Wafer size is critical in the return on investment (ROI) and cost calculations of the foundry business model. For high-volume devices, where the process is well established and the emphasis is mainly on PPAC (Power-Performance-Area-Cost), the solution would be to simply go for higher wafer diameters. Using large diameters also enables large dies to be manufactured while limiting edge losses. Bosch and Infineon have recently built 300mm Si wafer fabs not only to further reduce their costs, but also to maintain their leadership. For example, Infineon, in addition to its existing factory in Dresden (Germany), invested €1.6B in another 300mm wafer factory in Austria. Bosch’s and Infineon’s new 300mm fabs are mostly designed to fabricate IGBTs and MOSFETs for the automotive market. Cree | Wolfspeed is now moving to SiC 8’’. Sapphire and GaAs could follow suit in years to come but, first, sapphire and GaAs need to be fully established at 6’’ which is not yet the case. These advanced 300mm Si foundries cost several billion dollars and need large volume demand to get a positive ROI. However, less known is the existence of a market for depreciated fabs which also represent tremendous opportunities to produce devices at a lower cost for any volume. e.g., 20-year-old 6’’ Si foundries are now fully depreciated and can operate at a much lower volume and still be profitable. There is still a handful of facilities for Si 6’’ which are mostly used by the MEMS industry, for low-volume applications. Globally, there are still many foundries retaining small-size wafer capabilities. This analysis is well detailed in the new semiconductor manufacturing report released by Yole Développement (Yole) this month: 6’’ and Below: Small-Dimension Wafer Market Trends.
Yole estimates that a little more than 35% (~800) of existing fabs can still process 1 to 6’’ wafers worldwide (both Si and MtM materials). In general, advanced technology node (<28nm) devices can be fabricated in brand-new fabs offering state-of-the-art cleanrooms and ultra-precise equipment. For other non-mainstream applications, which have low demand and/or are based on less-advanced technology nodes (>28nm) and where the process changes are rapid, depreciated fabs could be adapted. With the increased use of non-Si substrates, more and more players are offering specific processing equipment for a given material and/or equipment that can adapt to different materials and diameters, including 4’’ and 6’’. This equipment can be astutely used in fully depreciated sub-6’’ Si wafer fabs, creating opportunities for small and large players to offer highly competitive device fabrication cost. In addition, it is economically better to buy an existing fab rather than to build a new one from scratch.
These are reasons why wafer fabs need never die!
ATREG, as a unique service provider in the fab ecosystem, plays a big role in facilitating fab and manufacturing line dispositions to keep such facilities immortal.
Yole recently met with Stephen Rothrock, Founder, President & CEO of ATREG, to discuss today’s fab ecosystem, ATREG’s services as well as the company’s plans for the coming years.
This interview was conducted by Gaël Giusti, PhD., and Vishnu Kumaresan, PhD., Semiconductors & Manufacturing Market Analysts at Yole.
Gaël Giusti (GG): Could you please introduce ATREG, its history, and current activities?
Stephen Rothrock (SR): Certainly, Gaël! ATREG was founded in 2000 and is headquartered in Seattle, Wash. We are the go-to global firm for the sale or acquisition of advanced technology infrastructure-rich manufacturing assets, including semiconductor fabs and cleanrooms. During the first six years of our existence, we focused exclusively on 200mm, mainly for Japanese companies, which established our global reach right out of the gate. We were the first company to transact on 300mm fabs in 2010 and sell an operational 300mm fab in 2011. Over time, we started adapting to new market needs and facilitating transactions on other types of manufacturing assets such as A&T, advanced packaging, and OSAT, as well as adaptive re-use of facilities into data centers. But ATREG offers much more than just transactional assistance. We advise global advanced technology companies on their overall, long-term manufacturing strategy, so they remain cost-effective amid a very competitive global industry.
Over the last 20 years, ATREG has completed close to 40% of all global operational wafer fab sales in the semiconductor industry. We work with some of the world’s largest and most reputable companies in the semiconductor, assembly & test, display, and electronics sectors around the globe, including ADI, Fujitsu, GLOBALFOUNDRIES, IBM, Infineon, Matsushita (Panasonic), Maxim, Micron, NXP, ON Semiconductor, Sony, Qualcomm, Renesas, Texas Instruments, and Western Digital to name just a few. Our complex transactions encompass land, buildings, intellectual property (IP), complete tool lines, and business unit elements (supply contracts, workforce, etc.) We leverage our unique blend of deep market knowledge, extensive industry relationships, and global transaction expertise to help clients drive strategic decisions, achieve their transaction objectives, and maximize ROI for our clients.
GG: We believe that economically, it is more profitable to buy an existing fab rather than building a new one from scratch. What is your take on this hypothesis?
SR: Yes indeed, in most cases, it will make more sense for a semiconductor company to buy an existing brownfield fab available on the market, and then upgrade the tools or convert it, depending on the geometry and the products to be manufactured. The relevance of brownfield fabs has increased because of the growing need to accelerate time-to-market for certain products.
A good example is the former TI 150mm / 200mm fab based in Greenock, Scotland (GFAB). We facilitated its sale to Diodes in Feb. 2019 who integrated all of TI’s current fab operations and workforce. Under a multi-year wafer supply agreement, Diodes continues to manufacture TI’s analog products until TI transfers into its other existing fabs. The majority of the tools in this facility were also convertible to 8” and TI had already begun some of the tool conversion. This made the site very attractive to companies looking for 8” production, as Diodes was at the time.
On the flip side, if local tax incentives and government grants are available, a greenfield fab is a strong option, for example, the Cree | Wolfspeed 200mm SiC greenfield fab being built on the Marcy Nanocenter in Upstate New York in the U.S. It is expected to be finished in early 2021, with production starting in January 2022. Foreign companies such as TSMC also want a piece of the greenfield cake, announcing plans for a 5nm fab in Arizona with support from that state and the U.S. federal government. The facility will supposedly have a 20,000 wafer-per-month capacity and create over 1,600 jobs directly and thousands more indirect jobs.
GG: There is a consensus among equipment suppliers that equipment types available on the secondary market are an indication of trailing-edge technology. We observe that most of this available used equipment is for 6” wafer diameter. From a fab point-of-view, do you think there are more fabs that are disposed of at 6” and sub-6” diameter ranges?
SR: It depends on what you are comparing it against, in terms of either 8” or 12” fab transactions. When you compare operational fab transactions from 2010 until now, there have been just over 50% more deals at the 8” wafer size than at 6” and below. This is partially due to the significant consolidation that we’ve seen within the industry, coupled with the significant cost advantages 8” holds over traditional 6” fabs that leads to an increased buyer pool for 8” facilities. It’s also important to note that there are fewer operational 8” fabs than there are 6” and less, so as a percentage of total fabs, 8” transactions are much more frequent. 12” fab transactions have actually been quantitatively less frequent than 6” and sub-6” transactions over the last decade. However, as there are even fewer 12” fabs than 8” fabs, the percentage of fabs that have been sold is still higher than at the 6” and less level. The low cost and high efficiencies obtained at 8” and 12” means there is much more demand for those fabs if they become available, which is likely a driving force behind why there is a higher percentage of them sold.
GG: Following up from the previous question – the ON Semi fab that is up for sale in Belgium is an 8”-convertible, 6” manufacturing line. Do you see the future of 6” and sub-6” fabs is to increasingly go towards 8”
SR: ON Semiconductor has indeed engaged ATREG to help dispose of its operational, automotive-qualified Oudenaarde manufacturing facility. Nearly 71% of the fab’s 320 available front-end 150mm tools are 200mm-convertible, but this conversion will carry a significant cost for the facility’s buyer. The Flanders Government is proposing some very attractive incentives to help finance that conversion.
Several 6” transactions that have occurred in the last decade have involved 200mm-convertible equipment lines, and it often makes a facility significantly more attractive for traditional analog / logic / power production sites. However, nearly half of the existing fabs at or below 6” contain some sort of specialty product (II-VI, III-V, MEMS, or other) which often don’t require conversion. Wide-bandgap is only recently coming into volume production at 6” and very few companies are pursuing this at 8” (notable leading-edge being Cree | Wolfspeed’s greenfield fab currently being built in the Mohawk Valley in Upstate NY).
GG: In your opinion, are there many depreciated fabs with well-established processes that are ‘good enough’ for their end-application that they do not have to keep upgrading to the latest technology nodes? Could you please cite some examples?
SR: In ATREG’s experience, the long-term viability of processes at 6” and below varies significantly in different fabs. As mentioned earlier, specialty fabs have seen strong performance at the 6” level where companies such as Cree | Wolfspeed, II-VI, and others continue to operate. However, for CMOS, BCD, and other analog or power products that are transferrable to 8”, we have seen that even extremely well-run 6” fabrication facilities are not cost-competitive with their 8” counterparts. This has been a driver for much of the movement towards 8” fabs.
Vishnu Kumaresan (VK): ATREG is uniquely positioned in the fab ecosystem. How do you see your service offering evolving as the global semiconductor industry continues to shift over the next five to 10 years?
SR: It is true that ATREG serves a very specific niche of the semiconductor market assisting with both “front-end and back-end“ fabs. Our experience, knowledge of fab sales, and extensive industry relationships have been the cornerstones of our strength for the past 20 years. We know the global advanced technology manufacturing asset market inside and out and it is the reason why so many semiconductor companies come to us to ensure the successful divestment and acquisition of their manufacturing assets. No other company has our long-standing experience and market expertise to do this kind of specialized work. As long as there are manufacturing assets changing hands worldwide, ATREG will be there to assist and evolve its service offering to meet the needs of the advanced technology industry.
VK: In general, what do you think is the health of the 6”-and-less fab ecosystem?
SR: We have observed two main drivers within the 6”-and-less fab ecosystem worldwide. The first movement is the push from China to increase their domestic semiconductor production. The government is offering companies impressive incentives to produce locally, however many Chinese companies are behind technologically compared to their leading counterparts in other parts of the world. This has meant that 6” and less is still feasible for traditional technologies that might not be cost-competitive with 8” in other regions. The second driver can be found in the specialty markets that are still reliant on 6”-and-less facilities for the majority of their production. These technologies include MEMS, GaN, SiC, GaAs, and many other compound semiconductors that have more niche applications. I think that until these materials become cost-competitive at larger wafer sizes, there will remain a strong demand for 6”-and less-facilities.
VK: In your opinion, do the current disruptive technology trends (AI, IoT, 5G, etc.) push the 6”-and-less fab ecosystem?
SR: It is indisputable that AI, IoT, 5G, and the like will push the technology further, and therefore the 6”-and-less fab ecosystem. One area where we see 6”-and-less production being driven by those is the automotive segment as electric vehicles (EVs) continue to grow. Wide-bandgap semiconductors are frequently being used in EVs for their higher efficiency and operating temperatures based on two leading technologies in this field – SiC and GaN. These technologies are currently being manufactured at 6”-and-less wafer sizes and most production will likely continue at this size for the foreseeable future. As the demand for smart cars increases, so will their components and compounds to meet the ever-increasing demand from consumers.
VK: What is the impact of COVID on the fab ecosystem and semiconductors in general?
SR: This question is actually timely as we recently featured a guest column on this very topic in our latest newsletter authored by Christian Dieseldorff, Senior Principal Analyst for Semiconductors at SEMI. At the beginning of COVID, there was obviously a lot of worry about what effect it would have on demand and production. But like many in the industry, we’ve been surprised at how strong semiconductor has performed over the past six months, especially memory and foundry. On the production side, fab operations were very rarely affected as near all were deemed essential operations, and companies have continued to view fabs with a long-term outlook as they consider how to move forward with their operations.
VK: What message would you like to pass on to our readers from the semiconductor industry?
SR: We too often see companies trying to handle the disposition or the acquisition of a manufacturing asset in-house by themselves. This generally results in a much lengthier than needed process ending up in unsuccessful transactions, leaving fabs on the market for too long, and allowing the selling company little leverage in the disposition process. Semiconductor manufacturing asset transactions are very unique and extremely complex to navigate. You wouldn’t necessarily know it, but a lot of creativity is sometimes required to structure a successful win-win fab transaction for both parties. It takes a lot of trust, patience, market knowledge, and key industry relationships to make this work. ATREG puts its 20-year transaction experience at the disposal of semiconductor companies to ensure the success of their global manufacturing strategy.
Stephen Rothrock founded ATREG in 2000 to help advanced technology companies divest and acquire infrastructure-rich manufacturing assets, including wafer fabs as well as MEMS, solar, display, and R&D facilities. Over the last 20 years, ATREG has completed close to 40% of all global operational wafer fab sales in the semiconductor industry. Recent global acquisitions and sales have involved Fujitsu, GF, IBM, Infineon, Matsushita (Panasonic), Maxim, Micron, NXP, ON Semiconductor, Sony, Qualcomm, Renesas, and TI.
Prior to ATREG, Stephen established Colliers International’s Global Corporate Services initiative and headed the company’s U.S. division in Seattle, Wash. Before that, he worked as Director for Savills International commercial real estate brokerage in London, UK, also serving on the UK-listed property company’s international board. Stephen holds an MA degree in Political Theology from the University of Hull, UK and a BA degree in Business Commerce from the University of Washington in Seattle, USA.
Gaël Giusti, PhD., is a Technology & Market Analyst specializing in Semiconductor Manufacturing and Equipment & Materials at Yole Développement (Yole). As part of the Semiconductor & Software division at Yole, Gaël’s expertise is focused on thin film growth and related applications, equipment, materials and manufacturing processes. He is involved daily in the production of technology & market reports and custom consulting projects. Prior to Yole, Gaël served as a R&D engineer at Sil’Tronix Silicon Technologies for 5 years where he was in charge of upscaling a CVD process to develop epitaxial AlN thin film on sapphire for RF applications. He also worked on transparent conducting thin films for optoelectronics applications as a post-doctoral researcher at LMGP (Grenoble, France). Gaël holds a master’s degree from ENSICAEN (Caen, France) as well as a PhD in Materials Science from the University of Birmingham (UK).
Vishnu Kumaresan, PhD., is a Technology & Market analyst in the Semiconductor Manufacturing Team of Yole Développement, France. He focuses on the semiconductor manufacturing domain, covering both equipment and material segments. His scope includes mainstream microelectronic applications as well as More-than-Moore applications. Having lived and worked in 4 countries, he has more than 11 years of experience in the electronics industry, covering semiconductor, display, and software technologies. Prior to joining Yole, he worked as an epitaxy engineer in an advanced startup in the micro LED display industry and has previously gained corporate experience at IMEC, CNRS, Saint Gobain and Infosys. Vishnu obtained his PhD in Material Physics & Chemistry from Pierre and Marie Curie University, France, and master’s in Microelectronics from National University of Singapore and Technical University of Munich, Germany.
Wafer fabs never die: small-dimension wafer is still a growing market.