Data centres: must we choose between digital tech and ecology?

In Europe, data centres are becom­ing hubs of digit­al devel­op­ment; driv­en by the wide­spread adop­tion of cloud tech­no­logy, a rise in online ser­vices and the grow­ing integ­ra­tion of arti­fi­cial intel­li­gence into pub­lic and private ser­vices. Their pro­lif­er­a­tion is accom­pan­ied with an unpre­ced­en­ted level of invest­ment: accord­ing to Euronews, plans to set up data centres have increased by 168% in one year, focus­ing around hubs in Lon­don, Frank­furt and Par­is¹. The phe­nomen­on is linked to digit­al sov­er­eignty and eco­nom­ic and fisc­al issues, with each Mem­ber State seek­ing to attract these stra­tegic foundations.

At the same time, their impact on loc­al resources is begin­ning to show. The European Com­mis­sion estim­ates that data centres account for nearly 3% of the elec­tri­city con­sumed in the European Uni­on². In Ire­land, this share will reach 21% accord­ing to fig­ures from the nation­al energy author­ity for 2023³. Oth­er ten­sions are emer­ging, par­tic­u­larly in Mad­rid and Mar­seille, where sev­er­al pro­jects have raised con­cerns about their water con­sump­tion for cool­ing, giv­en the fre­quent droughts in the region⁴.

Faced with these changes, the European Com­mis­sion is impos­ing trans­par­ency require­ments on energy and water con­sump­tion through the Energy Effi­ciency Dir­ect­ive, to con­trib­ute to the tar­get of an 11.7% reduc­tion by 2030⁵. In response, some indus­tri­al ini­ti­at­ives are seek­ing to reduce their foot­print through heat recov­ery for dis­trict heat­ing, pass­ive cool­ing and soft­ware optim­isa­tion, among oth­er meas­ures. In Saint-Denis, for example, heat from a data centre sup­plies the urb­an net­work, while oth­er pro­jects are look­ing at link­ing it to trans­port or leis­ure sys­tems⁶.

In order to sort out the facts from the fic­tion regard­ing these inter­twined con­cerns of tech­no­lo­gic­al devel­op­ment, envir­on­ment­al con­straints and ter­rit­ori­al strategies, two research­ers offer some explan­at­ory insights. Cécile Diguet, an urb­an plan­ner at Stu­dio Dégel who headed the urb­an plan­ning and devel­op­ment depart­ment at the Par­is Region Insti­tute for nine years, co-authored Sous le feu numérique. Spa­ti­al­ités et éner­gies des data centres (Under digit­al fire: Spa­ti­al­it­ies and ener­gies of data centres), pub­lished in the journ­al Géo­car­re­four⁷. Clé­ment Mar­quet, a soci­olo­gist at the CNRS, is inter­ested in the mater­i­al­ity of digit­al infra­struc­ture. He is the author of Ce nuage que je ne saur­ais voir. Promouvoir, con­test­er et réguler les data cen­ters à Plaine Com­mune (The cloud I can­not see: pro­mot­ing, con­test­ing and reg­u­lat­ing data centres in Plaine Com­mune), pub­lished in the journ­al Tracés, in which he ana­lyses the social and polit­ic­al dynam­ics linked to data centres in Plaine Commune.

In France, the annu­al elec­tri­city con­sump­tion of data centres was estim­ated at 10 TWh in 2023. Accord­ing to pro­jec­tions by the French Elec­tri­city Trans­mis­sion Net­work (RTE), this con­sump­tion could reach 28 TWh per year by 2035. If all the pro­jects announced go ahead and data centres use their full capa­city, con­sump­tion could even rise to 80 TWh per year, or around 15% of France’s cur­rent nuc­le­ar pro­duc­tion⁸.

“The cloud relies on data centres,” points out urb­an plan­ner Cécile Diguet, “these build­ings that need to be con­struc­ted, con­sume a lot of elec­tri­city and are neces­sary to oper­ate and cool the serv­ers. This cool­ing is essen­tial because serv­ers gen­er­ate a lot of heat.” She adds: “The vocab­u­lary of digit­al tech­no­logy – “imma­ter­i­al”, “wire­less”, “dema­ter­i­al­ised” – delib­er­ately obscures the mater­i­al real­ity and impact of digit­al tech­no­logy.” Soci­ology research­er Clé­ment Mar­quet adds: “There is noth­ing imma­ter­i­al about the cloud: all data exchanges rely on phys­ic­al infra­struc­ture.” Fur­ther­more, “in the industry, it is com­mon to say that the cloud is someone else’s data centre.”

The mater­i­al dimen­sion goes bey­ond build­ings alone. Clé­ment Mar­quet points out that “the man­u­fac­ture of com­puters requires the extrac­tion of min­er­als, with destruct­ive effects on eco­sys­tems, soil, water and air.” He points out that “chip engrav­ing is very water-intens­ive” and “elec­tron­ic waste causes sig­ni­fic­ant pol­lu­tion.” These effects, although indir­ect, are struc­tur­ally linked to the func­tion­ing of the cloud.

Ter­min­als, includ­ing com­puters, tele­vi­sions, smart­phones, con­nec­ted devices, remain the largest source of digit­al energy con­sump­tion, account­ing for more than 50% of glob­al demand. Net­works and data centres share the rest: approx­im­ately one-third for the former and just under 15% for the lat­ter⁹. In 2022, data centres con­sumed between 240 and 340 TWh, or 1 to 1.3% of glob­al elec­tri­city demand¹⁰.

This dis­tri­bu­tion is chan­ging rap­idly under the influ­ence of gen­er­at­ive AI, cloud com­put­ing and stream­ing, which are sig­ni­fic­antly increas­ing the con­sump­tion of digit­al factor­ies¹¹. Clé­ment Mar­quet dis­tin­guishes between dis­trib­uted and cent­ral­ised infra­struc­ture: “Ter­min­als con­sume more, but in a dis­persed man­ner. Data centres con­cen­trate their demand, which cre­ates spe­cif­ic ten­sions on net­works.” This pool­ing causes loc­al pres­sures that are invis­ible in over­all bal­ances. Cécile Diguet points out that “digit­al tech­no­logy relies on three major infra­struc­tures: net­works (around one-third of the car­bon impact), data centres (around 40%) and equip­ment. They also con­tain hard­ware and net­works, mak­ing them a cent­ral and vis­ible infra­struc­ture, unlike net­works, which are often bur­ied underground.”

“Data centres there­fore cause loc­al­ised envir­on­ment­al prob­lems that can­not be reduced to a glob­al com­par­is­on between ter­min­als, net­works and data centres,” notes Clé­ment Mar­quet. “The rap­id rise of gen­er­at­ive AI is exacer­bat­ing this pres­sure: these infra­struc­tures, which are very energy-intens­ive but less depend­ent on the net­work, tend to be scattered across rur­al areas. Their size and intens­ity nev­er­the­less pose many challenges.”

The loc­a­tion of data centres is primar­ily based on eco­nom­ic cri­ter­ia: tax­a­tion, energy costs, avail­able land, polit­ic­al sta­bil­ity and leg­al cer­tainty. Envir­on­ment­al argu­ments, often high­lighted in com­mu­nic­a­tion strategies, are not decis­ive at European or glob­al level. Clé­ment Mar­quet points out that “attract­ive­ness policies play an import­ant role,” cit­ing North­ern Vir­gin­ia as an example. Dub­lin and Ams­ter­dam have benefited from favour­able tax­a­tion, and Spain is pos­i­tion­ing itself thanks to act­ive sup­port from the pub­lic authorities.

In France, cer­tain meas­ures are enhan­cing attract­ive­ness: “For loc­al author­it­ies, the prop­erty tax is what makes these infra­struc­tures so attract­ive,” explains Cécile Diguet. “For inter-muni­cip­al com­munit­ies, the bene­fits are lim­ited. It is mainly the State that inter­venes through reduc­tions in elec­tri­city taxes, in par­tic­u­lar the TICFE, which is an import­ant lever for operators.”

How­ever, this attract­ive­ness does not guar­an­tee coher­ent loc­al integ­ra­tion. Clé­ment Mar­quet notes that “these policies often neg­lect land use plan­ning: they take little account of integ­ra­tion or loc­al con­di­tions.” They may also be accom­pan­ied by spe­cif­ic statuses (crit­ic­al archi­tec­ture, nation­al interest), inhib­it­ing envir­on­ment­al require­ments. Ire­land is a case in point: its favour­able tax regime has attrac­ted many centres, but their con­cen­tra­tion has sat­ur­ated the net­works, lead­ing to a morator­i­um in Dub­lin in 2021¹². In the Neth­er­lands, some pro­jects have been sus­pen­ded due to their excess­ive impact on land and energy.

Data centres host essen­tial ser­vices in major sec­tors such as health­care, defence, gov­ern­ment and fin­ance. Their secur­ity has become a mat­ter of digit­al sov­er­eignty and nation­al secur­ity, with their pro­tec­tion con­sidered a pri­or­ity for eco­nom­ic and social sta­bil­ity, “espe­cially when they host sens­it­ive data (health, gov­ern­ment depart­ments, mil­it­ary),” says Cécile Diguet. “Some are clas­si­fied as OVIs. The mil­it­ary often uses its own data centres and net­works. Data centres are essen­tial to guar­an­tee digit­al sov­er­eignty, and the state can cre­ate them to secure stra­tegic data.”

The des­ig­na­tion of an oper­at­or of vital import­ance (OVI) is inten­ded to pro­tect these crit­ic­al facil­it­ies. In Europe, gov­ern­ments are seek­ing to secure stra­tegic data in a con­text of grow­ing threats, par­tic­u­larly with the rise of gen­er­at­ive AI. But Clé­ment Mar­quet points out that “this status is often high­lighted, par­tic­u­larly in the con­text of com­pet­i­tion around AI,” adding that “this has promp­ted coun­tries such as the United King­dom, and soon France via Art­icle 15 of the Eco­nom­ic Sim­pli­fic­a­tion Act, to recog­nise data centres as struc­tures of nation­al import­ance.” This recog­ni­tion has con­crete effects: “On the power of loc­al elec­ted offi­cials, land arti­fi­cial­isa­tion and envir­on­ment­al pro­ced­ures.” This is a dif­fer­ent logic from that of OVIs, which are ‘linked to the pro­tec­tion of state data and more mod­est projects.’

The European energy mix, such as nuc­le­ar power in France and hydro­elec­tri­city in Sweden, may seem con­du­cive to redu­cing car­bon foot­prints. But this does not guar­an­tee a lower over­all impact. Oth­er factors come into play, such as facil­ity design, thermal man­age­ment, use of mater­i­als and rap­id growth in digit­al usage.

Cécile Diguet high­lights the lim­it­a­tions of this reas­on­ing: “There are major dif­fer­ences between coun­tries: France high­lights its use of nuc­le­ar power, but this also has its own impacts, par­tic­u­larly on waste. Gains in energy effi­ciency are quickly off­set by increased usage. Great­er effi­ciency does not neces­sar­ily mean lower con­sump­tion. At the loc­al level, data centres can com­pete with oth­er elec­tri­city needs, such as trans­port or industry. The impact is there­fore also felt at the region­al level.”

Energy optim­isa­tion cer­tainly leads to sav­ings, but these are often off­set by grow­ing demand linked to tech­no­lo­gies such as gen­er­at­ive AI. This rebound effect partly can­cels out the envir­on­ment­al bene­fits. Resource use also remains an issue, regard­less of the energy source¹³.

Clé­ment Mar­quet shares this view: “The energy­so­urce mix is cer­tainly an import­ant factor, but it var­ies from coun­try to coun­try. It is not enough to guar­an­tee a low impact. The envir­on­ment­al impact of digit­al tech­no­logy is not lim­ited to CO₂ emis­sions: it also includes resource con­sump­tion, water, waste, etc.” Fur­ther­more, “the grow­ing needs of digit­al tech­no­logy, par­tic­u­larly in rela­tion to gen­er­at­ive AI, are adding to oth­er low-car­bon elec­tri­city needs (mobil­ity, industry, heat­ing). This com­pet­i­tion may slow down the energy trans­ition and pre­vent the decar­bon­isa­tion of oth­er sec­tors at the loc­al level.’

Aicha Fall