Europe's chip ambitions: Manufacturing

Back in 2018, I wrote about the European Processor Initiative (EPI), one of Europe's recent ambitious plans to re-gain some expertise in the chip industry. The chip industry is one of the leading industries of the world where Europe is lacking far behind other major powers, the USA and other Asian countries such as China, Taiwan and South Korea.

The task of designing a high-performance and low-power SoC on the basis of the ARM instruction set was meant to bolster its own chip designing capabilities and to have more control over the chips it uses for High-Performance Computing (HPC), Automotive and Artificial Intelligence (AI). The EU founded the company SiPearl to do all of the design work, hoping that their products will be sought after elsewhere in the world once they are in production. There is also an accelerator planned on the basis of RISC-V. The first iteration of these chips was originally planned to be available in 2021, but they got postponed to 2022. 

Designing a chip on the leading edge of technology is hard and costly enough. But building a manufacturing infrastructure is even harder and much more costlier. The CEO of ASML, the famous Dutch lithography equipment maker, said in an interview recently about the new plans of the EU to also build up their manufacturing capabilities in this sector: "Wanting to build a factory for 2-nanometer chips in Europe is the same as saying you want to build a rocket to send people the moon," - and did I mention already that they don't come cheap, too?! Such a modern factory (called fab in the chip sector) would cost about 20 billion EUR. And you don't just need the fabs, you need the suppliers and enough customers which want to build their chips there, too, to run it at 85 - 90 % utilization.

European chip makers don't seem to be too keen on the idea, read the Politico article for more details about their stance. In short, they would rather like to see some investments in older fabrication technology where capacity is also scarce, but the EU is rightfully insisting that the industry could do that already with their own money. For a big strategic investment they want something at the forefront of technology which would also serve as the technological foundation for EPI. And I support the commission with their more ambitious stance as you need to think of tomorrow and the day after tomorrow, not just the present needs which can be solved by the industry themselves by simply investing more. Europe could provide desktop and laptop chips, not only for government and industry PC's, also for gaming systems. Apple and Microsoft are transitioning to the ARM architecture nowadays. By 2025 - 2030 we all could be using such a PC. The hunger for chips is only to get bigger elsewhere, too. Not necessarily at the leading edge, but still. The world is also ripe to break the dominance of TSMC which cannot satisfy all of the demand in the market currently and the near future, which is already felt in the car sector since the beginning of the year. While TSMC heavily invests into capacity expansion and new processes (100 billion USD over the next three years), all of their leading fabs today are located in Taiwan and the dangers of such a concentration became apparent last year with the pandemic and this year again with a drought which could impact the access to water for the company.  

Of course there is the danger of mismanagement and failure to compete successfully with these players, also no one wants to repeat the mistakes the GDR made with the build-up of its chip industry. The high costs alone dictate that the EU is unlikely to tackle to build up the manufacturing infrastructure all by itself, instead it is highly likely that they want to support one of the leading foundry companies to invest in Europe. This would also open the door for fabless companies from all over the world to utilize such new capacity. There is clearly demand for leading edge capacity. Publicly only Intel seems to be the most interested of the considered companies, Samsung declined already and the dominant TSMC is more cautious about its plans and recently committed itself to similar efforts in the USA. 

Intel would be a tough choice. Not only did the company lose its technological lead during the last decade to TSMC, it also just entered the foundry business seriously a couple of weeks ago under their new CEO Pat Gelsinger who is praised a lot for his former technical background at Intel and his leadership skills. But Intel is still highly accustomed to itself as their only customer of their processes. Changing a philosophy is simple, it is about the thinking of the employees and changing their internal processes that might take a lot more time and effort. There is also a high risk of Intel simply failing to deliver on their promises on time, it would not be unprecedented in the recent past: Their 10 nm process debacle is still on everyone's mind which failed to materialize for more than five years. Intel is also at the forefront of the US CHIPS initiative. There is potential for cooperation with the USA to spread some of the costs on more shoulders for some of the technologies which come to light under this initiative. But don't count on it, the construction of a fab and other manufacturing infrastructure is a highly local affair, the possibilities to share seem to be more of use for EPI.

My first choice would be TSMC. Even though this would be another first for the Taiwanese company, it would strengthen their global footprint and would diversify some of their risks further. They would also gain access to major European vendors and European know-how in other related industries, such as the chemical industries. With a presence in Asia, the USA and Europe it would become a truly global company. They also have the deep pockets for such a fab expansion. In addition, their roadmap execution was excellent during the last 10 years.