Europe faces the Airbus vs air miles decision for chips
Image credit: Airbus
The European Commission is taking a second look at reinvigorating chipmaking in the region.
It’s hard to think of a more globalised industry than chipmaking. The devices themselves probably do more air miles than any other product. Worth more than their weight in gold, air-freighting even half-finished chips from country to country makes sense. Wafers can start off on one continent, get despatched to a fab in Taiwan and then move on to a packaging plant on the other side of the South China Sea.
Now that multichip modules have become mainstream in phones and a growing portfolio of other products, a packaged memory could wind up being flown to a second packaging plant in a different country before it starts to make its way through the distribution chain and into a shipping container once final assembly has taken place.
At the end of 2019, it was hard to see that trend reversing even though China and US were coming to blows over the consequences of that globalisation: that strategically important technologies in effect belong to no-one. Faced with the prospect of the American Century being the last one, the US started to reconsider its trade-off between corporate profitability and national muscle. You can hardly expect to conduct a war successfully using drones and AI if the other side is likely to be equally good at knocking them out.
At the same time, after two decades of surplus, semiconductors were already moving into a shortage of supply. The big chipmakers learned a harsh lesson at the end of the 1990s: oversupply is bad for margins. And by bad I mean really bad. In all honesty, they had been learning that lesson roughly every five years for the previous few decades, but as the growth periods were so good the odd recession just happened to be good at knocking out the weaker competition. But the post-1995 slump in price just kept on with only a brief respite during the dot-com boom. Then, as geopolitical storm clouds gathered, prices were showing signs of turning around and going up, largely because the ability of equipment manufacturers to demand the next shipment be cheaper was dissipating. For once, the fab owners could call the shots.
Couple that with a pandemic that put a number of globally important supply chains on hold, followed by a flurry of activity when carmakers and other manufacturers realised they had overdone the order cancellations, and you wind up with a proper shortage on your hands. There is every chance this will turn around next year. Already, Gartner expects the ever-volatile DRAM and flash sectors to see prices drop sharply by the end of 2022 because capacity is opening up faster than predicted increases in demand. In principle, it is a supply situation that is likely to resolve within a year, assuming no other shocks. However, governmental bodies in the US and Europe remain keen to build even more capacity and find ways to cut those air miles for chips.
CEOs in technology like to talk about the opportunities of disruption. Normally, they mean technological disruption. The last time the European Commission took a long hard look at bringing the region back into the chipmaking elite, the potential disruption was very much one based on technology. For years, Europe’s fab-owning companies decided, like many others, they were better served going to foundries in Taiwan and China. A decade ago, people in the industry and civil servants under Neelie Kroes, the then commissioner for competition and the digital agenda, looked at the possibility of getting either a local company or an outside player such as Intel or TSMC to get involved in an “Airbus for chips”.
The point of disruption was what seemed the imminent increase in wafer size from 300mm to 450mm. Wafer transitions are rare, but because they can as much as halve the cost of making chips, they can prove to be lucrative for companies that are able to spend the money to make the jump for their advanced products. The competition just falls away.
The Airbus-of-chips plan was to take advantage of the research infrastructure in Europe through lithography specialist ASML, and the Imec and CEA-Leti institutes, to encourage a player such as TSMC to set up shop away from Taiwan with a shiny new 450mm plant. The plan had multiple problems. None of the companies in the region in the business of actually selling chips cared: they were perfectly happy emailing layouts to Taiwan. Such a megafab would not employ that many people. They would be highly paid but the numbers barely moved the needle for local politicians. That at least overcame the potentially ugly national scramble as to which nation would host the big fab. The final nail in the coffin was that the chipmakers went cold on 450mm, deciding to play safe with 300mm wafers and trying to eke out savings there, not least because the equipment makers they had hammered down in purchase negotiations year on year saw little point in giving big chipmakers even more leverage by taking out even more competitors.
Today, the disruption is geopolitical. Taiwan looks more than a little vulnerable to a military power that has never recognised the island’s independence and has done whatever it can to make sure no-one in the United Nations recognises its independence either. Supply disruption because of Covid makes the argument for a local supply even easier. And that is roughly how Europe is now looking at European Chips Act that will partially mirror the kind of inward investment the US is planning following TSMC’s decision to build a relatively small plant in Arizona.
Can it work this time? It might be possible to bootstrap a mega fab using Imec’s experience, which includes running some fairly big pilot lines in order to support the work on extreme ultraviolet (EUV) lithography with ASML as well as all the process pathfinding it does. But running foundries is hard. Just ask any competitor of TSMC. The Taiwanese giant has suffered the odd wobble, but the last major issue was way back with the 45nm generation. They are now getting ready for the N3 or ”3nm” node. In the meantime, other foundries have found it hard to keep up and have largely given up trying.
A further problem is that the major players in advanced silicon design are mostly in the US, Korea and Japan. The picture is further complicated by a process of disaggregation in the chip designs themselves. Monolithic integration is not dead, but it has looked better. EUV and by extension ASML is the only game in town for anyone looking to get advanced designs made. But it’s power hungry and expensive to run. This has made the economics of integration far more delicate. Some circuitry just does not shrink all that well. Whereas chipmakers used to eat the cost of that for the convenience of getting as much as possible onto one die, that simply does not work now. So you have to have chips coming from multiple lines and package them together. This implies those chips still doing a fair amount of travelling, and the regions that benefit will have the largest, longest-established infrastructure.
Europe, like the US, could favour certain industries. The US made it more attractive for military-focused designs to move back onshore. Europe could take advantage of its automotive and other industries. To some extent that has already happened. In mid-September, Infineon cut the ribbon on a 300mm plant in Austria for power electronics that will most likely go into generators for renewables and electric vehicles. But for the rest of the car? Those chips will continue to rack up the air miles, rather than heading to a local Airbus of chips.
Sign up to the E&T News e-mail to get great stories like this delivered to your inbox every day.