UT Bram Nauta

Bram Nauta is a professor of IC design at the University of Twente.

2 February

Instead of paying some company for a fab that will manufacture chips for US companies, Europe should build a chip ecosystem of its own.

As we all know, there’s a shortage of chips in the world. Building additional production capacity will take several years due to the long chain of industrial players involved. The fab needs a clean building, inside we need wafer steppers and other machines. Moreover, these machines must be built from parts that are in short supply. In addition, the world needs to keep innovating toward more advanced CMOS nodes and develop even more advanced machines. All this is a complex process that’s hard to accelerate.

In the meantime, we’ve started to realize that chips are used in just about every product except toilet paper. This warrants looking at them through a pair of geopolitical glasses. Unlike bolts and nuts, they’re hard to design and produce, and it takes a long horizon to get their production in order. Without access to chips, tech companies are at risk. The knowledge to design and fabricate these components has become a strategic asset.

The chip production chain is a global one, to which many countries have contributed for years. In Europe, we installed the bulk of our fabs some 20 years ago. ST in Grenoble and Globalfoundries in Dresden have made their last leaps down to twenty-something nanometer. Oops, I almost forgot the Intel fab in Ireland. The European technologies are excellent for applications that do not need a lot of computing power. But our global market share has dropped to roughly 10 percent. Soon the chip market will double again, and we’ll be at 5 percent. Completely marginal.

Some politicians say we should build a modern wafer fab in Europe. This may sound smart but let’s look at the potential customers: they’re US-based fabless companies like Qualcomm, Broadcom, Nvidia and AMD. Another big one is Mediatek in Taiwan. Europe sold its last decent fabless company, Dialog, to Renesas in Japan. We have no significant fabless semiconductor industry anymore and thus no customer base for these modern fabs. Why would we build a fab in Europe if the demand is in the US? Especially considering the fact that we would be paying an existing player to do it.

It’s a bit like we want freedom of the press for democracy and then buy a printer to print newspapers for other countries. Does that solve your strategic problem? It doesn’t. You need newspaper agencies and critical journalists – that’s the key part. If Europe wants to play a role in this industry, we need to first create our fabless chip ecosystem. We need more chip designers. We’re strong in analog and RF, but we need more, including digital and software. Universities can deliver these.

We should also dare to jump ten years ahead in technology development. The 3nm race already has a winner, but in ten years, technology will be very different again: vertically standing transistors, most likely stacked in many layers. We can start figuring out how to produce these future structures. No, I don’t mean Science and Nature style of research at 0 kelvin with 1 out of 1000 samples being functional, but things that work at room temperature and can be mass-produced at an acceptable cost. Institutes like Imec and Leti, together with universities, can do this.

Don’t tell me this plan is hopeless. I clearly remember the days that Qualcomm and Broadcom were baby companies. In one or two decades, the world will be a different place again. Invest the money that we otherwise would spend to buy a fab, create this ecosystem of technology and designers and make it happen. Focus on the chips of the future but also build on what we’re good at today. Yes, it requires both public and private money, but most of all: we need the attitude to invest and the desire to work in this fantastic technology. And one day, we may wonder how we ever used toilet paper without chips.