Multicultural Diversity Workforce and Global Technology Collaboration Empowered Semiconductor Manufacturing Excellence in Taiwan: A Manufacturing Engineer’s Perspective

Taiwan is an island about the size of Maryland off the coast of mainland China with a population of about 24 million. Taiwan has the same official language (Mandarin) and same cultural roots as China. In 1949, the civil war in China brought 2 million people with a variety of Chinese cultures to Taiwan, where traditional Chinese values and business philosophies were kept [1]. The diversity in culture and acceptance of cultural differences were deeply rooted in Taiwan’s society.

With support from the United States (US) Agency for International Development (USAID) in the 1950s and 1960s [2] and foreign investments such as Philips from the Netherlands and Texas Instruments (TI) and Radio Corporation of America (RCA) from the US in the 1960s and 1970s, Taiwan has transitioned from an agriculture-based economy to a labor-intensive manufacturing economy. To further advance to a technology-driven, value-added economy, a top-down, bold, and risky industrial policy was made in 1974 by then Minister of Economic Affairs of Taiwan Mr. Yun-Suan Sun, an electrical engineer educated at Harbin Institute of Technology in China, that semiconductor manufacturing would be a key direction of Taiwan.

This policy on semiconductor manufacturing was consistently carried out over the next 50 years. The Industrial Technology Research Institute (ITRI) [3], a government-sponsored applied research organization established in 1973 and located in Hsinchu, Taiwan, was assigned the task of starting semiconductor manufacturing from scratch [4,5]. In 1976, a 10-year technology license agreement was signed between ITRI and the RCA in the US for the complementary metal-oxide semiconductor 7.5 μm technology node [6], which was three generations behind the state-of-the-art at that time. ITRI engineers were selected and trained at RCA on semiconductor manufacturing. In 1977, a 3-inch wafer integrated circuit (IC) production line was established in ITRI [4]. In three months, the yield (a quantitative measurement of quality) of this production line was better than that of RCA, which later left the semiconductor business. High yield has been the hallmark of Taiwan’s semiconductor manufacturing excellence since then.

With a successful industrial policy and modest but consistent research and development (R&D) support from the government (rather than a large one-time funding and catchy slogans from politicians), the semiconductor industry in Taiwan continues the relentless R&D, adopts the best technology in the world, welcomes international culture, and remains low key to become a world leader in both yield and technology in semiconductor manufacturing. Three key pillars supporting this success in semiconductor manufacturing are good government policy, global technology collaboration, and multicultural diversity. Such success in manufacturing has greatly transformed Taiwan’s economy and improved people’s standard of living [7]. This is a wonderful story that “multicultural diversity” is part of the transformation of a manufacturing industry and opens many opportunities for the prosperity of society.

Hsinchu is not only the location of the original ITRI and the first IC production line in Taiwan but also the site of Hsinchu Science Park [8], where the Taiwanese government provides tax incentives and prepares the land with infrastructure (road, electricity, water, etc.) for high-tech companies (mostly in the semiconductor industry) to lease. Taiwan Semiconductor Manufacturing Corporation (TSMC) was an early tenant of the Hsinchu Science Park. Adjacent to the Hsinchu Science Park are two leading research universities, National Tsing Hua University (NTHU) and National Yang Ming Chiao Tung University (NYCU), which provide talent—a critical part of the supply chain. Many international and domestic semiconductor design and manufacturing companies have converged around Hsinchu. A unique semiconductor manufacturing cluster with exceptional manufacturing talent has been created in Taiwan.

Manufacturing talent is the most critical part of this cluster. People make the difference in Taiwan’s semiconductor manufacturing excellence. The high concentration of manufacturing talent and networks among world-leading companies and universities in semiconductor design and manufacturing make Hsinchu an ideal site for semiconductor manufacturing R&D. It mimics the high concentration of artists in Paris and New York [9] and the cluster of innovative automotive R&D companies around Detroit [10]. This success story of continuous and consistent policy in Taiwan is a role model for countries determined to play a role in semiconductor manufacturing.

Without the Covid-19 pandemic and worldwide shortage of IC chips, this success story and excellence in semiconductor manufacturing might still be unknown to most of the world. Due to the shortage of IC chips, the supply chain bottlenecked, factories idled, and inflation rose [11]. The average lead time for IC chips increased from 10 to 12 to over 25 weeks [12]. Compounded with the Ukraine war, concerns of geopolitical conflict between Taiwan and China, and the US policy prohibiting the export of advanced semiconductor technology for weapons to China, there is a great desire for the US and many others to have a strong domestic semiconductor manufacturing capacity. Industrial policies for semiconductor manufacturing, such as the US CHIPS and Science Act, European Union (EU) European Chips Act, South Korea K-semiconductor Strategy, and China 14th Five-Year Plan, have been created for domestic semiconductor manufacturing to ensure supply chain resilience and national security. This major shift in manufacturing national policy is still evolving.

In this paper, the multicultural diversity and global technology collaboration for semiconductor manufacturing in Taiwan is first introduced. The foundry model for contract manufacturing of semiconductors, the technology node and assessment of causes for failures, and challenges working in semiconductor fabs are then presented. Two models for excellence in manufacturing and three pillars for semiconductor manufacturing excellence are discussed. Finally, building a workforce and future research in semiconductor manufacturing are outlined.

Culture is a critical element in manufacturing [13]. In the 1980s, the US recognized that Japan automakers had better quality, lead time, launching products on time, and overall operational excellence than US automakers. There were two groups at Massachusetts Institute of Technology (MIT) and University of Michigan (U-M) studying the Japanese automotive industry. Toyota stood out within the Japanese auto industry. As a result of this study, the MIT team published a book entitled “The Machine That Changed the World: The Story of Lean Production—Toyota’s Secret Weapon in the Global Car Wars” [14] which outlined the success of lean production and quality (techniques learned from the US after World War II) in Toyota. The U-M team has published a series of books, for example [15–17], by Professor Jeffery Liker on the Toyota Production System, lean culture, lean leadership, continuous improvement, working with little inventory, designed in quality, safety, and ergonomics. Industry around the world has adopted lean principles and achieved significant success in manufacturing operations. The semiconductor industry in Taiwan is no exception. The continuous improvement and lean principles of TPS are the foundation in semiconductor manufacturing operations.

Another example is the manufacturing of internal combustion engines and cars in the early 20th century. The core automotive technology was invented in Europe (Germany, Italy, and France). Innovations of the transfer line by Henry Ford in the US brought drastic advancements in productivity and quality, compared to traditional European factories, and elevated the US automotive manufacturing to the top of the world.

A hypothesis for the success of Taiwan’s semiconductor manufacturing is attributed to the global technology collaboration as well as the diversity and balanced mixture of the US, Japanese, European, and traditional Chinese cultures in the workforce. This is elaborated in Secs. 2.1–2.4 and summarized in Sec. 2.5.

US culture and technology are leading in innovations and excellence in (1) IC design, (2) software for IC design, (3) higher education, (4) professional training, and (5) IC manufacturing equipment.

The US dominates IC design. In addition to Intel and Apple, two leaders in IC design, Table 1 summarizes the top 10 fabless IC design companies, which are dominated by the US.

The US also dominates the Electronic Design Automation (EDA) software for IC design. The top three EDA software companies in 2021 are Synopsis (32% market share), Cadence (30%), and Siemens EDA (13%) [19]. Siemens EDA was Mentor Graphics, which was acquired by Siemens in 2017. The cost to design an advanced technology node is high. For 5 nm, the cost is about US$542 million dollars in comparison to $298 million for 7 nm, $174 million for 10 nm, $106 million for 16 nm, and $70 million for 22 nm [20]. The cost for design IC with technology nodes below 3 nm will be even higher. The fabless IC design needs to work with a manufacturing foundry two to three years in advance before the start of production of advanced technology nodes. Such simultaneous engineering between fabless and foundry is a role model on partnership in design and manufacturing [21].

The US is the global leader in higher education, with outstanding research universities which innovate technologies and educate leaders in engineering, science, business, law, etc.—all essential in semiconductor manufacturing. The research of the most advanced semiconductor manufacturing technology is mostly confined to the industry. There are enormous potentials and opportunities in manufacturing research and education partnership with the semiconductor industry to change this paradigm. This will be elaborated in Secs. 7.2 (on education) and 8 (on research).

US companies led the world in professional training of job-related skills and company-wide initiative. The Six-Sigma program, with the green-belt and black-belt certificates of General Electric, is a good example [22]. Intel also has a great professional training program in semiconductor manufacturing.

US, Japan, and the Netherlands are three dominating countries in semiconductor manufacturing equipment, as shown in Table 2. Semiconductor manufacturing industry is very conservative and rarely changes suppliers. There is no Taiwanese company in the top 10. All major equipment for semiconductor manufacturing in Taiwan is imported. Almost all semiconductor manufacturing equipment is shipped by airfreight. Speed is essential in the semiconductor manufacturing operation to be discussed further in Sec. 4.

US culture greatly influences Taiwan in technology and education. US companies design the IC and provide the software and equipment to Taiwan’s fabs. Taiwanese students are educated in the higher education mimicking that of the US. Many leaders in Taiwan’s semiconductor industry have advanced degrees in the US and working experience in the US semiconductor industry.

Japan ruled Taiwan for 50 years (1895–1945). Japanese culture is adopted in Taiwan. Many Taiwanese and Japanese are well connected personally and culturally.

Japan is a world leader in high purity materials for semiconductor manufacturing. Four examples are the photoresist (especially for extreme ultraviolet (EUV) lithography), hydrogen fluoride etchant, silicon wafer (Shin-Etsu Chemical and SUMCO as top two suppliers in the world), and advanced 3D-packaging materials. Japan is also outstanding in semiconductor equipment, as evident in Table 2.

Four unique Japanese cultures [24] are:

• Kodawari (pursuit of perfection): Commitment, insistence, and attention to details on everything. Small things (such as cleanness) can make a big difference in yield.

• Consensus on and commitment to goals: Taking a long time to build the consensus on goals. Highly committed and dedicated to achieving the goal once it is set.

• Punctuality: People’s time is well respected.

• Ichi-go ichi-e (one time, one meeting): Treasure and cherish the time and opportunity of a meeting.

Such Japanese culture is well respected and deeply rooted in some Taiwanese. A workforce with some members with such Japanese culture in dedication and care is important to achieve high yield in semiconductor manufacturing.

The Netherlands occupied southern Taiwan in 1624–1662 [25]. Spain occupied northern Taiwan in 1626–1642 [26]. Taiwan is also known as Formosa, which is “beautiful island” in Portuguese.

The Netherlands especially has a close relationship with Taiwan. Philips and ASML are two examples. Philips setup manufacturing operations in the early 1960s in Taiwan and had a major operation with over 40,000 employees [27]. More importantly, Philips brought in key manufacturing technologies, such as the precision mold making and injection molding for plastics, to Taiwan and elevated Taiwan’s technology level in manufacturing. Philips is also one of TSMC’s original investors and the only one from outside Taiwan. ASML is the world leader in lithography and a key partner of TSMC in the immersion deep ultraviolet (DUV) and EUV lithography equipment [28]. ASML and TSMC grow together to advance the technology node in semiconductor manufacturing.

Taiwan has many outstanding small and medium-sized enterprises (SMEs), also known as “Hidden Champions” (Mittelstand in Germany) [29], in the manufacturing supply chain. These manufacturing SMEs are dedicated, dynamic, agile, highly innovative, high-tech, and well-networked—a business culture found in Hidden Champions in Germany, Switzerland, and northern Italy. Many SMEs in Taiwan are specialized in their domain of expertise, innovating technologies, and becoming the backbone of the semiconductor manufacturing industry. Hundreds of these outstanding SMEs in Taiwan are suppliers with timely service and reasonable cost across all semiconductor manufacturing operations such as the equipment installation, equipment repair and maintenance service, materials supply, fab and clean room design and construction, and many other specialties. These SMEs in Taiwan are specialized in their domain of expertise, innovating technologies with customers, and becoming the backbone of the semiconductor manufacturing industry.

Every year, Taiwan’s Ministry of Economic Affairs honors the Mittelstand Award to top Taiwanese SMEs. Many past awardees are in semiconductor manufacturing. For example, Gudeng Precision, a past awardee, co-developed with TSMC the vacuum EUV reticle pods to reduce the contamination and improve yield in production.

Another culture of Japan and Germany that is rooted in Taiwan is the love, respect, and successful examples of manufacturing, especially among SMEs. Table 3 summarizes the manufacturing share of gross domestic product (GDP) of Taiwan, the group of seven (G7) and EU, as well as three major manufacturing countries (China, South Korea, and India) in descending order. Taiwan’s manufacturing share of GDP is the highest, 32.9%. In addition to semiconductor foundry, many hidden champions of manufacturing in Taiwan contribute to such high activity in manufacturing. Many manufacturing SMEs are family businesses in Taiwan. The second or third generation of family leaders who have personally witnessed the dedication of the first-generation entrepreneurs are well trained (many with international education) and coached to lead their family SMEs.

Taiwan is fortunate to keep traditional Chinese culture without being influenced by the first cultural revolution (1966–1976) and the on-going second culture revolution (1992–date) [30] of rapid modernization and the changing political environment in China. A strong work ethic is one of the hallmarks of traditional Chinese culture. The Hsinchu area has a high concentration of Hakka, a Chinese ethnic group with the tradition of dedicated work ethic. ITRI, TSMC, and the semiconductor manufacturing supply chain in Hsinchu all benefit from such dedication to work in fabs and labs in early years.

The culture of traditional Chinese merchant guilds [31,32] was kept in Taiwan. China has 10 great merchant guilds namely: Shanxi, Hui, Western Zhejiang, Ningbo, Dongting, Jiangxi, Guangdong, Shaanxi, Shandong, and Fujian. Many of these merchant guilds moved to Taiwan in 1949. These successful merchant guilds were trustworthy and grew business together with their customers. This is reflected in TSMC’s values and business philosophy [34].

TSMC’s values and business philosophy [33] are the roadmap for success in semiconductor manufacturing. The four core values of TSMC are Integrity, Commitment, Innovation, and Customer trust (ICIC), which existed since the start of TSMC in 1987. ICIC is also a guiding principle for recruiting new TSMC employees. The 10 key business tenets of TSMC are (1) integrity, (2) focus on core business: IC foundry, (3) globalization, (4) long-term vision and strategies, (5) treating customers as partners, (6) building quality into all aspects of business, (7) unceasing innovation, (8) fostering a dynamic and fun work environment, (9) keeping communication channels open, and (10) caring for employees and shareholders, and being a good corporate citizen.

For example, TSMC’s treating customers as partners [33] and stated “Since the company was founded, we have treated our customers as partners and have never competed against them. This policy is the key to our current success and will be crucial to our continued growth. At TSMC, customers come first. Their success is our success, and we value their ability to compete as we value our own.” The focus in never competing with customers is the foundation of TSMC’s success in semiconductor manufacturing.

Culture is critically important in manufacturing operations [34]. Taiwan’s unique history and location as the intersection and melting pot of US, Japanese, European, and traditional Chinese culture has built a diverse and capable workforce in semiconductor manufacturing. Not all engineering graduates in Taiwan want to work in semiconductor fabs in Taiwan. The Human Resource has an important task to identify, recruit, and select manufacturing talent to build a workforce [35]. There is only a small pool of recruits who excel in the fab environment and have the personality to attend to details and pursue perfection (Japanese culture) and the ability to work with a network of high-tech and specialized SME suppliers (European culture of hidden champions), and partner with equipment, materials, and software companies from the US, Japan, and Europe. Over the years, manufacturing talents were educated, selected, and trained with values and business tenets (traditional Chinese culture). This generation of workforce with multicultural diversity is a role model for other countries to emulate in semiconductor manufacturing.

There are two competing business models in the semiconductor industry. One is the integrated device manufacturer (IDM), which integrates the IC design and manufacturing within the same company. At the inception of the semiconductor industry, companies designing new IC chips were required to build their own fabs for manufacturing. The Mead-Conway chip design [36] revolution in 1978 enabled the EDA software and made the foundry model possible. As Morris Chang, founder of TSMC, said “the design part could be separated from the technology. But he didn't advocate the advent of pure-play foundries, but he did make the point which would lead to the conclusion that you could start up a pure-play foundry” [37].

Another business model is the “fabless + foundry” model which entails the fabless IC design companies partnering with a foundry, a contract manufacturer, to manufacture their IC designs. In 1987, TSMC started as a pure-play foundry focused only on semiconductor manufacturing. The start of TSMC’s foundry service matched well with pioneering fabless IC design start-ups such as Adaptec (founded in 1981), Altera (1983), Xilinx (1984), Array Technology Inc. (1985), Qualcomm (1985), and NVIDIA (1993) [6]. TSMC provided the critical foundry manufacturing service for these start-ups. The “fabless + foundry” approach has sparked the creation and growth of more fabless IC design start-ups and transformed the semiconductor industry.

In logic IC, Intel has been the leader since the invention of the 4-bit 4004 microprocessor as the central processing unit (CPU) in 1971 [38] and dominated the IDM. In the past two decades, the fabless (e.g., Qualcomm, Broadcom, Advanced Micro Device (AMD), NVIDIA, and MediaTek) and foundry (e.g., TSMC, GlobalFoundries (GF), United Microelectronic Corp (UMC), Semiconductor Manufacturing International Corp (SMIC), and Powerchip Semiconductor Manufacturing Corp (PSMC)) together are gaining market share [6,39]. The market share of the top 10 foundries [40] is listed in Table 4. TSMC has the largest share (58.5%) of the foundry market. Altogether, Taiwan has 66.9% of the global foundry business. The top 10 customers of TSMC, as listed in Table 5, are dominated by US companies, including the device maker (Apple), fabless design companies, and even IDM (Intel). TSMC business with China on advanced technology nodes was stopped by US export controls.

In memory and automotive IC, IDM is still the dominating business model. The top three memory IC companies (Samsung, SK Hynix, and Micron) and top five automotive IC companies (Infineon, NXP, Renesas, Texas Instruments, and STMicroelectronics) are all IDM. These companies all invest heavily in their fabs and only utilize foundry (for about a third of the revenue in 2022) when necessary. It is uncertain if the foundry model can be successful in memory and automotive IC manufacturing in the future.

IDM companies can also provide foundry services. Intel and Samsung offer foundry services in addition to being IDMs. Both need to maintain a complex business relationship, which has collaboration (in foundry business) and competition (in product design) with their customers (in manufacturing) and competitors (in product), respectively. For example, Intel is competing in four fronts with (1) NVIDIA on high-performance computing graphics processing unit (GPU) for generative artificial intelligence learning, (2) AMD on the x86-architect CPU, (3) Apple, Qualcomm, and others on the Advanced RISC Machines (ARM)-based CPU for mobile devices, and (4) TSMC, GF, UMC, and others on the foundry service. The fabless + foundry versus IDM in semiconductor manufacturing, particularly for memory and automotive ICs, continues to evolve [39].

The technology node [42] was defined as the transistor’s gate length, which was considered as the minimal feature length. As the gate length reached 45 nm, the definition of technology node was changed to a representative number of a new generation of the logic IC chip. A shorter distance in a technology node represents a smaller feature size and a higher density of transistors in the logic IC. Another parameter that quantifies the miniaturization of the IC is the metal-pitch, the distance between the center of two metal lines in the IC. According to the semiconductor roadmap of Interuniversity Microelectronics Centre (IMEC), an international R&D organization in semiconductor technology, the current metal-pitch is about 40, 28, and 22 nm in the current 7, 5, and 3 nm technology node [43]. IMEC roadmap predicts an end of the metal-pitch at about 12–16 nm, while transistor and material innovations continue to evolve in advanced technology nodes [43].

Table 6 illustrates the companies which can achieve production at various technology nodes. From 180 nm to 130 nm, seven companies (ADI, Atmel, Rohm, Sanyo, Mitsubishi, ON, and Hitachi) could not advance. At 90 nm, six companies (Cypress, SkyWater, Sony, Infineon, Sharp, and Freescale) failed to advance. Renesas, Toshiba, Fujitsu, and TI did not advance to 45/40 nm, with only 10 companies in the US, Japan, South Korea, Taiwan, China, and EU left. In 32/28 nm, Panasonic, STM, and HLMC could not advance and ended the Japanese and European presence in advanced logic IC manufacturing. IBM, a pioneer of semiconductor, failed to advance in 22/20 nm volume manufacturing. GF and UMC decided not to further invest below 10 nm. Intel, Samsung, and TSMC were in three-way competition in 10 nm and 7 nm. In 2023, Samsung and TSMC remained as two competitors in volume production of the state-of-the-art 3 nm technology node with Intel planning to jump two technology nodes to join the competition.

Companies failed to advance the technology node due to yield and profit. IBM sold its non-profitable semiconductor manufacturing to GF in 2014 and focused only on semiconductor R&D. Panasonic sold its money-losing semiconductor business to Nuvoton (Taiwan) in 2019. Due to the high cost of EUV equipment and R&D, GF and UMC decided to stop advancing below the 10 nm technology node.

Leadership, determination, decision, and people are the root causes of failures in semiconductor manufacturing in Table 6. The government and its industrial policy play a key role in the semiconductor industry due to the large investment and long payback period. Leaders in government with the vision, execution, communication (with citizens and the legislative branch), and consistency in good policy (for decade-long support) are essential to build a profitable domestic semiconductor manufacturing industry.

Accurate and quick decisions must be made in the competitive semiconductor manufacturing industry. These decisions are often very technical, high-stakes, risky (with many technical and business unknowns), and associated with significant financial consequences if a wrong decision is made. Decisions must be made swiftly and decisively because the technology is rapidly advancing. Examples of landmark decisions include Intel’s refusal to make CPUs for Apple iPhone in 2006, AMD’s separation of design and manufacturing to create the GF in 2009, TSMC’s decision to develop its own 180 nm Cu process instead of licensing the technology from IBM in 1997, and ASML and TSMC’s partnership to co-develop the immersion DUV lithography in 2002. These key decisions have long and significant impacts on the success of a company. Outcomes of some key technical decisions, such as Samsung’s gate-all-around transistor for 3 nm technology node, and business decisions, like Intel’s IDM 2.0 to compete in the foundry business, are still unknown.

Manufacturing engineers are critically important to achieve high yield, recommend the correct schedule, provide detailed analyses of risks and investments, and execute the plan in operations for business leaders in semiconductor manufacturing. There are often conflicting recommendations with different estimates, views, and levels of optimism in manufacturing. Business leaders need to select and trust a recommendation and make high-stakes decisions. This is very hard for business leaders without in-depth background in manufacturing and extensive operation experience in the semiconductor industry. Honesty and integrity are critically important for manufacturing engineers making recommendations. A good guideline is the description of the Integrity in TSMC’s Business Philosophy [44], which states that: “Integrity means: 1) We tell the truth. 2) We believe that the record of our achievements is the best proof of our merit. Hence, we do not brag or boast. 3) We do not make commitments lightly. Therefore, once we make a commitment, we devote ourselves completely to meeting that commitment…”

Speed is critical in semiconductor manufacturing because of the high cost to build a fab of advanced technology nodes and the need to quickly turn capital investment into revenue [45]. For example, the 3 nm fab is estimated to cost US$20 billion dollars. For TSMC, a new plant in the fab is strived to complete in one year in Taiwan (other countries may take longer). This requires coordination of construction and facility to move in hundreds of machines per month in an operation schedule spanning 24 hours a day and 7 days a week (so-called 24/7).

Capacity is also critical in semiconductor manufacturing. Customers demand rapid delivery. Fabs need to have the equipment and facility (both need extensive capital investment) ready for customers once a product contact is signed. For example, the number of EUV equipment already in the fab (and the yield in operation) determines the quantity of IC that can be produced per month for advanced technology nodes.

A trend in the foundry business is that the number one leader in the market will have the lion’s share of the market (e.g., TSMC in Table 4). Both fabless and IDM need to use the most advanced manufacturing technology to achieve the best performance of their IC and be competitive in the market. Without using the most advanced manufacturing technology, the IC products of fabless and IDM are not competitive in the market, even with good designs. Number two foundry will have some market share. The profit margin is much lower for number three. Intel, Samsung, and TSMC are fiercely competing for the top in semiconductor manufacturing.

Yield in production determines the cost and profitability of the fab. Yield is a real technical challenge for manufacturing engineers in advanced technology nodes because of the complexity of IC design (e.g., hundreds of layers), processes (e.g., chemical mechanical planarization (CMP) and diamond dressing for every layer), and equipment (e.g., multi-patterning DUV and high numerical aperture (NA) EUV equipment). Examples of yield issues in semiconductor manufacturing include the launch of Samsung’s 3 nm gate-all-around transistor and Intel’s struggle in EUV lithography volume production.

Manufacturing talent makes the difference for high yield by solving issues in production. The base equipment is the same for the semiconductor manufacturing industry. Three keys to build a workforce of manufacturing talent include recruiting and retention (in competition with the high pay job in software companies), education, and training (e.g., Intel’s training program and TSMC Newcomer Training (NCT) Center). These will be elaborated in Sec. 7.

Semiconductor fabs have four groups of manufacturing engineers in R&D, process, equipment, and operation. Most semiconductor manufacturing R&D follows Intel’s tick-tock product development cadence or its variants [46]. Two groups work in parallel on two advanced technology nodes ahead of production in the so-called mother fab. For example, when TSMC, Samsung, or Intel’s 5 nm is in production, a R&D group in the mother fab is working on the upcoming 3 nm line and another R&D group is working on the 2 nm line simultaneously. Once the process is developed and yield is adequate for initial production, the equipment and process knowledge in the mother fab are moved to the production fab, which, for TSMC, can be in the US, Japan, or another city in Taiwan. Some R&D manufacturing engineers move from the mother fab to the production fab to setup the equipment and continue improving the yield of the production line.

The process engineers setup and optimize the production process of the equipment for different IC designs in the foundry. In contrast to the IDM model, the foundry requires frequent setup of equipment. In IDM, once a production line is qualified for a specific product, there are minimal changes to the process. Once a product is obsolete, equipment in the IDM production line may be sold or repurposed. On the contrary, process engineers in the foundry have a more difficult job due to the frequent setup changes and requalification of the equipment for production. Top process engineers excel in the foundry model by learning from each equipment setup to gain valuable knowledge in process and equipment. Such knowledge is necessary to optimize processes for new IC designs.

An example of such frequent equipment setup is the number of test wafers (also known as dummy wafers) used to setup the equipment (to maintain similar thermal and flow conditions inside the processing chamber) from one IC design to another. Such test wafers are usually the reclaimed wafers from prior test runs. TSMC, for example, uses over 1 million test wafers per year for equipment setup. Test wafers are an essential part of the foundry operation for equipment setup.

Equipment engineers ensure all equipment is operational with minimal defects in fabs with 24/7 operation. There are two distinct approaches on working with equipment suppliers. One approach is to modify the base machine with additional technologies and keep the equipment and process knowledge confidential. Another approach is to work closely with equipment suppliers as partners. Service engineers from the equipment supplier would be on-site to provide critical support to the operation, especially during the equipment setup and limited production phases. Once the yield of the equipment in operation is stable, service engineers from equipment suppliers need to be on-call 24/7 in case there are issues in the fab. Regardless of the approach, the knowledge of the optimized production process and equipment is a treasure and always kept confidential.

Operation engineers (or systems engineers) integrate, synchronize, manage, and optimize the flow of the entire fab production operation. The automation, cleanliness, quality control, robotics, supply chain, system design, capacity planning, system balancing, logistics, purchase quality assurance, predictive maintenance, production planning and control, cyber security, sustainability, supplier management, etc. are all system-level tasks and critical for the yield, operation, and circular economy of the fab. The adoption, training, and implementation of data analytics and manufacturing system technologies are critical and have revolutionized the fab operation.

The continuous improvement in yield during the ramp-up is critical for the delivery, capability, and profitability of semiconductor manufacturing operations. Quality control, identifying and solving equipment and operation problems, and learning from defects and errors are examples of key issues that affect the ramp-up time and productivity. Thousands of engineers and technicians working coherently as a team 24/7 to improve the yield to launch a new technology node is common and critical in semiconductor manufacturing.

For process, equipment, and operation engineers striving to learn and improve, the foundry model has an advantage over the IDM model on the knowledge of equipment for yield improvements and process/operation optimization. Taiwan has a cohort of such knowledgeable, capable, and dedicated manufacturing engineers for specific semiconductor manufacturing equipment and operation as the backbone of its workforce.

Jobs are tough in semiconductor fabs. The discipline, concentration, and team-coordination are essential and often military-like. The lifestyle of manufacturing engineers in semiconductor fabs is different from the normal 8-hour per day and 5-day per week work routine [47]. A common schedule in the semiconductor fab is working two consecutive 12-hour days followed by two consecutive days off. This is repeated year-round. The 12-hour work inside the fab requires wearing the cleanroom suit, an overall garment also known as a “bunny suit” in the Class 10 to 1000 cleanrooms [48]. The fab cleanroom is an isolated (without easy phone and internet access to outside) and demanding working environment with major responsibilities to keep a high yield year-around.

Taiwan has trained manufacturing talent and established a generation of manufacturing engineers that excels in semiconductor manufacturing. Three examples exemplify the dedication and capability of this manufacturing talent. Taiwan is prone to earthquakes, which is a risk for fabs [49]. On 21 September 1999, a 7.3 Richter scale earthquake devastated Taiwan and shut down production in all semiconductor fabs. The response was swift and impressive. Most fabs were running near full capacity in just 10–14 days. The second example is the traffic jam near the foundry after each earthquake in Taiwan, because engineers rushed to take care of their equipment in the fab. Metrology equipment needs to be recalibrated after earthquakes with a Richter scale of 3 or higher. The third example is TSMC’s Nighthawk Force project in 2014, on which the 24/7 R&D sped the introduction of the 10 nm technology node. Key suppliers also needed to team up to provide the 24/7 support.

A manufacturing job in the fab is analogous to working in a dairy farm, which requires discipline, selflessness, teamwork, care, and other attributes. In comparison, jobs at other high-tech companies (e.g., Amazon, Apple, Baidu, Google, Microsoft, and SAP) have a more regular schedule and possibly higher pay. Developing manufacturing talent and building a dedicated workforce by recruiting, selection, education, and training is essential for a competitive semiconductor fab operation. This will be discussed in Sec. 7.

The late Prof. S.M. Wu of the University of Michigan in Ann Arbor wisely summarized three key characteristics of globally competitive manufacturing companies. One is knowledgeable leadership who possess in-depth knowledge of products and manufacturing processes. The second is empowered workers, who have adequate education and training and are empowered to take responsibility for the performance of the process. The third is superior technology—technology that is the most effective in improving both the product and process.

While serving at the Office of Advanced Manufacturing (OAM) in the National Institute of Standards and Technology (NIST) in 2017, the author analyzed the per capita manufacturing GDP of key industrialized countries and identified seven outstanding countries: Japan, Switzerland, US, Sweden, Germany, Denmark, and Netherlands. There are four common characteristics of these countries:

• Respect for hands-on skills in the society and dedication of skilled technicians and engineers in manufacturing.

• Solid manufacturing education from high-tech technicians to PhDs.

• Manufacturing high value-added products.

• Outstanding industry leaders and government leaders in manufacturing national policy.

For example, in the US, Alexander Hamilton (1757–1804) who wrote the REPORT ON MANUFACTURES is a role model for government leaders in manufacturing national policy.

The three pillars government policy, multicultural diversity, and global technology partnership for semiconductor manufacturing excellence are illustrated in Fig. 1.

Two government leaders played critical role in Taiwan’s success in semiconductor manufacturing. In addition to Mr. Yun-suan Sun, who established ITRI on applied research in 1973 and made the key policy decision to focus on semiconductor manufacturing in 1974, the other leader is Mr. KT Lee, who studied Physics at the Cavendish Lab in the University of Cambridge. As Ministers of Economic Affairs and Treasury, he established the Hsinchu Science Park in 1980 and guided Taiwan’s government policy in science and technology. Together, they recruited leaders to ITRI, trusted them, and helped spin-off semiconductor technology developed in ITRI under government funding to become TSMC, UMC, PSMC, and others. The photo in Fig. 2 shows an ITRI group trained at RCA in 1976. Most of them graduated from ITRI and became successful leaders in the semiconductor industry. Morris Chang was recruited from TI/General Instrument to become ITRI’s President in 1985. He recognized that the high yield and manufacturing talent were true competitive advantages in Taiwan, and a pure-play foundry focused only on IC manufacturing could be profitable. He founded TSMC in 1987 and continued in a leadership position in ITRI until 1994.

The coordinated development of ITRI and TSMC is a successful example of an industrial policy. Based on ITRI’s 3-inch wafer line in 1977, IC products could be found in digital watches, singing Christmas cards, etc. In 1984, the UMC was a spin-off from ITRI with both IC design and foundry services. TSMC was established in 1987 with a 48.3% share from the Taiwanese government (less than 50% to avoid becoming a state-owned enterprise), 27.5% technology share from Philips, and the balance from several Taiwanese major conglomerates encouraged by the government to invest (24.2%). TSMC leased ITRI’s 6-inch wafer production line (as TSMC’s Fab 1) with 2.5 and 3 μm technology nodes (two generations behind the state-of-the-art at that time [6]) to start the IC production as a pure-play foundry in 1987. In 1988, TSMC started to make a profit. In 1989, TSMC moved to its new 6-inch wafer fab (Fab 2, which is still in production as of 2023) in the Hsinchu Science Park and continued the race toward advanced technology nodes in semiconductor manufacturing.

Taiwan is the most unique case and the benchmark for the successful industrial policy in semiconductor manufacturing. The industry policy in Taiwan also avoids the protection of the local semiconductor industry or intervention of companies’ operations. Over-protection hurts the competitiveness of the government-sponsored companies. This is a true test of the wisdom of government policy makers who manage the funding and timing of government subsidies to decide when, and for what, to end government support. The Taiwan government sold most of its share (keeping only 6.38% as of 2023) once TSMC went public. Philips and all Taiwanese local investors sold all their TSMC shares by 2006. Most of the original investors had high confidence on TSMC—an indication of challenges for Taiwan’s industrial policy makers who guided the creation of TSMC as well as early leaders of TSMC. Currently, some countries have an industrial policy to create a semiconductor manufacturing capability for high yield 24/7 production of IC chips in advanced technology nodes, which advance every 3–4 years, require extensive investments, and need to payback the investment from sales of a lot of IC chips.

There is no genius in semiconductor manufacturing. Great leaders in semiconductor manufacturing require extensive experience and continuous improvement. A workforce in semiconductor manufacturing needs to be built methodically over years. Three essential steps—evaluation/selection, education, and training—are required to establish a high-performance semiconductor manufacturing workforce.

Changing a person’s cultural tendencies is difficult. Selecting the right person (with the potential to be educated and trained) to embrace cultural diversity in semiconductor manufacturing is the foundational first step to building such a workforce. People have different talents [50]. The traditional grade point average, standardized test, personal essay, and reference letters are not the best way to identify potential manufacturing talent anymore. New evaluation methods need to be developed to recruit the right people, particularly for people with the aptitude to excel in semiconductor fabs. One potential approach is to apply advanced evaluation methods based on the technology- and data-driven survey methodology, the data analytics for evaluation, and a database linking the performance to survey data [51].

During the Covid-19 pandemic, online lecture materials and video recording of lectures for key undergraduate and graduate manufacturing courses have been developed. The flipped classroom pedagogy has been tested [52]. Video-enhanced pedagogy coupled with hands-on labs for experiential learning has been adapted [53]. All these have paved the way for innovations in semiconductor manufacturing education, which has the dynamic and hands-on advanced manufacturing curriculum to teach the breadth and depth of knowledge at the undergraduate and graduate levels, respectively.

In undergraduate education, there is a lack of connection to semiconductor manufacturing-related processes, equipment, and materials. New topics connecting semiconductor materials and processes, as well as pedagogy innovations to connect to semiconductor products, need to be developed.

In graduate education, there are many outstanding graduate courses taught by talented and dedicated instructors. But these courses are not distributed to create bigger, broader impacts in graduate manufacturing education. New courses related to semiconductor manufacturing processes (such as advanced packaging and metrology) and systems (such as data analytics and automations) still need to be developed. A majority of engineering workforce in manufacturing advanced technology nodes requires a minimum of master’s education in Taiwan. Personnel cost is small in comparison to the cost of fab and value of high yield. China and Taiwan have both established Semiconductor Colleges for focused graduate education.

There are three opportunities for the semiconductor manufacturing education reform:

• Cyber-based modular semiconductor manufacturing curriculum: The manufacturing course can be divided into modules. Each module has lecture materials and lecture video, which can be delivered online.

• Product-oriented pedagogy: The traditional manufacturing lectures are based on processes or systems. This can be adjusted to be based on products (e.g., semiconductor, battery, electric vehicle, etc.) by selecting and integration of modules from processes and systems related to products. Many students adopt better in learning based on a specific product because they can see the physical product.

• On-demand learning [54]: The modular approach can tailor the course material to meeting a student’s individual needs. Personalized exams and homework problems can be created for evaluation.

Training of professional skills is fundamental for new or transferred equipment, process, and operation/system engineers. The US was the leader in professional training, but company budget cuts in professional training since the 2000s has greatly weakened the US’ competitiveness in manufacturing. The augmented reality, virtual reality, and mixed reality-based training and digital twins will be a critical part of the curriculum.

In Taiwan, TSMC’s NCT center is the model for hands-on semiconductor manufacturing professional training. NCT has state-of-the-art production equipment and a group of full-time training instructors, all with years of fab experience. Before entering the fab, all new TSMC manufacturing engineers are required to be trained for months at NCT to gain the depth of knowledge on working effectively in the fab.

Manufacturing research is essential for innovations of the next-generation heterogeneous IC fabrication and packaging. Semiconductor manufacturing is an integration of technologies from all manufacturing disciplines as well as cross-disciplinary integration with material science, physics, and chemistry. Topics of research changes, but the fundamental knowledge and problem-solving approaches in manufacturing remain the same.

Customer demands drive the technology advancements and research in very high precision and production volume. This has distinguished semiconductors from other products in manufacturing research in innovative equipment and high yield production. Lithography dominates the evolution of semiconductor manufacturing to achieve the small feature size. Advanced lithography equipment is expensive and complex to operate. The immersion DUV equipment may cost about US$60 million and takes a team of four engineers to operate. The high NA (0.55 versus traditional 0.33) EUV equipment may cost US$400 million and take a team of over 100 engineers to constantly monitor and adjust the production operation centering around this machine. Such a level of cost for equipment and complexity of operation is the epic of advanced manufacturing.

Another key semiconductor manufacturing research area is in advanced packaging of IC chips to a heterogeneous device and be placed in a printed circuit board. An example of packaging an advanced CPU with over 200,000 Cu pillars (for power and data signals) is to connect to several thin (about 20 μm in thickness) redistribution layers made of low dielectric constant (Dk) and dissipation factor (Df) Ajinomoto Buildup Film (ABF) or glass material. Research to adopt new materials in manufacturing processes (e.g., joining and CMP) and metrology (e.g., X-ray based in-line and post-process inspection) are all critical to advance the yield in manufacturing and performance of microelectronics (e.g., speed, transistor density, functionality, etc.).

Fundamental knowledge is the same for research in semiconductor manufacturing. An example is illustrated in the review of assembly of microelectronics for semiconductor manufacturing [55]. Topics such as the design of 3D packing for manufacturing, electro-physical and chemical processes for deposition of Cu, forming and joining of many miniature Cu pads and thin ABF/glass films in 3D packaging, abrasive dressing and polishing in CMP, precision machines to fabricate lens and components for EUV equipment, inspection of surface quality and defects, optimization of complex fab production operations, and standards for measurements are all essential in the current and future semiconductor manufacturing R&D.

Environmental impacts and sustainability of semiconductor manufacturing are important topics. Hsinchu has the highest level of acid rain in Taiwan [56]. Due to the growth of the industry, high energy and water consumption, chemicals for extreme cleaning requirements, and other needs for advanced technology nodes, semiconductor manufacturing could account for 3% of total greenhouse gas emissions by 2040 [57]. A detailed study on semiconductor manufacturing’s impact on the environment requires information from fabs on materials and operation information. IMEC has started the semiconductor technologies and systems program [57] to gather critical information from fabs. A vision paper on the green transition of the IC industry [57] has been published.

This paper analyzed the semiconductor manufacturing excellence in Taiwan, described the consistent industrial policy, and hypothesized that the good government policy, global technology collaboration, and diverse mixture of US, Japanese, European, and traditional Chinese cultures were three pillars for such success. TSMC’s value and business philosophy were highlighted as a roadmap to success in semiconductor manufacturing. Manufacturing talent in the semiconductor industry was recognized as the most critical part of the success in Taiwan.

South Korea is another country representing semiconductor manufacturing excellence. There is a great generation of semiconductor manufacturing engineers in South Korea. It will be valuable to benchmark and compare South Korea’s paths to achieve their success in semiconductor manufacturing [58]. The success stories of great generations of semiconductor manufacturing engineers in Japan in the 1970s and in the US from 1970 to 2020 are equally worth further studies.

China accelerates Taiwan’s transformation to advanced manufacturing. The manufacturing industry in Taiwan is sensitive and greatly affected by the neighboring China, a powerhouse in manufacturing with the same official language. Many Taiwanese manufacturing sectors folded or moved to China. When China can mass-produce a specific product at a lower cost with equal quality, it is a sign for Taiwan to stop manufacturing that specific product. But Taiwan did not retreat from the competition in manufacturing with China by moving to the seemingly easier service industry. Over time, Taiwan’s manufacturing industry has elevated the technology and service levels to co-exist with China by producing high value products with better services, e.g., quick turn-around time and trust (e.g., the ICIC of TSMC). An analogy is—if China manufacturing is like the big trees, then Taiwan manufacturing will be orchids under the trees. Taiwan’s transformation to advanced manufacturing in the shadow of the fierce competition from nearby China and Korea is another good future study of the government policy.

A key goal of this study is to encourage government policy makers and manufacturing engineers that the workforce and excellence in semiconductor manufacturing can be established. It will not be easy. Most countries have a much better position than Taiwan in 1974. In a speech entitled “The CHIPS Act and a Long-term Vision for America’s Technological Leadership” by the US Secretary of Commerce Gina Raimondo [59], she stated that “If we don’t invest in America's manufacturing workforce, it doesn’t matter how much we spend … It starts with training and inspiring a generation of engineers and scientists who are excited about manufacturing.” Focused and consistent efforts from leaders in government, industry, and academia are necessary to co-create good policy, multicultural diversity, and global collaboration—three legs in Fig. 1—for steadfast and continuous (rather than short-term spikes) in R&D and workforce development to advance the technology node-by-node without dreaming shortcuts in semiconductor manufacturing.

The author is deeply indebted to the support from NTHU leaders, especially Professors Hocheng Hong and Hung-Yin Tsai, and the Yushan Scholar Program. Many anonymous manufacturing engineers who work in the semiconductor industry and provide their insights of Taiwan’s semiconductor industry are greatly appreciated.

No data, models, or code were generated or used for this paper.