Posted on 06/25/2021 6:01:56 AM PDT by blam
This is interesting for what it says about the state of European semiconductor manufacturing:
German tech giant Bosch has opened the doors to a new semiconductor manufacturing facility in Dresden, which it hopes will plug the significant hole currently plaguing automotive supply chains.
The site – which was estimated to cost the company in excess of €1 billion (AU$1.57 billion) to build – will reportedly begin producing much-needed chips for vehicles by early September this year. It is so far unclear how the product will be distributed, and if German manufacturers – including Volkswagen, BMW, and Mercedes-Benz – will be given priority in the queue.
“The new [factory] is the single largest investment in the company’s history. This cannot be stressed too much. Its size and additional production capacity alone are impressive. The very latest methods of data-driven continuous improvement in production make the Dresden plant a smart factory,” said German Chancellor Angela Merkel.
“To put it another way: in this plant, natural and artificial intelligence have joined forces with the internet of things to form a productive symbiosis.”
Here Alte Jungfer Merkel is no doubt repeating some generic technomarketingspeak handed to her by an aide, as I suspect Merkel is as ignorant of current developments in artificial intelligence as I am of the songs she sang in the Freie Deutsche Jugend.
Here’s a Deutsche Welle English video segment on the fab, for which the headline says “Europe aims for independence from Asia”:
How cutting edge will the Bosch fab be?
The 300mm fab(12 inch wafers) has been under construction since June 2018 and as recently as October 2019 it was stated that the Dresden wafer fab was expected to start operations in spring 2020…
The fab makes use of 5G communications and artificial intelligence [there’s that phrase again -LP] for extensive automation and will be used for the manufacture of power semiconductors and ASICs for automotive applications down to a minimum geometry of 65nm, according to a company spokesperson.
300mm is industry standard, and Dresden makes sense, as Infineon, Global Foundries and X-FAB already have fabs there, so there’s a skilled labor pool to draw on. The strange thing here is spending $1.2 building a new 65nm fab, since that’s about seven process node generations behind the cutting edge. 65nm was cutting edge way back in 2005.(Many companies are presently running 7nm and testing 3nm)
You don’t need cutting edge process technology to build automotive integrated circuits, which tend to use older process nodes, but 65nm doesn’t give you much headroom for the long haul. It also doesn’t really do much to “compete with Asia,” since it’s hopelessly behind not only leaders TSMC and Samsung, but also about half the fabs run by Chinese companies SMIC and Unigroup. It’s hard to see investing $1 billion in building a older technology fab as opposed to contracting out production to foundries. (Of course, foundry capacity is highly constrained right now, and maybe Bosch leadership was able to see that coming when they broke ground in 2018.)
Could it still turn out to be a decent investment? Possibly. In the short haul it will relieve the current capacity crunch and let Bosch gain share from competitors who can’t book foundry production. Plus semiconductor manufacturing seems like the sort of thing European governments like to subsidize. And indeed, the DW piece notes that “The German government also invested in the plant.” (It also wouldn’t surprise me to find out that various inscrutable, multi-acronym EU agencies unknown to voters in Barcelona and Gdansk are also kicking in money.) Fueled by taxpayer money, building a 65nm fab could still be a profitable proposition for Bosch.
But fabs close all the time, even 300mm fabs. Building a fab dedicated to such an old process node makes profitability a much more challenging proposition, especially given the next (inevitable) industry downturn. You only have so many years before everything is obsolete. Bosch has given themselves a very short runway to profitability for this investment.
Back in the 1990s, Europe was competitive with American and Japanese semiconductor companies, with fabs built in the UK, Ireland, France, Germany, the Netherlands and even Italy. Today, of those countries, only Ireland (Intel) and Germany have any 300mm fabs at all, despite having a highly educated workforce and modern technological infrastructure. (Russia evidently has one as well, owned by Crocus Nano Electronics (about which I know next to nothing), also running 65nm, that started production in 2015.) Europe’s semiconductor industry has been passed not only by Taiwan and South Korea, but also by China.
While Europe was integrating, its semiconductor manufacturing capacity was stagnating. Highly integrated, yes, but also highly regulated and highly unionized.
As I’ve said before, building cutting edge fabs is a very expensive game to play. Collectively, the world has decided that Europe is not a suitable location for one.
I've said this for years...and, have advocated new factories to run 'old' technologies. Making money will be the problem.
I hadn't realized how far behind Europe had gotten. we're headed in that direction if we're not attentive.
Silicon atoms are .2 nm across. At 3 nm that is 15 silicon atoms across. Not much room left at the bottom here.
I guess we'll go to Qubits, eh?
A $1 billion fab? I thought they were now in the $4 billion range.
Actually, making money will be easy as the equipment is “used” bought at deep discounts. That’s what analog companies use and they make a fortune.
$1B gets you a 65nm fab. $12B gets you a 5nm fab like the ones TSMC are operating. The margins on TSMC’s products (Apple, Nvidia, Qualcomm) are huge compared to the margins on automotive semiconductors. Plus there is a lot of competition when the market gets soft.
The sweet spot in auto semiconductors is in sensors with integrated electronics, which requires a totally different fab approach than popcorn parts, especially for MEMS. BiCMOS, GaN, GaAs, mixed-signal ICs.
Automotive sensors don’t have to fit in your pocket, they don’t have to perform 23.56 gigaflops, they don’t have to run off of a 5volt 2 ounce battery. You don’t WANT cutting edge for most car sensor parts. You want reliable, proven, and you want to use the same parts for a decade if you can.
I'm expecting that 7nm or better new fabs are in the minimum $12 billion range.
If I had Jeff Bezos's money I'd build the highest tech wafer fab possible with a separate research and a separate development facility on campus.
The new fab would have ample room to begin the next generation of chip production 'down the hall' from the current high-tech production without interrupting that production.
I've done this with very good success at Texas Instruments, decades ago.
Western Electric had a 40 year MTTF (mean time to failure)specification on much of the equipment they supplied to Bell Telephone, years ago.
"We've been trying to reach you about your phone's extended warranty" /runs for exit, ducking for cover>
Disclaimer: Opinions posted on Free Republic are those of the individual posters and do not necessarily represent the opinion of Free Republic or its management. All materials posted herein are protected by copyright law and the exemption for fair use of copyrighted works.