Asia Pacific (Japan, Korea, Taiwan, China, and Singapore) accounts for 84% of global chip exports and is home to 10 of the 16 semiconductor exporters and the top six suppliers, which includes Taiwan Semiconductor Manufacturing Company (TSMC). The leading pure-play foundry will celebrate the inauguration of its first fab in Kumamoto, Japan, on February 24, 2024.
Why is semiconductor production so concentrated in the countries in East Asia? Why was TSMC willing to form its Japanese joint venture but decided to have a stand-alone greenfield investment in Arizona to build advanced chips? Can TSMC successfully replicate its pure-play foundry business model in its second fab in Japan?
DIGITIMES Asia held an exclusive interview with Dr. Richard L. Thurston, the Founder and Principal Member of RLT Global Consulting and CEO of Hudson Valley Fast Fab ("HVFF"). Thurston retired from TSMC as Senior Vice President and General Counsel in 2014, served as a consultant for TSMC, and worked for Texas Instruments (TI) to negotiate joint venture and trade agreements with Japan in the 1980s and 1990s.
Along with DIGITIMES advisor Dr. Albert Lin, Dr. Thurston will also share his insights on Japan and whether it still has what it takes to succeed in the Angstrom era on the first DIGITIMES Asia GeoWatch forum, scheduled on March 27 Taipei time.
Q: TSMC will inaugurate its first fab in Kumamoto and has disclosed investment plans for a second fab there. Singapore is said to have re-pitched with even better incentives to attract investments from TSMC. Based on your experience in Japan and your work at TSMC, why is semiconductor manufacturing so successful in this part of Asia? Will TSMC be successful in replicating its pure foundry model in Japan?
In Japan, pure-play foundries have not yet been successful, and making this business model work in Japan will be challenging. In South Korea, Samsung has been able to do some foundry work with its excess capacity, but not as a pure-play foundry: it is still, principally, an IDM. Singapore has foundries because of the legacy of Chartered Semiconductor (acquired by Globalfoundries), UMC, and SSMC, a 1998 TSMC joint venture with Philips/NXP (I was on the SSMC Board of Directors from 2002-2013). But, of course, no one has been as successful as TSMC with the pure-play foundry model. There is a distinction between pure play and others that do some foundry work.
I joined Texas Instruments in 1984 as Asia-Pacific Legal Counsel. We operated several manufacturing plants in Japan, Taiwan, Singapore, and Malaysia. Therefore, I've been in a unique position to observe the successful development of the semiconductor industry in these Asian countries. A lot of it comes down to culture. It comes down to the focus and discipline that the corporation creates. It's the workforce, and it's also the customer base.
Quite honestly, I think that creating a pure-play foundry in Japan is going to be very difficult. Based on what I have read, TSMC is not creating a pure-play foundry but a customer-driven/dedicated facility. It is a well-thought-out hybrid joint venture. Sony requires more advanced, high-performance CMOS image sensors to drive its growing markets. This need should drive TSMC's Kumamoto facility.
There are many, and perhaps 10-20 factors attributable to TSMC's success, including technological breakthroughs, skills at transferring R&D from lab to mother fab, better yields, and focused customer engagement, to name a few. Examples of leading-edge customer engagement include Apple, Nvidia, Qualcomm, AMD, TI, Sony, Fujitsu, etc.
I lived on the ground in Tokyo, Japan, for Texas Instruments for 3.5 years, fought the trade war battles, and negotiated with the Japanese government and businesses – it was and remains a much different world within which to manufacture and sell semiconductor devices. When Rick Tsai was TSMC's CEO, we explored different manufacturing opportunities in Japan during 2005-2009. Fujitsu was probably the most promising because of its focus on advanced technology, and we explored a Joint Venture approach with them. We also explored the pure-play foundry model, but the Japanese government blocked that. They did not want TSMC to run a foundry in Japan. Why? Partly because of the culture, their mindset, the baggage of their manufacturing legacy, and the Japanese Government's heavily nationalistic industrial policy.
TSMC's approach with Sony is the right one to take: to focus on one or more joint ventures with local customer engagement.
Q: Perhaps the Japanese government's thoughts have changed after the pandemic. And right now, they are very eager to have their semiconductor industry revive again. Huge investments followed TSMC's announcement to build the first fab in Kumamoto in 2021, and many fabs in Japan are commencing mass production this year. Why did TSMC form a joint venture in the Kumamoto fab starting as a minority shareholder instead of a wholly owned subsidiary?
TSMC is very smart to continue doing joint ventures. Chairman Liu and CEO Wei understand that they don't have to put 100% of the equity capital into an investment. They can maintain control through technology, intellectual property licenses, other contracts, veto rights, etc. They can expand capacity smartly, reducing the risks of not bleeding themselves dry financially. They do not have to do all the heavy lifting in a foreign country with which they're not as familiar as in their home country. We had discussed this model often in the past. Although TSMC Japan has many good people who know Japanese culture very well, it is difficult for any foreign company, let alone a Chinese-ethnic-based firm, to be successful in Japan as a standalone. It doesn't work from a cultural standpoint, and we were well aware of that at TI.
Q: Yes. Well, then, why did TSMC choose to do it alone in Arizona? Should they have chosen to do a Joint Venture instead? Would that have been more conducive to progressing the fab?
Allow me to say that I don't know the specific nature of any of the discussions TSMC had with Federal and State governments or with any of their customers. But while I was there, in light of the Founder's concerns over the history of WaferTech, we always had talked about doing a standalone foundry in the United States. Part of this discussion was because of the bitter taste left in some people's mouths, especially that of the Founder. You will recall that TSMC formed a joint venture in Washington State with three partners in 1996, but that did not work for many reasons. Therefore, TSMC quickly exited that joint venture relationship to create the stand-alone foundry in the WaferTech operation. The Arizona initiative originated because most of the TSMC management team had thought a lot during 2005-2010 about where and when to expand production in the US. That is why we pursued multiple talks with IBM to acquire its Microelectronics Division, and the last and most serious talks occurred during 2012-13. It was all about setting up an advanced lab that we could control and manage much more successfully than a joint venture. While I was involved, Arizona became a location initially out of political "suggestion" with the lure of significant subsidization.
Q: You said the Japanese government was difficult. However, the semiconductor companies here in Taiwan were perplexed by the attitude of the American government because they said so much about the importance of chip resilience but did not help to accelerate the building of the fabs and allocating funds early. That's why I think a lot of companies, including TSMC, felt that they probably have misinterpreted geopolitical situations.
The dynamics in Washington, DC, sometimes are fast to change, but most often extremely slow in the follow-up. Meanwhile, American paranoia has caused an unusual focus on the most advanced process technologies rather than equally important legacy (something less than most advanced) technologies. Washington, DC, became overly focused on pushing TSMC to build its most advanced processes and linked large subsidies to that process node. And, unfortunately, with all the different agencies competing out there for limited funding and with 2024 election politics that have been ongoing for a while, monies have been diluted significantly – up to now (perhaps it might change). I think that altogether 39 states have gotten some CHIPS money through different tech hubs that may have diluted the pool too much. How many of those are going to be successful? I would like to say all of them, but I don't believe that.
Also, there has been some misreading of the tea leaves. It is hard to say who has been at fault. I believe TSMC has a strong and capable American Government Relations manager. MediaTek has a good person, too. It's somewhat confusing because they get different stories based on who they talk with on the "Hill." Keep in mind that the CHIPS office is like a startup. The CHIPS office doesn't have a year under its belt despite all of the bureaucracy going into it. They just started to compete with METI in Japan, Korea's MOTIE, the Singapore EDB, etc. There's always bound to be some miscommunication. And when too many politicians and lobbyists get their hands on these important funds, it becomes confusing. There has to be constant recalibration. You're correct, the important and larger companies that are part of the US supply chain need to be treated better than they have, at least publicly. Hopefully, a recalibration will occur, and we will receive a more positive outlook sometime in March 2024.
Q: You are one of the very few non-tech experts in the semiconductor industry, and you have accumulated experience helping countries and companies build up research and development hubs. What would you advise the Taiwanese companies to do to navigate the geopolitical uncertainties in the future?
Thank you for your compliment. Unfortunately, one must have several crystal balls. Japan is perhaps one of the more difficult countries to deal with consistently and successfully. Singapore is somewhat less complicated, although very bureaucratic in its way. The local team, whether in Japan, Singapore, or South Korea, must have a good balance between local and 'foreign" management. In Japan, a foreign company with only Japanese management cannot get its foot in the door of Japanese government organizations. Sometimes, a more assertive American or European management presence is needed.
For example, when I worked for TI, Jerry Jenkins (then Chairman, President, and CEO) took my advice and sent me and two other American managers to live in Japan and to assist with trade and commercial negotiations. We were ultimately successful in opening the market, negotiating trade agreements and joint ventures, and increasing TI sales in Japan from around US$200 million to US$1.6 billion by the time I left TI.
I spent a lot of time handling Japanese government relations successfully, and one of my best friends went on to become the vice minister of the National Police Agency and worked with me over the years. Norm Neureiter, who followed me, was also very successful. No one will ever get into the inner sanctum of the government. However, effective communication is essential, and it is crucial to have a strong government relations team that is on the job 24/7.
With Sony as a core partner, TSMC is doing it right. It's a good selection. Fujitsu and Renesas are also good potential partners. Companies must have full-time government relations and staff that manage it, understand it, and support the country manager and fab manager. A multicultural presence is a must-have to succeed in Japan and Arizona as well.
For TSMC, I had recommended to Chairman Chang and other senior TSMC management that we should send Rick Cassidy from the United States to Japan – when we first considered manufacturing in Japan. A strong Taiwan-based manager was sent instead, and he had some success. We should not deny that a cultural competition/divide requires more experienced foreign advisers to the local team. Regardless, an important element behind success is constant communication with government agencies – make them feel that they are part of the venture to understand your goals and objectives and never embarrass them.
