Almost twenty European countries want a 2nm fab built in Europe. Impossible, you say? Future Horizons analyst Malcolm Penn doesn’t think so.
Europe is dusting off its ambitions in advanced semiconductor manufacturing once again. After several attempts in the past, nineteen EU member states recently signed a joint declaration to collaborate on “strengthening Europe’s capacity to develop the next generation of processors and semiconductors. This includes chips and embedded systems that offer the best performance for specific applications across a wide range of sectors as well as leading-edge manufacturing progressively advancing towards 2nm nodes for processor technology.”
Europe isn’t the only one seeking to bolster its local semiconductor ecosystem. The US has recently persuaded TSMC to build a 5nm fab in Arizona and Samsung is mulling to commence 3nm production in Texas. Japan is setting up an advanced IC packaging and testing plant together with TSMC. And even though results so far haven’t been as lofty as its ambitions, China is more motivated than ever to establish its own semiconductor industry now that its access to advanced technology is being squeezed by the US – a situation that isn’t likely to reverse any time soon.
ASML CEO Peter Wennink is skeptical about this trend. “It has taken decades to build a global frictionless semiconductor ecosystem. If you’re going to break this up into pockets of innovation, you’re going to increase cost. The knowledge to manufacture advanced semiconductors and associated materials and equipment is now concentrated in a few places in the world, which together form this seamless ecosystem. If you think you can replicate that within a very short term – it’s simply not possible. If governments are determined to do this, it will take years,” he warned at a press conference discussing Q4 2020 and FY2020 results.
Yet, it’s not hard to see why governments are pressing on anyway: it’s difficult to overstate the importance of semiconductors. As building blocks of an ever-expanding array of applications, including (digital) infrastructure and military equipment, chips are not only pillars of economies, they also support the autonomy, security and sovereignty of nations. Being overly dependent on other countries is increasingly seen as a dangerous vulnerability. The corona pandemic and recent geopolitical developments, in particular, have instilled a much greater sense of urgency to reduce any high-tech dependencies.
Adding to these worries is the unrelenting scale-driven concentration of semiconductor production: leading-edge IC manufacturing is gravitating towards one or two companies. With Intel fallen behind and seriously considering outsourcing production, only TSMC and Samsung are left standing. TSMC is based on an earthquake-prone island claimed by China as its own territory, while Samsung of South Korea lives under constant treat of an erratic neighbor with nuclear arms at his disposal. Their dominant positions are also cause for concern with respect to pricing power, although the two powerhouses have been keeping each other on their toes so far.
Europe, too, has opened its eyes to these strategic threats. “Europe’s share of the 440 billion euro global semiconductor market is around 10 percent, well below its economic standing. Europe is increasingly dependent on chips produced in other regions of the world – notably those used for electronic communications, data processing and compute tasks, including processors,” the joint declaration states.
The elephant in the room, of course, is that the chances of Europe successfully resuscitating advanced semiconductor manufacturing look bleak at best. There hasn’t been a leading-edge fab operational on the continent for two decades. Europe hosts three indigenous chipmakers, but they’re happy with a fablite business model and production of legacy or limited-volume niche products. The likes of Samsung and TSMC have little appetite to come here – if there was a good reason to do it, they would have already done so.
So, can it be done? Yes, says CEO Malcolm Penn of market research firm Future Horizons. He calls for all stakeholders, both public and private, to come together in the same way they did to develop GSM, which became one of Europe’s greatest success stories in technology.
Airbus of chips
It wouldn’t be the first time Europe tries to catch up. In the mid-80s, the European Union supported the efforts of Philips and Siemens (now Infineon) to leapfrog a process node to wipe out their competitive disadvantage. Although not a total failure, the Megachip initiative didn’t manage to get Europe back on track. In fact, if anything, it made the two companies lose interest in high-end commodity chips. It took until well into the 00s for this reorientation to complete. The unfortunate ending of Megachip marked the start of the current European focus on outsourcing semiconductor production, except for specialized products that don’t require a process upgrade every year or two.
For a while, the EU seemed fine with that. No alarm bells rang in Brussels when the Crolles chip manufacturing alliance, consisting of NXP (formerly Philips Semiconductors), STMicroelectronics and Freescale, effectively fell apart in 2007. Neither did they when Qimonda, a memory maker split off from Infineon and at one point the world’s second largest DRAM manufacturer, went belly-up in 2009.
In 2012, however, Neelie Kroes rekindled efforts to boost European semiconductor manufacturing. The Dutch European Commissioner for Digital Agenda famously called for an “Airbus of chips,” by which she didn’t mean a pan-European merger the way the aerospace firm was forged from national champions but a comparable alignment of company, regional and European interests.
Kroes’ so-called 10/100/20 plan was to spend 10 billion euros of taxpayers’ money to unlock an industry investment of 100 billion euros in order to roughly double Europe’s ‘market share’ in global chip production to 20 percent in 2020. She asked leaders of the European electronics industry – the three chipmakers, but also representatives from research (eg Imec, CEA-Leti) and adjacent industries (eg ASML and Arm) – to draw up a roadmap to accomplish that.
Dividends, not assets
Europe’s chipmakers, however, had no interest whatsoever in Kroes’ plan. “They all refused point blank to cooperate and have anything to do with it. They killed the whole initiative, kicked it into the long grass,” says CEO Malcolm Penn of market research firm Future Horizons, which authored a report that provided Kroes with the ammunition she needed to give her lofty ambitions credibility in Brussels.
Infineon, NXP and STMicroelectronics “didn’t want to invest a dime in manufacturing capacity. Of course, they still wanted to get their hands on the EU money,” Penn continues. “It simply isn’t in their interest to tie up huge amounts of cash in manufacturing plants. And, more importantly, the stock market doesn’t want them to either. Shareholders want dividends, not assets.”
“The only way you can get over this brick wall is if chip companies’ customers force them to build a fab close to them. This is what’s happening in the United States right now. Companies such as Apple and Nvidia are leaning heavily on TSMC and Samsung to put in some local production.”
Europe has no such behemoths. In fact, very few sizable industries that require a constant supply of chips are left, and none of them require leading-edge technologies. The one sector that comes to mind is the automotive industry, but its chip needs are generally satisfied using mature process technologies. It’s likely that more and more leading-edge chips will slip into cars as they’re equipped with even more sophisticated driver assistance systems and, eventually, self-driving abilities, but, unlike Tesla, no European carmaker has taken that up. BMW, Peugeot nor Bosch are developing those kind of AI chips in-house.
Without any market pull coming from Europe’s industry, things seem pretty hopeless. Yet, Future Horizon’s Penn sees an opening. A pinhole, perhaps, but if it’s a proper fab Europe wants, “there’s an opportunity staring us right in the face,” he says. “To play catch-up now with FinFET technology would be foolish indeed. But there’s a technology transition coming up, to gate-all-around. You need dislocation points like this to get your foot in the door. Gate-all-around is new to everybody. There’s absolutely no reason why it couldn’t be successfully commercialized in Europe.”
In Penn’s mind, it doesn’t have to be all that difficult. Leuven’s Imec has the technology ready to go. The European Investment Bank and the European Commission could finance a sizable chunk of it, perhaps more than currently allowed by state-aid laws. Worried about competitiveness due to European wages and labor restrictions? Set up an environment in which special dispensations and rules apply, similar to a free-trade zone. “If it’s a matter of European national security, such exceptions should be possible.”
It will be crucial to get the stakeholders onboard, every single one of them: the R&D labs, the chipmakers and the electronics industry. And, most of all, the end-users, Penn concedes. “Without market pull, we would just drown in the water that we’re trying to push up.”
What Penn has in mind, is to take inspiration from the GSM project. In the late 80s, the European Commission successfully leveraged its institutional authority to have member states back a single standard for digital mobile telephony. Instead of going head-to-head, key R&D labs, universities and telecom companies aligned towards the single goal of digitizing mobile networks, which in turn convinced network operators to invest. This rare display of unity gave the development of GSM great momentum; it was adopted swiftly across Europe and eventually became a global standard.
“I’m absolutely convinced something like that could be achieved today in semiconductor manufacturing,” says Penn. “It all depends on the political will of the European Commission. If someone like Neelie Kroes is willing to pick up the chalice and champion an ambitious program like this, the most difficult part will be to decide where the fab will be built – because nineteen countries will insist it should be theirs.”
Main photo credit: TSMC