Acer founder: It’s too far-fetched to manufacture chips in the US

Stan Shih, the godfather of Taiwan's computer industry and founder of Acer Group, said today (1) that industrial division of labor is a major trend. Although the United States hopes to break the development of globalization and move semiconductor manufacturing back to the United States due to national security concerns, this goes against the general trend of industrial paradigm shift and may not be feasible. If it is done reluctantly, it will be relatively inefficient and lack competitiveness.

Shih Chen-jung said that two major events happened in the semiconductor industry last week. One was that Intel's market value was surpassed by Advanced Micro Devices (AMD), and the other was that the United States passed the chip bill. The subsequent inspiration and impact it brought to Taiwan and the semiconductor industry are worthy of everyone's attention.

Shih Chen-jung pointed out the difficulties that American semiconductor giant Intel will face when it enters the wafer foundry industry from the smile curve. He pointed out that the shift of industry paradigm towards vertical division of labor is an inevitable trend, and even the world's strongest semiconductor company cannot stop the shift of industry paradigm.

Shih Chen-jung said that an article published in Harvard Business Review in 1991 mentioned that the world was moving towards the development of "Computerless computer companies" (computer companies without computer factories) and "Fabless semiconductor companies" (semiconductor companies without wafer fabs).

Shih also pointed out that the paradigm shift in the personal computer and semiconductor industries was initiated by Acer and TSMC. In the personal computer field, Acer launched its own brand in 1983 and started a new business model of ODM; in the semiconductor field, TSMC was founded in 1987 and started an innovative business model of professional wafer foundry.

Shih Chen-jung pointed out that after the Harvard article was published, Intel did not agree and still insisted on moving towards a vertical integration model. After 30 years of evolution, even the world's strongest semiconductor company could not resist the paradigm shift and was defeated, and its market value was surpassed by its competitor AMD.

Shih Chen-jung said that the US chip bill was based on the perspective of US defense and national security, so it is necessary to develop semiconductor manufacturing in the US at all costs. However, if we consider the economic and industrial development, the US's competitiveness is not in manufacturing. "No matter how strong it is, it can't be stronger than the weakest link." Manufacturing is the weakest link in the US, and long-term development is not feasible. There is a big gap in competitiveness compared to Asian manufacturing.

Shih Chen-jung has been involved in the semiconductor industry for more than 50 years (his first job was at Universal Electronics in the semiconductor packaging industry, and later when he founded Techtronic Semiconductor, he worked in the DRAM industry). Based on his many years of industry observation, he pointed out that, taking the memory industry as an example, the business cycle is obvious, and the capital demand for investment equipment is huge. When making money, they will expand investment, which often leads to oversupply in the industry, and then they will stop investing, resulting in undersupply. This is the characteristic of the US capital market. In the long run, the United States has been surpassed by Japan and South Korea, which continue to invest regardless of the business cycle.

He further analyzed that the reason why Micron is still competitive in the memory field is that it has carried out industrial division of labor with Taiwan. Therefore, in the high-tech industry, the United States develops technological innovation and outsources manufacturing to Taiwan partners, which can create the competitiveness of many American international brand manufacturers.

Shih Chen-jung said that wafer manufacturing also needs industrial division of labor to be competitive, and it needs economies of scale and diverse products for customers to fully load production capacity. Although Intel hopes to enter the field of wafer foundry, customers may worry that their products will conflict with Intel. If Intel only wins a few customers, it will not be enough to make the wafer foundry business successful. Industrial division of labor is a competitive ecosystem, just like Acer's successful improvement in competitiveness after the century change in 2000, when Acer and Wistron were separated.

Regarding AMD's successful market value surpassing Intel, Stan Shih pointed out that AMD separated its manufacturing department into GlobalFoundries, which focused on the development of processor design. In the beginning, the two parties were tied together, which made them uncompetitive. Later, the two parties separated and developed independently. AMD also found TSMC for foundry, following the trend of industrial division of labor. Its competitiveness was improved, and it grew and declined with Intel.

The U.S. Senate and House of Representatives passed the Chips and Science Act last week, finally integrating the semiconductor and technology research and development industry revitalization plan that had been debated and negotiated for many years through bipartisan consensus. It is expected to be sent to the White House this week and take effect after being signed by President Biden.

From the perspective of policy subsidies, the Chips and Science Act demonstrates the highest level of attention from both the government and the public in the United States to the semiconductor industry. It directly subsidizes high-end chip manufacturing with $52.7 billion, including certain rewards for software such as design and research and development. In addition, it provides a 25% tax credit for foreign semiconductor manufacturers such as TSMC and Samsung to set up factories in the United States.

Another part of the bill is scientific research grants, with a total amount of up to 200 billion US dollars. The main player is the National Science Foundation (NSF), which received a huge amount of funds of 81 billion US dollars to reward scientific research and development in robots, artificial intelligence, high-speed computer computing, and of course semiconductors. It claims to ensure the United States' global leading position in advanced technology, and has instructed the NSF to formulate special projects to cultivate the most cutting-edge scientific research talents starting from elementary school.

The US government regards semiconductor manufacturing as the most critical competitive industry, and links it with national security and US-China competition, revealing that the US political and military institutions are highly concerned about the backwardness of the US semiconductor industry. The Chips and Science Act reflects this sentiment. Although the United States is the place where computer chips were invented and dominated the world before the millennium, its chip manufacturing has been surpassed by TSMC and Samsung in recent years. In addition, the Chinese government has invested in semiconductor manufacturing with the whole country's strength, which has made the US government and industry suddenly wake up like a sleeping lion, and fully subsidizing semiconductor manufacturing has become the biggest consensus in the US political arena.

How significant the actual effect of the chip bill is remains to be seen. Take TSMC's investment of US$12 billion to build a factory in Arizona as an example. From the Trump administration's request for TSMC to build a factory there, TSMC conducted business and policy negotiations, proposed a specific factory construction plan in 2020, and started the factory construction project in full swing in 2021. It has been more than 3 years since then. Although Trump and Biden both encouraged TSMC and Samsung to build factories in the United States, the relevant federal bills have been delayed. Even if the two parties have a consensus, some members of the Congress have continuously proposed new versions of the bill. Different bills have been negotiated back and forth between the Senate and the House of Representatives. The US president has made a lot of big words, but has been unable to implement specific policies.

If the Democratic Party had not faced a political crisis in the November midterm elections, which could have resulted in a major defeat, forcing the Democratic Party leaders to take steps to integrate the House and Senate members who were each singing their own tune, and ultimately Senate Majority Leader Chuck Schumer to strongly integrate the multiple semiconductor and technology bills that had been proposed, the U.S. federal government's semiconductor revitalization plan would probably have continued to be delayed. By the November midterm elections, all the House members would have been replaced and one-third of the senators would have been replaced, and everything would have to start all over again. However, TSMC's plan to officially start production in 2024 would continue as usual. In comparison, everyone can see that the U.S. government's problem is that it talks more and does less, and that symbolism is more important than substance.

In addition, the consensus among all sectors on the budget of US$52.7 billion is that it is a drop in the bucket. This year, the total capital expenditure of the global semiconductor industry is close to US$200 billion. TSMC's capital expenditure this year alone will exceed US$40 billion. The company has announced a three-year plan of US$100 billion. The U.S. federal government's budget of more than US$50 billion, which will be divided into five years, to subsidize the entire semiconductor industry is really small compared to the industry's own investment scale. Even with the increase in state governments, it can only have an encouraging effect at most, and cannot have a structural impact on the technological advancement of semiconductors and the competitive situation of the industry.

The U.S. industrial subsidy bill, which is much ado about nothing, has taught Europe, Japan and all governments that claim to control the semiconductor industry a clear lesson. The technology and industrial development of semiconductors have long exceeded the limits that governments and politicians can influence. Countries need to strengthen their national defense and control semiconductors for national security. Traditional tools such as rewards and forcing industry investment will only produce half the result with twice the effort. The government must truly understand the core of the semiconductor industry, establish long-term and stable partnerships with industry leaders such as Intel, Samsung and TSMC, and discuss feasible solutions to ensure confidentiality and security. This is the side effect and real significance of the Chip Act legislation.