Go mainstream!

Profit in the semiconductor industry is concentrated at the most advanced CMOS nodes. Looking at the results and expectations of TSMC, the mature nodes remain important, but the ‘extra’ growth is in the sub-7nm technology. Europe doesn’t play a role there. European industry got stuck at 16nm, if we even reached that node. 16nm is nice for automotive, RF and analog, but if we want to play a bigger role in the global semiconductor industry, we must expand to the most advanced nodes.

This isn’t easy. If I talk to industry, I don’t hear much of an ambition to go there. “We’re happy with our technology.” “We don’t do these big digital chips anyway.” It makes sense from an existing company’s business perspective. In Europe, we’re strong in analog/RF circuit design, and doing that in 7nm doesn’t make sense. The transistors aren’t any better than in the older nodes, and likely even worse, due to relatively large interconnect parasitics. Doing your first layout of a simple inverter in FinFET technology without experience may take you one week minus some sleepless nights. So, the current European industry doesn’t even try.

However, the growth in semiconductors is expected to be in computing and AI, and this requires a lot of digital computing area in advanced CMOS. Area means turnover and market share. AI will make decisions for us, it will replace civil servants, car drivers, teachers, lawyers, and even writing columns like this will be done by AI. If a nation, a continent or at least a large group of people with shared values becomes so dependent on AI, you’d better control it.

To start controlling AI, you first need hardware and design these big chips in sub-7nm CMOS. You need talent to do this and a lot of venture capital. In Europe, we barely have talent mastering the design of these kinds of chips. They’re way too big and too expensive for universities. There are a few exceptions of academic projects that still master the bit, but in general, we don’t educate complex digital hardware engineers anymore. The second problem is venture capital. Europe has a lot of pension money in reserves, but the venture part of that is invested in the US, not here.

So, we’re kind of stuck. We have ESMC building a 16nm fab in Dresden and Intel in an identity crisis, wondering what company it wants to be, having just pulled the plug from its German very-advanced fab investment. 16nm it is. And it’s meant for the automotive industry, which is of strategic importance in Europe and especially Germany.

Meanwhile, our European cars are threatened to be rush-passed by much cheaper new Chinese car brands. A decade ago, we were still laughing about the ugly and unsafe Chinese cars and Vorspung durch Technik was the German credo. How things have changed. Modern cars are driving batteries with some sensors and a lot of silicon chips. Europa has no battery industry and no advanced chip industry. European cars will become oldtimers.

Europe wants to invest in chips, but the result is a lot of opportunists who jump on the money without much of a vision. Best case, they want to invest in mature IC technologies (and this is good!), but in most cases, they escape to adventures like photonics, quantum and superconducting. Yes, we should invest in these technologies, but by the time they’re mature, Europe has become a tech desert already. And none of these technologies can replace CMOS anyway.

If Europe wants to survive in semiconductors, we must go mainstream CMOS. Start educating more digital chip designers. Make analog and RF designers confident with FinFET technologies. Most importantly: make sure we have venture capital. If we don’t spend it now, there won’t be much to be spent later.