Semiconductors: can European industry regain ground?

The glob­al mar­ket for elec­tron­ic com­pon­ents will be worth $625bn in 2024. Dur­ing the Cov­id-19 pan­dem­ic and the res­ult­ing short­age of chips, Europe was reminded of its depend­ence on Asia and the United States. In an attempt to lim­it this depend­ence, the European Uni­on announced the launch of the CHIPS and Sci­ence Act in Feb­ru­ary 2022, which aimed to stim­u­late European pro­duc­tion. Some three years later, how is the sec­tor faring in Europe? We take stock with Sébas­tien Dauvé, dir­ect­or of CEA-Leti, who has just launched the FAMES pilot line in Gren­oble, fun­ded by the European Uni­on and France.

Sébas­tien Dauvé. After a decline in 2023, the mar­ket has been grow­ing since 2024, but this growth masks a two-speed evol­u­tion. The mature semi­con­duct­or sec­tor, such as micro­con­trol­lers, which were in short sup­ply for industry and the auto­mot­ive sec­tor in 2022–2023, is now sat­ur­ated. At the same time, we are see­ing an explo­sion in the mar­ket for very advanced node com­pon­ents (less than 5 nm), stim­u­lated by very strong demand for chips for data centres and arti­fi­cial intel­li­gence applic­a­tions: graph­ics pro­cessing units (GPUs) and high band­width memory (HBM). The invest­ments in these fields are stag­ger­ing: Taiwanese com­pany TSMC, for example, has announced that it will invest $100bn in the United States over the next four years.

The invest­ments required to deploy new indus­tri­al resources are such that we can­not ima­gine a major change in the short term. It is there­fore still char­ac­ter­ised today by a very strong inter­de­pend­ence on a glob­al level: a com­pon­ent can be designed on one con­tin­ent and pro­duced on a second, while the raw mater­i­als are sup­plied by a third. The United States, for example, excels in the design of integ­rated cir­cuits. Japan has taken the lead in the pro­duc­tion of wafers (the semi­con­duct­or wafers on which elec­tron­ic com­pon­ents are prin­ted) and pro­cess gases, and China is unavoid­able for the sup­ply of rare earths. Taiwan and South Korea, through the TSMC and Sam­sung foundries, dom­in­ate chip pro­duc­tion – TSMC is even the only one to mas­ter the most advanced nodes (2 nm), which are highly innov­at­ive and in great demand today.

Europe accounts for just under 10% of glob­al semi­con­duct­or pro­duc­tion, thanks to man­u­fac­tur­ers such as the French com­pany STMi­cro­elec­tron­ics, which ranks around 10^th in the world. Although it does not have the capa­city to pro­duce advanced nodes, it is rather well posi­tioned in the design and pro­duc­tion of ‘More than Moore’ com­pon­ents, con­sist­ing of sensors, imagers, power and tele­com com­pon­ents, and micro­con­trol­lers. This class finds applic­a­tions in vari­ous sec­tors of activ­ity, such as auto­mot­ive, industry, defence and health. The con­tin­ent also has near-mono­pol­ies in spe­cif­ic fields: the Dutch com­pany ASML, for example, is the only play­er to have mastered the man­u­fac­ture of EUV advanced litho­graphy equip­ment, which is essen­tial for foundries.

Finally, Europe bene­fits from act­ive research and sig­ni­fic­ant innov­a­tion cap­ab­il­it­ies, par­tic­u­larly through its RTOs (Research and Tech­no­logy Organ­isa­tions, includ­ing CEA-Leti, the Bel­gian IMEC, the Ger­man Fraunhofer, the Finnish VTT, etc.), a unique organ­isa­tion­al mod­el cap­able of driv­ing innov­a­tion from the earli­est stages of research through to pre-industrialisation.

We knew that this object­ive was very ambi­tious… I would say that in the short and medi­um term, it is more a ques­tion of main­tain­ing our pos­i­tion in the cur­rent mar­ket and retain­ing our sov­er­eignty over the most stra­tegic devel­op­ments when we have the means to do so: in par­tic­u­lar, those relat­ing to defence and cyber­se­cur­ity, but also quantum com­put­ing, in which Europe is mak­ing good progress.

This is bad news for Europe, because it would be in our interest to have more play­ers estab­lished here. In micro­elec­tron­ics, the notion of eco­sys­tem is very import­ant. In Gren­oble, we are for­tu­nate to have an eco­sys­tem that has reached crit­ic­al mass, bring­ing togeth­er the entire value chain, from start-ups to large groups, and this is invaluable.

We must con­tin­ue to invest in our strengths, from tech­no­lo­gic­al innov­a­tion to pro­duc­tion, but also strengthen the links between the semi­con­duct­or and the areas of applic­a­tion that are of vital import­ance for Europe (industry, auto­mot­ive, health, etc.), which have now become fully aware of the import­ance of components.

Europe is also ahead of the game when it comes to tak­ing envir­on­ment­al and energy issues into con­sid­er­a­tion: these two con­straints rep­res­ent sig­ni­fic­ant oppor­tun­it­ies for innov­a­tion. The CEA is, for example, lead­ing the European GENESIS pro­ject, bring­ing togeth­er 50 part­ners, with the aim of accel­er­at­ing eco-innov­a­tion in semi­con­duct­or man­u­fac­tur­ing pro­cesses. We also aim to reduce com­pon­ent con­sump­tion by a factor of 1000 by 2032.

Above all, how­ever, we must not for­get that the micro­elec­tron­ics mar­ket is cyc­lic­al by nature: what is true today will not neces­sar­ily be true tomorrow.

A major trend is emer­ging in which Europe could play to its strengths: edge AI, arti­fi­cial intel­li­gence man­aged not in data centres but on peri­pher­als, smart­phones, con­nec­ted objects, indus­tri­al units, etc. These embed­ded applic­a­tions require elec­tron­ics that are both very energy effi­cient and cap­able of per­form­ing the infer­ence phase, or even the learn­ing phase, loc­ally. How­ever, tra­di­tion­ally, the units ded­ic­ated to cal­cu­la­tion and those ded­ic­ated to memory are sep­ar­ated on the chips: 80 to 90% of the energy is con­sumed in the trans­mis­sion of data between the two. Edge AI will there­fore require innov­a­tions in elec­tron­ic archi­tec­ture, in which Europe has a role to play. It will also be closely linked to sensors, which are a European strength.

FAMES allows us to build 2,000 m² of addi­tion­al clean rooms and to acquire around a hun­dred new items of indus­tri­al equip­ment, rep­res­ent­ing an invest­ment of 830 mil­lion euros, sup­por­ted by both the European Com­mis­sion and the French State. It entered into oper­a­tion­al ser­vice at the begin­ning of the year. We have kept per­fectly to the planned sched­ule, which was very tight. It is import­ant to emphas­ise this: when you know Asia, you know that oper­a­tion­al exe­cu­tion is for­mid­able there. This proves that we are cap­able of doing just as well.

Its primary pur­pose will be to pre­pare FD-SOI tech­no­lo­gies for 10 or even 7 nm nodes. This tech­no­logy is cur­rently pro­duced by Glob­alFoundries and STMi­cro­elec­tron­ics, in 22 and 18 nm respect­ively. The tar­get mar­ket remains mod­est on a glob­al scale, but it is a par­tic­u­larly inter­est­ing solu­tion for embed­ded applic­a­tions that are look­ing for frugal­ity. But FAMES should also enable us to pre­pare the ‘next step’ for European man­u­fac­tur­ers, by accel­er­at­ing the devel­op­ment of oth­er tech­no­lo­gies deemed key for the next 5–10 years: non-volat­ile embed­ded memor­ies, which will play an essen­tial role in the afore­men­tioned nomad­ic AI uses, radi­ofre­quency com­pon­ents, which will sup­port the trans­ition to 6G applic­a­tions, and het­ero­gen­eous 3D integ­ra­tion, which will exploit stack­ing to integ­rate new func­tion­al­it­ies on a single chip. It should be added that we will par­ti­cip­ate in the oth­er pilot lines provided for by the Chips Act.

These pilot lines aim to pre­pare for Europe’s indus­tri­al future in the field of semi­con­duct­ors in the short, medi­um and long term. There is some­times talk of dif­fi­culties in col­lab­or­at­ing at the European level: this is far from being the case in the field of micro­elec­tron­ics. We work in close col­lab­or­a­tion, mak­ing the most of our com­ple­ment­ary skills, in order to respond effect­ively to the stra­tegic urgency we face.

Interview by Anne Orliac