“Low-carbon computing is possible!”

Paul Benoit and Miroslav Sviezeny found­ed Qarnot Com­put­ing in Decem­ber 2010. Their idea was to use warmth emit­ted by com­put­er servers, known as “waste heat”, to heat entire build­ings. Rather than recov­er­ing this heat and trans­port­ing it some­how, they chose to trans­form the servers them­selves into “radi­a­tors” that can be installed direct­ly on premis­es. To do so, a cen­tralised dig­i­tal plat­form sends com­put­er cal­cu­la­tions – finan­cial cal­cu­la­tions, 3D imag­ing etc. – to the servers and brin­ing the results back togeth­er to recon­sol­i­date them. These “com­put­er radi­a­tors”, called QH‑1, are equipped with a ther­mo­stat and a switch, allow­ing them to be adjust­ed on demand.

Three years on, Qarnot test­ed its first mod­els in homes Paris, fol­lowed by a new build in Bor­deaux. Next, they moved on to fundrais­ing, pres­ti­gious clients, the devel­op­ment of a dig­i­tal boil­er called QB‑1, and a part­ner­ship with the Casi­no group to heat their ware­hous­es. Today, the com­pa­ny employs around 70 peo­ple, includ­ing 50 engi­neers who devel­op new “eco­log­i­cal heat” applications.

Qarnot says it sells “eco­log­i­cal heat”. Can you explain what that is?

Quentin Lau­rens. First and fore­most, Qarnot pri­mar­i­ly sells a com­put­ing ser­vice, but we do make use of fatal heat that is a con­se­quence of the activ­i­ty of servers. After cre­at­ing the “com­put­er radi­a­tor”, i.e. heat­ing a build­ing with com­put­ing pow­er, Qarnot extend­ed its range with a dig­i­tal boil­er that pro­duces hot water at over 60°C. As a term to define our prod­ucts and the dif­fer­ent mar­kets they address, we now say that we pro­vide low car­bon com­put­ing. This means that we pro­duce heat or hot water from the tem­per­a­ture released by com­put­er proces­sors. Our exper­tise lies large­ly in the plat­form that dis­trib­utes the cal­cu­la­tions to be per­formed in a way that is total­ly trans­par­ent to the occu­pants of the premis­es where the devices are installed.

Until now, our activ­i­ty has main­ly been of inter­est to two types of users: large con­sumers of com­put­er pro­cess­ing such as banks, car­toon stu­dios, local author­i­ties or social hous­ing land­lords, and ener­gy com­pa­nies that install our devices in their build­ings. But now there are new sec­tors that are begin­ning to take inter­est, par­tic­u­lar­ly in fields like arti­fi­cial intel­li­gence and machine learn­ing, flu­id mechan­ics soft­ware and med­ical research.

Would you say that eco­log­i­cal heat makes dig­i­tal tech­nol­o­gy “green­er”?

Be care­ful, we are not say­ing that we are zero car­bon or zero emis­sions. How­ev­er, we do that our tech­nol­o­gy is more than just anoth­er low car­bon solu­tion – our mod­el is a real par­a­digm shift. For a data cen­tre, you have to con­struct a build­ing, sup­ply it with ener­gy for oper­a­tions and to cool the servers, con­nect it to inter­na­tion­al net­works, install a gen­er­a­tor with fuel, etc. How­ev­er, the Qarnot mod­el removes the need for an exter­nal data cen­tre and its sub­se­quent cool­ing. The only thing need­ed is an ener­gy source for the pro­cess­ing pow­er, which pro­duces heat, and a fibre optic con­nec­tion. We try to opti­mise the life of servers and micro­proces­sors by using ‘peak pow­er’ (the max­i­mum pow­er of a device) ensur­ing that good per­for­mance and heat emis­sion are com­bined. In doing so, we esti­mate that we reduce CO₂ emis­sions by 81% com­pared to a tra­di­tion­al data cen­tre mod­el because we don’t have any neg­a­tive exter­nal­i­ties, it min­imis­es the effects and impacts.

How did you cal­cu­late this reduc­tion in emis­sions?

Today, we do not have a pre­cise cal­cu­la­tion, so it is rather an esti­mate that we have cal­cu­lat­ed our­selves. We have just launched the devel­op­ment of a cal­cu­la­tion method that allows us to go to a great degree of finesse in mea­sur­ing the car­bon foot­print, by inte­grat­ing each of the para­me­ters (car­bon foot­print of the source of elec­tric­i­ty depend­ing on the coun­try, mate­r­i­al used, rate of heat recov­ery, etc.). With this mod­el, we will be able to pre­cise­ly cal­cu­late con­sump­tion and quan­ti­fy green­house gas emis­sions of a com­put­er cal­cu­la­tion or an activ­i­ty. This method­ol­o­gy is cur­rent­ly being cer­ti­fied. As such we will be able to esti­mate the cost of com­put­ing in a data cen­tre, con­sid­er­ing the dif­fer­ences in elec­tric­i­ty prices and emis­sions across France, which pro­duces nuclear ener­gy, and oth­er coun­tries that pro­duce elec­tric­i­ty in coal-fired pow­er stations.

What are the pos­si­bil­i­ties for devel­op­ing eco­log­i­cal heat?

We recent­ly devel­oped and improved our QB‑1 boil­er, and we are cur­rent­ly con­duct­ing tests with the Open Com­pute Project to recov­er servers from large data cen­tres when they renew their fleets and recon­di­tion them into boil­ers. For the moment, we are main­ly sell­ing com­put­ing or heat­ing ser­vices. In the near future, we plan to sell the boil­er as a stand-alone ser­vice, i.e. a cus­tomer could buy boil­ers and use them for both com­put­ing and heat­ing water.

On the user side, there are many mar­kets that are inter­est­ed and with whom we are devel­op­ing projects. Farm­ers are test­ing the Qarnot solu­tion to heat green­hous­es to help pro­duce toma­toes and straw­ber­ries all year round. Brew­eries are con­sid­er­ing using it to heat water. Heat­ing net­works, par­tic­u­lar­ly in north­ern Euro­pean coun­tries, are increas­ing­ly inter­est­ed in recov­ered ener­gy. We already have projects with sev­er­al cities. There is no short­age of projects. All these chal­lenges are excit­ing for our engi­neers because they respond to their envi­ron­men­tal con­cerns. Eco­log­i­cal heat is still in its infan­cy, there is much yet to come.