and proper AI’s functioning. National digital transition’s efforts and the work of data centres are using enormous energy volumes (mostly, electricity) to train and run AI models; and the consumption is rapidly growing. For example, gigawatt-scale clusters have emerged in specific “digital regions” in North America, Europe and Asia Pacific.
Modern “digital centers”
There are presently three types of “digital and/or AI centers”:
a) commercial, i.e. corporate-like data centres; they run by businesses or institutions for their own profitable use;
b) so-called co-location data centres, which lease space to customers to house their own computing and storage equipment and/or to provide both the space and computing equipment; and
c) hyperscale data centres, that are presently operated by major digital technology companies to meet both their own needs and/or provide cloud-based services to their customers.
The share of these centers’ electricity consumption in the world is presently almost equal, i.e. about 100 gigawatts in 2025, according to the IAE.
Recent advances in AI models are making data centres around the world larger and more power-intensive, raising the importance of the availability of national/regional electricity generation capacity and grids.
However, the existing infrastructure, decision-making policy frameworks and talent pools that enabled the top digital markets to flourish do not have adequate strategies to counter the threatening circumstances.
Thus, as the electricity consumption for digital and of AI deployment means is rising, e.g. by 2028 the US “digital-electricity demand” is projected to reach 12 percent of the national energy grid’s capacity.
Besides, numerous countries around the world are still heavily investing into digital and AI implementation’s efforts; the trend is going to increase demand for the “digital/AI electricity” demand.
International energy’s accounts
The IEA’s Energy and AI Observatory provide up-dated information on the growing links between the energy and AIs deployment, New and fast-moving progress in digital and AI technologies require new approaches to comprehensive analysis of the implications of AI on energy demand, as well as the AI’s effect on the energy sectors’ efficiency, resilience and competitiveness.
The IEA has initiated development of a first of its kind platform contributing to analysis of global insights into energy industry and digital technology sectors’ consumption.
For example, data centers’ demand in some global regions/states (as a share of total power consumption, with more than 2 percent) reveal the following picture: the US -8.9; UK -5.1;
the EU – 4.8; Australia and New Zealand – 3.7; Japan and Korea -3.2; and China -2.3.
Bottom-line: – Data centers make up 8.9% of the total power share in America, based on analysis from the IEA; – With 53.7 gigawatts of installed data center capacity, the US leads globally by a wide margin; – In the UK, 5.1% of total electricity demand is driven from data centers as billions in foreign investment flows into the country.
As to the installed data centers’ electricity capacity (in gigawatt, GW) in the world, there are four main regions: United States with 53,73; China with 31,90; European Union with 11,92; and Japan and South Korea with 6,58.
Source and citations from: https://www.iea.org/data-and-statistics/data-tools/energy-and-ai-observatory?tab=Energy+for+AI; and https://www.visualcapitalist.com/how-much-power-do-data-centers-use/