Opinion: Next Crolles wafer fab helps keep Europe in the deep tech game
The news that ST’s Crolles site in France will gain another 300mm wafer fab and a base for further FDSOI process development is a good thing for Europe, the companies concerned and their customers (see ST, GlobalFoundries to build 300mm FD-SOI fab at Crolles).
However, the idea that it will help Europe get closer to producing 20 percent of the world’s chips by 2030 is far-fetched. But on the positive side it represents some realistic thinking by the politicians, and a willingness to execute by chip companies, that could help keep Europe in the deep tech game for another decade.
The news of a wafer fab that can produce about 51,000 wafers per month at full capacity is illustrative of the power of political will and tax-payer subsidy. Both Jean-Marc Chery, CEO of ST and Thomas Caulfield, CEO of Globalfoundries confirmed that without French state support the investment would be “challenging.” I take this to be short-hand for “would not happen.”
That an alliance and subsidies are required to build a next fab at Crolles and to provide a base for the development of FDSOI (fully-depleted silicon on insulator) manufacturing process is no surprise. After all the original 300mm wafer fab at Crolles was developed 20 years ago by an alliance between ST, Philips (later NXP Semiconductor) and Motorola Semi (subsequently Freescale and then acquired by NXP Semiconductor).
No European company could afford to go it alone at the leading-edge then, and none can now. Chipmaking costs at the leading-edge and close to it have been increasing exponentially since then.
For now, the CEOs of Globalfoundries and STMicroelectronics have not said much about the cost of their joint-venture wafer fab planned for Crolles or how much subsidy they will be getting from the French government under the European Chips Act.
Speaking with the press on a conference call ST CEO Chery said the fab would be a multi-billion euro investment and that France’s promised support was “significant.” Meanwhile the French government reportedly disclosed that the joint investment would be worth about €5.7 billion (about US$5.8 billion) without saying how much money it would be putting up.
If GF and ST have to find €4 billion between them then that would leave the French state providing €1.7 billion or about 30 percent of the cost. This would seem to represent a minimum level of subsidy in the modern era of wafer fab building.
It is also positive that Crolles next 300mm wafer fab is aimed at application sectors for where there is European demand and world-leading capability: in automotive electronics, IoT and communications infrastructure.
Crolles is set to be the second chip manufacturing site that is likely to benefit from the €43 billion European Chips Act (see European Chips Act could include powers for EU control). Intel is set to get about 40 percent subsidy support or about €6.8 billion (about US$6.9 billion) from the German government towards the €17 billion initial cost of building two side-by-side wafer fabs in Magdeburg, Germany.
The higher level of subsidy could be due to the fact that Magdeburg will be a completely new semiconductor manufacturing lacking in infrastructure that needs to be developed.
FDSOI down to 18nm
One thing we did find out from the conference call with GF and ST CEOs is that the fab ownership will not be 50:50 but in proportion to the ownership of equipment in fab and the output from the fab. The facility is expected to ramp quickly to full capacity by 2026 with a manufacturing capacity of up to 620,000 300mm-diameter wafers per year with ST taking about 42 percent of the wafers and GF taking about 58 percent of the wafers. GF CEO Caulfield said the ownership of fab was essentially in the equipment installed and that these were good figures of merit for that.
We also found out that the fab will make multiple processes including 40nm bulk CMOS, Globalfoundries’ 22FDX process and that it would provide an environment for both companies to take FDSOI down to 18nm. There was no mention of Samsung, which also has a finger in the FDSOI pie.
The ramp to full production by 2026 is aggressive and implies a start to production in 2024 and indeed the start could not be much sooner than that.
One concern is that some key pieces of 300mm chip production equipment are on 24-month delivery lead times while the chip making industry goes through a global spasm of capacity expansion planning. The supply of complete equipment chains could be a limiting factor in the roll out of the GF-ST fab but Caulfield did say that in some cases the companies have already put orders in with equipment companies.
20 percent by 2030?
Both CEOs were keen to say that their fab would support the aims of the European Chips Act to get European chip production up to 20 percent of global production by 2030.
According to market research firm IC Insights, Europe had 5.7 percent of the world’s installed wafer capacity as of December 2020, and given the lack of leading-edge production in Europe, by value Europe’s production could be considerably lower (see China rises in global IC wafer capacity ranking).
It is obvious that if Europe wants to gain market share it needs to invest in chip production proportionately faster than other regions. But the investments now being made by Intel, GF and ST under the European Chips Act – while welcome and unusual in Europe – are relatively minor by global standards.
Here we are cheering for the possibility of two or three production-scale digital wafer fabs getting built in Europe over the next several years while a single foundry company such as TSMC seems to throw 300mm fabs down each and every year.
If the US government can fund its own Chips Act it will unlock large investments in the US by Intel, TSMC and Samsung. Taiwan and China continue to invest much larger sums. At the same time regions such as India are eager to join in with chip making. The balance of semiconductor payments is now seen by many countries and most, larger geographic regions, as key to both short and long-term economic well-being.
Deep-tech in Europe
The conclusion must be that Europe’s share of global chip production will, at best, stop falling – but at least Europe has started to move after many years of taking the globalization dividend (see Opinion: Time for Europe to wake from a 30-year slumber). And it must be acknoledged that the visionaries within the European Commission have managed to managed to get chip company support in a way that Commission vice president Neelie Kroes could not a decade before.
With Intel eager to operate a foundry service at the leading-edge and GF and ST keen to operate in the More-than-Moore space, just behind the leading-edge, the European politicians have covered twin aspects of the need to keep up with global competition.
And it should be recognized that many future technology developments such as silicon photonics, quantum computing and spintronics are built upon know-how and capability in semiconductor manufacturing.
What we can say is that the European politicians, Intel, Globalfoundries and STMicroelectronics are helping to secure the chance of continued deep-tech capability in Europe for another few years. Further investment, from TSMC and/or Samsung, would be welcome. And at some time before the end of this decade it will be time to go again. The price of deep-tech progress and capability is constant support and subsidy.
Related links and articles:
ST, GlobalFoundries to build 300mm FD-SOI fab at Crolles
Globalfoundries, ST mull joint European wafer fab – report
Opinion: Time for Europe to wake from a 30-year slumber
Intel to get 40% subsidy for German wafer fabs
TSMC confirms German fab talks
GlobalFoundries valued at $25bn in IPO pricing
India lobbies Intel, Globalfoundries, TSMC for fabs
European Chips Act could include powers for EU control
Spain wants to develop 5nm chip manufacturing – report