Near Dresden in “Silicon Saxony” there are semiconductor manufacturing plants from Infineon, Bosch, X-Fab and Globalfoundries. Globalfoundries, the third largest chip contract manufacturer in the world, processes over 40,000 wafers per month in its Fab 1 and creates microstructures down to 22 nanometers. Fab 1 was created in the 1990s as a factory for AMD processors. Since then, Globalfoundries has invested more than 12 billion US dollars in Wilschdorf.
In the US state of New York, Globalfoundries manufactures even finer structures down to 12 nanometers. The two largest chip contract manufacturers (foundries) TSMC and Samsung Electronics are already proficient in 5 nanometer technology. So far, however, Intel has produced the finest structures within the EU: namely 14 nanometer processors in Leixlip, Ireland. Intel is currently building another large fab there, presumably for 7 nanometer chips. Globalfoundries, in turn, plans to introduce the particularly economical 12-nanometer technology 12FDX in the next few years and wants to double the capacity of the busy Fab 1 beforehand. This fits in with the European Commission’s plans to increase chip manufacturing capacity in the EU. This is also about digital sovereignty, i.e. independence from other countries when it comes to the supply of chips.
Dr. Manfred Horstmann heads the Globalfoundries Fab 1 in Dresden; he has worked there since the AMD era and knows the industry very well. So we asked him a few questions about chip manufacturing.
c’t: Are the expansion plans of Globalfoundries only about projects within the framework of the EU project IPCEI microelectronics or about more?
Dr. Manfred Horstmann: Let me start with the “big picture”: Investments in microelectronics have been massively promoted around the world for years. An economy that “can” microelectronics in terms of research, development, production and the necessary know-how for systems and integration has an important key to competitive and sustainable industries.
But Germany and Europe are “late to the game” – for more than two decades. With the first IPCEI, after initial delays, we have found a very good instrument that supports the European chip sector in making high investments – and thus even more so the industries whose future depends on crisis-proof access to microelectronics. The second IPCEI should start quickly: The rest of the world doesn’t wait for us in Germany and Europe to sort out our complex processes. We want to implement our Dresden expansion plans within the framework of the first IPCEI Microelectronics, which is still ongoing, and the planned second.
c’t: All other large semiconductor manufacturers and contract manufacturers are currently expanding their capacities by billions of dollars. Are you not worried about the famous “pig cycle”, i.e. about overcapacity?
Horstmann: I think we are currently experiencing a huge global digitization surge. Corona has accelerated this: digital weaknesses came to light like under a magnifying glass. If we look back to 2020 in the future, we will surely conclude that there have been as many changes this year as there have been any other decade.
The possible uses of microelectronics are growing almost every week. Think of the quality of cell phone cameras, the precision of medical devices, or the vanishing latency that comes with 5G / 6G. This opens up completely new fields of application and markets. Or think of everyday robotics, where we are still at the very beginning. In short, I don’t see any signs of the return of the pig cycle, the foundries in particular are too differentiated for that.
c’t: The EU expressly wants to set up a chip fab for the production of particularly fine structures. Could Globalfoundries also manufacture structures smaller than 12 nanometers in Dresden in the future?
Horstmann: Nanometers are not an end in themselves – especially when the single-digit nanometer digits are driven by marketing. The real core of the European debate is: How do we ensure that our industries do not get into trouble because we do not have crisis-proof access to the chips we need? The events of the last few years show that this crisis resistance is no longer given.
However, the depth and breadth of European industry will have completely different needs than 7, 5 or even 3 nanometers well into the next decade – according to the motto “be smarter, not just smaller”. We have geared our Dresden Fab to this: Three strong platforms (22FDX and later 12FDX, 28SLP, 40/55) with differentiated functionalities such as high voltage (HV), high frequency (RF), radar, 5G / 6G, analog-to-digital converter ( BCD), non-volatile memory (NVM), MRAM, RRAM, micro-LED, Edge AI, encryption, secure payment and image sensing (ISP). We believe that a Dresden technology corridor from 55 to 22/12 nanometers will provide solutions for 90 percent of the chips that European industry will need well after 2035. We have the solutions ready or in the pipeline that will be needed in the next 15 years.
c’t: The European chip manufacturers Infineon and STMicroelectronics as well as some industry experts criticize the EU plans to locate a fab for particularly fine structures in the EU. How do you see it
Horstmann: The debate about “2 nanometers by 2030” distracts us from important current tasks. I welcome the EU Commission’s ambitions in microelectronics. It has recognized that European skills and production capacities in this area are systemically relevant. For many years the focus was only on research and development, which is now changing after a dozen years of discussions in Berlin and Brussels. However, I don’t see that the demand from Europe for chip structures below 10 nanometers can fill a fab. Semiconductor factories are only successful if they have an “economy of scale” and produce with very high capacity utilization. The European demand is mainly in the areas of cars, Industry 4.0, the energy transition, 5G / 6G, AI / machine learning. Here we can further strengthen or establish new European positions with production capacities well above 12 nanometers.
c’t: What are the most important cost factors that make chip production more expensive in the EU than elsewhere?
Horstmann: That can be summed up in a nutshell: energy, regulation, labor and social costs. However, we can compensate for such location disadvantages by “doing better”, for example through higher quality and productivity. A few percent more yield per wafer compensates for some additional costs.
c’t: Globalfoundries operates its own gas-fired power plants with partners to supply power and heat to Fab 1. Is a higher CO2 tax a problem? How will EU plans for climate neutrality affect 23 or 28 years from now?
Horstmann: We are in the process of fundamentally modernizing our energy supply (electricity, heating, cooling) and making it future-proof. The aim is to reduce CO2 by orders of magnitude. We are also examining whether hydrogen can be used in the plants in the long term.
c’t: Intel boss Pat Gelsinger is said to have discussed a higher funding rate for the settlement of new chip fabs in the EU with EU internal market commissioner Thierry Breton. Would that be helpful?
Horstmann: Compared to the rest of the world, the German and European budgets for microelectronics are indeed rather modest. However, I see the will in the federal government to significantly increase its commitment. At the moment I’m more concerned about the speed and consistency in implementing the industrial policy strategy. We get bogged down in optimizing our processes. A stronger focus on results is urgently needed.
c’t: There is a lot of discussion in Germany about “chips for cars”. However, such automotive chips do not have a large share of the overall semiconductor market – how much is it roughly?
Horstmann: You’re right: Apple alone consumes more chips in a year than the entire global auto industry. Automotive is one of the fastest growing areas and is becoming more and more important for us in Dresden. At Radar, for example, we cooperate with Bosch.
c’t: How long does it usually take “from the order to the finished chip” for your products?
Horstmann: That depends on various factors: If the product is already on the line and fully qualified, then we can deliver additional wafers after about eight weeks – provided we have the necessary capacity. At the moment we’re running at 100 percent capacity, so it looks different. If the product is new, it will take six to twelve months; if special certifications are required, as is the case with automotive, it can take longer. And if new production facilities are required, we have to assume significantly longer times.
In c’t 14/2021 we show you how you can surf without being bothered by cookies and trackers. c’t editor Mirko Dölle found out how Apple’s AirTags can be remodeled as a stalking kit. We have also upgraded the Raspi as a backup server, shed light on the technology and infrastructure for card payments and explain what you should pay attention to after UMTS has ended. You will find issue 14/2021 from June 18th in Heise shop and at the well-stocked newspaper kiosk.