Local fabs for local people

Spain is not exactly famous for its presence in the chipmaking business. But it wants to be, declaring at the WEF’s Davos conflab that the government has set aside out of an electronics-focused budget of €12bn some €9bn to help get local fabs built and, in principle, reduce both the country’s and Europe’s dependence on Far Eastern and US suppliers.

According to news agency CE Noticias Financieras, economy minister Nadia Calviño, who was standing in for Pedro Sanchez at home while the Spanish prime minister was in the Swiss resort, claimed the Spanish government has made "preliminary contacts" with operators in different sectors.

There is one obvious issue, if the Spanish government wants at least one 5nm-capable fab €9bn does not go very far. It would give the country the ability, should EU subsidy rules allow, to go 50/50 with a willing partner, which does provide a more realistic number. But even, that is pretty much just one advanced, but not bleeding-edge, fab. Attracted by the US government’s entreaties to build a fab outside its native Taiwan, TSMC estimated its moderately sized fab and packaging centre in Arizona, also focusing initially on processes down to 5nm, would cost a little over $17bn. TSMC took a lot of persuading, choosing to focus efforts in Taiwan with only older processes being passed out to factories in China and Singapore. And that Singapore fab is there because of an old joint venture with Philips Electronics: the Dutch company helped get Morris Chang’s company set up in the first place several decades ago. 

Calculations in 2021 by Scotten Jones of consultancy IC Knowledge estimated TSMC's US fab would be around 17 per cent more expensive to build and operate compared to an equivalent 'gigafab' back home. About 10 per cent of that is down to it being a smaller fab, which means higher operating costs per wafer. Only 7 per cent is due to local cost differences. The bigger issue for any company putting an advanced fab in place is the cost of getting the people to run it and stick around.

Back when globalisation was somewhat more fashionable among governments than it is now, the talk was all about clusters. In the EU, it was clear that any attempt to clamber back into advanced semiconductor manufacturing would involve building on existing clusters of expertise, such as the fabs in Dresden, not very far from Intel’s planned fab in Magdeburg, the Imec-ASML grouping in Benelux or the STMicroelectronics-Leti  in Grenoble, France. There is good reasoning for this. Silicon Valley works because it was full of people experienced in chipmaking and is now full of people who do the designs for clusters that have sprung up elsewhere. The UK has its own electronics-design clusters that people in the industry are keen to protect.

Imagination Technologies and Global Counsel, for example, published their take on what the UK’s semiconductor strategy should be a few weeks ago. Given Imagination’s involvement, it is not a surprise that the emphasis is very much on IP development rather than onshoring manufacturing. There is not much risk of the latter. The UK government does not need a whole lot of persuading not to build fabs as it has shown little interest in the past beyond providing some tax breaks when Korean producers briefly showed an interest in moving production outside that country: the jobs simply aren’t there and it’s expensive. Any lobbying in that direction is pushing at a door so wide open, the biggest risk is stumbling straight through. And, other than the defence sector, which is arguably one of the bigger influences on the US strategy, there is not much appetite for onshoring or reshoring in the electronics sector overall. 

The reality is that the level of technology involved relies pretty much on a global system remaining intact. When a single lithography scanner, needed to define where the transistors go on the chip, costs more than a large passenger airliner the economies of scale are very much against nationalism. Plus, any real-world system needs a variety of different types of chip, which could and probably will come from anywhere. That gets even trickier when one major trend is multichip modules: putting multiple chips into one package. This takes advantage of the different strengths of newer and older process technologies. Though this is one area where onshoring might make a degree of sense, the wide range of packaging options even in this one corner of manufacturing would make it difficult to justify setting up lots of small, local factories. They are for the foreseeable future going to be mostly clustered in the Far East.

The problem for UK industry is finding a way forward for the remaining industry even if the options are relatively cheap for a government to implement. Imagination wants the UK government to “protect and promote UK IP abroad” and “expand enforcement and cooperation on IP rights registration”. Making that stick in a world where countries see competition between them as increasingly important is easier said than done. And the UK government is not just not making friends; it is having some trouble understanding it might want to keep some existing friendships intact and forego the brinkmanship and sabre-rattling over deals it willingly signed just a couple of years ago.

Similarly, companies like Imagination want to make it easier to organise visas and work permits for engineers coming from overseas. Though the government makes warm noises about points-based systems, it shows little appetite for fast-tracking applications or making them cheaper. 

Another recommendation goes deeper than the concerns of the semiconductor industry alone, and some of the required action is supposedly under way. This is to try to close the gap between research and commercialisation through the UK’s own version of Darpa as well as the expansion of the British Business Bank into more advanced sectors.

However, it is another aspect of the chasm-crossing problem that British firms face where again UK government simply stepped aside. For a while, it looked as though there would be a viable pilot line, similar to those set up in the EU to assist with prototyping and low-volume production, for compound semiconductors based around Newport Wafer Fab. Ron Black, former CEO of Imagination worked on an attempt to secure that instead of the fab being sold to Chinese-owned Nexperia, a move that raised issues of security around advanced development that could happen there, even if the new owners wanted to accommodate it.

One issue is that the analysis by the UK government has focused more on national security rather than commercial security, at first nodding the deal through before doing another U-turn just days ago that it would conduct a another review over national-security concerns. Realistically, national security is a fairly weak argument for blocking an acquisition of that nature. What the debate has missed every time is that fabs of this kind are resources with a bigger effect on the local technology industry than ministers consider. All that the endless vacillation has done is push UK fabless companies to use facilities in the US and elsewhere.

Like Spain, the UK’s problem of gaining greater traction in semiconductors is one of “I wouldn’t start from here if I were you”. Unfortunately, the trajectory in the UK does not look that promising even with more of a head start at the design end of the business.