Earlier this month, the European Commission presented its ‘Digital Compass’, which stated the intent to have manufacturing capacities below 5nm nodes and aiming at 2nm, and the production of 20% by value of global semiconductor production in Europe.
This is clearly driven by concerns about overdependence on chip production in Asia and the potential for being caught up in any future US-Asia political crossfire, which could lead to future chip shortages. But building a $30 billion 5nm or 2nm fab is not really going to solve the issue, as some industry commentators have noted.
One of these is industry analyst Yole Développement, who said, “In reality, a $30 billion fab plant is a perilous way to secure European technological sovereignty.” In its snapshot of the semiconductor industry, Yole said constructing such a state-of-the-art facility, without support from TSMC or Samsung, will take at least 10 to 15 years and demand tens of billions of dollars of investment. In addition, it may not even end in success. Citing Intel’s challenges, Yole said, “Intel has encountered difficulties on delivering its own advanced 7nm manufacturing process – if the US computing giant can’t easily succeed at this process node, should Europe pursue an even riskier jump to 5 nm?”
Evolution of semiconductor process technologies and key players (Source: Yole Développement)Yole’s view is that the only way for the EU to establish a 5nm fab in Europe would be to follow a similar path to the US, and team up with TSMC or Samsung. The joint effort would require fewer funds from Europe – around $10 billion – and dramatically cut development times from 10 to 15 years to around 3 to 4 years.
But this won’t necessarily solve the sovereignty issue. With or without TSMC or Samsung, Yole questions how much demand there will be in Europe for 5 nm, in three or four years’ time. “All the volumes of wafers to be produced for the continent’s telecommunication, automotive and other industries simply won’t fill such a fab, killing the rationale of this huge investment.”
So, what is the solution?
Well, the thinking is that there will be an increase in manufacturing outsourcing to TSMC and Samsung anyway, simply because it’s just going to take too long to build a fab, install and commission equipment. Yole said, “The best way to technological sovereignty is to invest wisely and create a robust manufacturing ecosystem that will deliver semiconductors to European companies in the long run.”
The analyst firm notes Europe is already home to cutting-edge chip manufacturers and equipment suppliers, each with annual sales from $4 to $16 billion. These include STMicroelectronics, Infineon Technologies, NXP Semiconductors and Ams as well as GlobalFoundries with its Germany fab, Intel and ASML. Meanwhile, several smaller but strong industry players co-exist in Europe, including semiconductor foundries, Tower Semiconductor and X-Fab, each generating between $500 million and $1 billion in sales every year.
After this, there are at least 200 companies, including the IDMs (integrated device manufacturers), subsidiaries of system makers and semiconductor equipment manufacturers, are sprinkled across Europe. The likes of Elmos, Murata Europe, Besi, EVG, Soitec and Siltronic each earn less than $500 million annually but sustain Europe’s manufacturing and technology independence.
Yole added, “Given this rich and established tapestry of industrial semiconductor players, the European Union and national states need a solid strategy to strengthen these suppliers. A first, important step, is to build the intermediary nodes, 14nm or 7nm foundries, in Europe that will support the continent’s current automotive, telecommunication, IoT and industrial applications. These facilities could be EU-funded but also co-developed with ST, Infineon, NXP, Ams and ASML.”
In addition, the EU should invest in activities such as heterogenous integration, advanced packaging and chip partitioning from 14 nm to 7 nm. This would avoid the urgent need for 5 nm and 2 nm chip production while enriching Europe’s technology know-how. A further step could be to leverage Europe’s strong R&D effort in emerging computing and photonics where a large number of companies and startups have a role to play.
This strategy would strengthen Europe’s semiconductor industry and provide a path toward 5nm and below in the longer term. It would also address the issue of more manufacturers being bought by US- and Asia-based companies, which is a growing trend. For example, in 2014, LFoundry shut down its fab in Rousset, France, in 2016 Italian fab, Avezzano, was sold to China’s SMIC and then in late 2020, Nvidia, US, proposed to acquire UK-based Arm. Earlier this year, Japan’s Renesas bought Dialog, UK, and Taiwan’s Globalwafers is in talks to buy Europe’s key wafer substrate provider, Siltronic, Germany.
Yole points out the fallacy of trying to build a 2nm or 5nm fab, stating, “The prospect of financing a 5 nm foundry in Europe while the continent’s finest semiconductor businesses are being subsumed by US- and Asia-based behemoths is akin to building a cathedral in the desert.” It believes that investing in home-grown businesses and infrastructure that will help Europe’s technology ecosystem to flourish is the only way forward, “otherwise the continent could one day become an industrial desert with an empty 5nm cathedral at the center.”
This kind of sentiment is echoed by others. In a recent interview with the Netherlands-based Bits & Chips, Imec chief strategy officer, Jo De Boeck, said, “Nonetheless, it’s going to take years and a lot of effort before Europe is ready to even consider building a fab. You can’t simply skip multiple nodes; we need to take steps towards it. At Imec, we believe Europe should continue to strengthen our strong base in R&D, manufacturing equipment and materials. When the basic technology is ready, we should bring companies together and start pilot production lines for promising applications that will be relevant for our industries and society. Once production-readiness is in reach, perhaps building a leading-edge fab makes sense. Another option is to consider bringing in a leading-edge player to accelerate this process.”
Indeed, Europe needs to take a more pragmatic, step-by-step approach. Having the big vision is great, but it will be a long-term iterative process rather than something that can be conjured up quickly by throwing money at constructing a leading edge fab. Europe needs to connect the dots between its excellent research and knowledge base with the established industry players already here, and then build upon the strong foundation it already has.