Story of the month
DATA CENTERPOWER CONSUMPTION
40% of global data centers are located in North America, mainly in USA, half of which being located in few states. While Western Europe accounts only for 26% of this global distribution as of today, one could expect growth in Europe as the electrification and Artificial Intelligence applications will also be developed here…
DATA CENTER AND DATA SOVEREIGNTY
In short, a data center is a centralized repository for the storage, management, and dissemination of data and information. More precisely it relies on “white” space including servers, storage, AC units, etc., and “grey” space where back-end infrastructure as transformers, chillers, or generators are located.
Most of the data center are currently into unregulated markets enabling to connect to nuclear plants power sources as well as benefiting from easier access to building permits. Data centers denomination pertains hyperscale data centers, supporting large scale cloud applications, multi-tenant “colocation” center with multiple enterprise into one location and even enterprise data centers, smaller-scale facilities.
We expect double-digit mid-term growth forecasts supported by strong data generation outlook (27% CAGR by 2027E according to IDC). Moreover, the lowering vacancy rate to 2-3% across US & Europe further testifies for this structural increasing demand with additional capacities being absorbed with virtually no lag.
Data sovereignty considerations is to trigger additional demand for localized data centers, especially in European countries with data being subject to the laws and regulations of the country in which it is located.
AFFORDABLE POWER PRICES AND FISCAL INCENTIVES TO FOSTER EUROPEAN DYNAMIC
As European power demand is down 10% since 2008 consumption peak, to what extent could data centers and specifically AI data centers contribute to a reversal?
AI data centers are proven to be significantly more energy-intensive than traditional ones. This arises notably from the increasing implementation of GPUs into servers to aid in workload acceleration for AI and machine learning heavy processes.
To that extent, AI servers are especially energy- intensive during their training phase. A ChatGPT query is indeed 10 times more power intensive than a traditional Google search (from 0.3 Wh to 2.9Wh).
Data centers’ power demand growth (8-14% CAGR, according to GS) is poised to be especially strong in countries with affordable and abundant baseload power like Spain, France, and the Nordics (through solar, nuclear, hydro output) and in countries where large financial services and tech companies operate, willing to offer incentives (tax breaks e.g.) such as Ireland, UK, and Germany.
However, with data centers’ representing only 1% of 2023 global power demand, one could argue that bringing back European power demand to growth is to be driven not only by data centers dynamic but primarily by macro-trends like further pick-up in electrification and developments in AI…
