From Talent Shortage to Workforce Excellence in the CHIPS Act Era: Harnessing Industry 4.0 Paradigms for a Sustainable Future in Domestic Chip Production

Semiconductor manufacturing encompasses a diverse set of expertise from various talent pools, including fab operators and quality technicians to electrical and materials science engineering PhDs performing specialized silicon design. A skills gap exists across the entire value chain within the industry. The main challenges the semiconductor industry is facing is explored, which is also summarized in Figure 2.

1. 1.

   It is a career path that very few Americans are familiar with due to decades of outsourcing and doesn’t carry the same prestige as software jobs in big tech. In a survey of 18 industry executives from various member companies of SEMI, 60% felt that the entire semiconductor sector has a poor perception compared to other tech companies [7], which tend to be more attractive in brand name. There is also a lack of awareness amongst employees in how dynamic the industry is, bearing the ability to switch to between roles as well upskilling for more lucrative opportunities.

   Even if the US merely focuses on maintaining domestic production for critical applications, this would entail up to 90,000 talented personnel required to staff new fabs [8]. Recruiting the right talent is a prevailing challenge despite extensive private and public investments towards reshoring semiconductor production. Furthermore, the U.S. has access to the largest talent pool across the globe, as it offers the best system of research institutions worldwide with endless opportunities. As a result, majority of international STEM students choose American colleges and universities to pursue their graduate degrees.

2. 2.

   Subsequently, only 46% and 36% of engineering Master’s and PhD students are U.S. citizens, respectively. Limited number of domestic students end up pursuing advanced engineering degrees compared to international students, which spans a multitude of social and economic factors. For majority of domestic students, a Bachelor’s degree in any engineering discipline earned from U.S. universities is sufficient for a rewarding career, particularly when considering the time and capital investment spent towards graduate degree programs and a general lack of awareness about various scholarship opportunities to fund graduate programs.

   To further accelerate the growth of the semiconductor ecosystem, there are three key strategies to remain competitive in the talent war: reskilling, automation, and expanding the pipeline through workforce development programs [8]. Leading Integrated Device Manufacturer(IDM) companies such as Intel, Micron, and TI have made notable strides towards reskilling for both current and future workforce through rotational and educational programs. Nonetheless, deploying all these efforts is a long-term investment that requires major time and capital to produce a substantial outcome, which makes it inefficient in accruing enough staff to operationalize the new fabs. Additionally, the mere investments in building fabrications plants will do little to offset the talent shortage, as it projected that by 2030, the U.S. will experience a shortage of 300,000 engineers and 90,000 technicians [8].

3. 3.

   Talent retention issues. Many jobs within the industry, particularly fab roles, experience low employee retention due to impacting parameters affecting workplace desirability, such as lack of remote option, flexibility, salary, and compensation.

4. 4.

   To meet global chip demands for all market sectors, there is a lengthy learning curve for new or retrained employees to properly execute all the complex steps involved in the fabrication process within a fast-paced fab environment. Even though large portion of semiconductor jobs are manufacturing intensive, majority remain unfilled with high attrition rates compared to design and analytic positions.

5. 5.

   There is an income discrepancy between design engineers and quality engineers/manufacturing technicians working in fabs, as the former is isolated from the hazards of working around risky machinery and chemicals as well as having the ability to maintain a flexible work schedule. The new generation of talent is placing a higher emphasis on hybrid/remote work options, especially after the 2020 pandemic. This is a concern of industry executives as fab roles happen to be less flexible than design counterparts and it is where the acute talent shortage is. However, the industry cannot expect their existing engineers and those in early stages of the value chain to give up better paying, lower stress, flexible desk jobs to be reskilled for demanding fab roles, which can further exasperate this issue as it leads to more attrition and dissatisfaction [8].

Ultimately, leading-edge innovations in Integrated Circuit(IC) chip packaging increases manufacturing complexity, which necessitates more process automation along with expanding the US capacity to hire more workers familiar with automation tools amid a competitive talent war. This specifically applies to technicians who are responsible for monitoring equipment and running recalibrations, especially with advancements towards microscale tool operations. The high capital investment that goes into recruiting, training, onboarding, and retaining employees requires dedicated training tools/programs and automation of repetitive tasks as a suitable option to re-structure the workforce, as it can alleviate the demands for scarce talent in the ongoing competitive war while maximizing workplace desirability. Traditional levers of automation include Automated Material Handling System (AMHS), and advanced dispatching and scheduling, which are older automation mechanisms commonly used in semiconductor fabrication facilities (fabs) processes. Moreover, the increased design and packaging complexities, along with outsourced fabrication and ATP, leaves the chip in immense vulnerability, creating the opportunity for potential adversaries to perform malicious activities by tampering with the chips, such as through Trojan insertion. This presents a need for rigorous testing, assurance, and inspection standards to detect and classify defects and Trojans—which can be a mostly automated process if embedded properly. Since a major target of the CHIPS Act is on chip design and wafer fabrication, automation in general can also garner support for increased funding and investment towards a domestic advanced packaging ecosystem since it reduces labor rates, increases supply chain security, and stimulates economic gains of equipment manufacturers by establishing ATP facilities [9]. However, the entire value-chain, especially backend manufacturing, have yet to reap the full benefits of Industry 4.0’s digitization for streamlined manufacturing/ATP.

AR/VR is another Industry 4.0 technology that is sporadically used in the industry, which is becoming more accessible and widespread in various fields. The users are immersed within a digital environment that provides them with a real-world sense of simulated realms, allowing them to see and work with digital objects. The application of AR/VR has a transformative capacity in semiconductor education, workforce training/onboarding, and maintenance/manufacturing/assembly.

The intricate chip design and fabrication flow of advanced packaging and technology necessitates a talented interdisciplinary workforce with extensive knowledge, skills, and abilities in various STEM domains. Nonetheless, U.S. community colleges, technical/vocational schools, and colleges/universities are not supplying enough talent to meet industry demands. Current estimates predict that workforce needs will double as the CHIPS Act, both directly and indirectly, generates thousands of job opportunities within the next couple years [4], but there is not enough talent to fill them.

The promising capabilities of automation and AR/VR, especially when used in parallel, increases workplace desirability, decreases manufacturing defects, and optimizes defect detection through recent advancements in AI/ML, ultimately maximizing yields in production. Villani et al. [10] provides a high-level analysis of industrial automation through case studies, which briefly demonstrates that despite rapid progression in modern production systems, human workers remain critical in industrial workplaces and that automation features as a complementing element by handling complex machinery operations to reduce heavy workloads on human workers. Leveraging automation mechanisms serves to improve desirability in work environments as it streamlines tedious and repetitive tasks.

The roots of Industry 4.0 stems from the Industrial Revolution, which was propelled by the rapid development in science and technology in the 18^th century to streamline production and transportation through mechanization, with the emergence of the steam engine and the spinning wheel. Before Industry 1.0, iron production was arduous, which required workers to manually heat the iron ores with charcoal through a furnace. Eventually, this led to a timber shortage and called for a more efficient process and accessible fuel source. The discovery of coal revolutionized the industry as it fueled both the steam engine to provide mechanical power and revamped iron production [11]. By controlling the carbon content present in iron during smelting, the Bessemer process enabled the mass production of steel, which had superior properties to the brittle wrought iron. This transformed the global economy and infrastructure through the development of industrial machinery and complex railroad systems.

The growth of the steel industry paved the way for Industry 2.0, which further modernized industrialization through electricity, assembly lines, and conveyor belts. Henry Ford pioneered the idea of mass production through the innovative combination of the assembly line to produce his affordable Ford Model T vehicles, which ushered in rapid technological globalization by revolutionizing transportation and general manufacturing [12]. Before his unique invention, only one station sustained the full assembly, which then transitioned to a distributed labor system divided into organized steps, inspired by the conveyor belts at that time. This incited the development of industrial-scale assembly lines, which deployed a novel systematic method composed of a rotating belt that transported parts for each worker to install, leading to an equal division of labor. As a result, the industry experienced increased worker productivity and decreased costs, ultimately simplifying and maximizing through-put for factories.

The key hallmark of Industry 1.0 was “steam-power”, which harnessed the potential of mechanization to replace the “muscle-power” endured from manual labor [13], while Industry 2.0 realized the colossal power of sophisticated machinery and electricity in further bolstering the economy.

In contrast, Industry 3.0’s main pivot was directed towards improvising production systems through microelectronics and Information Technology(IT), therefore introducing industrial digitization with advent of the computer. The digitization of factories is rooted in embedded programmable logic controllers incorporated into machinery to allow partial automation and streamlined data collection [14]. Industry 3.0 proliferated the power of automation via the rapid evolution of technological systems, as it currently exists in various aspects of manufacturing, including material tracking, equipment control, and product flow management to advanced process control. The benefits of automated metrology, decision-making, and analysis extends onto faster time-to-data, increased productivity and throughput, as it presents in Industry 4.0.

Industry 4.0 is characterized by full automation, intelligent tools with decision-making power, and big data analytics [15] to efficiently maximize productivity across the entire value chain. It is an amalgamation of recent innovations, such as: IoT, collaborative robots, digital twin modeling, AI, AR/VR, and cloud computing [16, 17]. The seamless cooperation between each domain creates a reconfigurable and adaptive cyber-physical realm that aims to digitize manufacturing. Wan et al. [13] defines Industry 4.0 as a fusion of three prime technological domains aimed at developing ingenious dynamics within modern manufacturing, known as the CIOT collaboration. The domains are Collaboration Technology (CT), Information Technology (IT), and the Operational Technology (OT)[13].

• •

  Collaboration Domain - The exchange of big data is conceded through the CT domain, as it encompasses a wide range of communication protocols, including wireless networks such as Bluetooth and Zigbee as well as high-caliber cellular networks [13].

• •

  Information Technology Domain - To provide an intricate classification of OT data, the IT framework addresses the enhanced processing of complex datasets collected by humans and smart sensors through the aid of digital twin modeling, IoT, big data analytics, cloud computing, and artificial intelligence. With the current trajectory towards advanced manufacturing, traditional data analysis and synthesis techniques are projected to be overwhelmed by the enormous datasets that come from complex industrial processes [18], such as in 300 mm wafer fabs, especially as time progresses with growing production demands.

• •

  Operational Technology Domain - The OT serves to create a sustainable industrial operation/production ecosystem through effective management for maintenance of factories and creates a simpler interaction between employees and industrial equipment/materials.

All three ubiquitous domains form the basis for four potential innovative solutions applicable towards the semiconductor industry: AR/VR, remote servicing, Computer Vision(CV)/ML enabled automated inspections, and predictive maintenance, as illustrated in Figure 4.

1. 1.

   AR/VR for remote planning and digital twin modeling. Employees can congregate through the CT domain to layout plans through AR/VR and visualize ideas, as well as exchange assistance. Furthermore, AR/VR can be used to simulate the 3D digital twin model of various fab scenarios for optimized production and reduce delays in congregating collaborative meetings through a virtual setting.

2. 2.

   Remote servicing is functional throughout all industrial settings. In midst of natural disasters or travel barriers, AR/VR can aid in remote servicing, where a group of experts can join to combat issues, such as equipment failure. It facilitates real-time interaction between the equipment manufacturers and clients by overcoming geographical barriers.

3. 3.

   Predictive Maintenance enabled through a network of IoT sensors. The data captured through the sensors is processed through mechanisms of big data analytics. ML algorithms can be developed to analyze the data, which workers can monitor to stay attentive of probable equipment failures that can halt production and create intervention plans ahead of time. With the tool to tool communications, there is not just predictive maintenance, but also quality bound optimization and automated control adjustments between processes, such as course correction versus the compilation of tolerance failures during manufacturing. The aim is to reduce time spent by workers on diagnosing unexpected failures and repairs.

4. 4.

   CV/ML enabled automated inspections to monitor production quality through analysis of visual data from the production line to spot possible defects and faults. The trained algorithms can streamline failure analysis to ensure maximum yield without needing manual quality inspection (which carries a higher chance of missed defects due to human error).

Adoption of this integrated framework offers multiple unique venues that all play a crucial role in both front-end and back-end semiconductor manufacturing. Every century has carried innovations from all scientific domains to develop technology to supplement human productivity,which serves to optimize yield and maximize efficiency. The rest of this paper analyzes the feasibility of current WFD initiatives and its the pressing challenges, including talent retention and workplace desirability.

Current downward trajectory of undergraduate retention rates within engineering programs has initiated substantial strategic efforts pivoted towards drawing pre-college students towards STEM, particularly electronics, as early as elementary and middle school. Since the average American school curriculum leaves out critical engineering disciplines with the exception of computer science courses, many talented students gravitate towards software, which leaves them alienated from the microelectronics field. As a response, companies and organizations are teaming with schools in taking remarkable initiatives to cultivate meaningful connections with K-12 students, such as STEM@GF, an online platform by Global Foundries (GF) that provides guided science projects for kids to develop core engineering skills [22]. The following initiatives introduce core hardware concepts, and familiarize students with careers opportunities [8], [23]:

• •

  Gamified learning projects

• •

  STEM outreach programs, summer camps, and extracurriculars sponsored by industrial partners

• •

  Facility visits and invited professional as guest lecturers

• •

  Mentorships in career exploration and research

• •

  On-campus (high school) career fairs with industry visits to classrooms

• •

  Industry-specific STEM camps or industry-led short courses with involved projects for direct workforce preparation

Notable efforts are put forth by Samsung [24], which offers middle and high school outreach programs, known as the Semiconductor STEM Academy Experience, dedicated to exposing 6-12^th grade students to careers path in the semiconductor and manufacturing industry. Either students themselves or the school district can request a tour of the facility and time with subject matter experts such as the engineers/technicians or inviting Samsung for school events and presentations. Micron offers similar initiates such as the Chip Camp in select U.S. middle school districts, where students get to explore key processing steps in chip production (processing silicon wafers, photolithography, etching, and ion implanting), various STEM activities, as well as receiving mentorship by both Micron team members and engineering intern students. Notable student group projects include designing a solid-state drive using Lego parts under multiple constraints to stimulate innate creativity. For high school graduates, Samsung provides high school graduates a 10-week internship in a technician career track, enabling them experience working in the industry. The first two weeks involve a mentor guiding the interns to help them determine if this is a suitable career choice for them. The final week consists of a showcase event with intern presentations, and select students will be invited to transition into their Fab Apprentice Program, an earn-as-you-learn initiative. This program allows the aspiring technicians the opportunity to work with 2 days a week while completing their associates degree in engineering technology or manufacturing, and upon graduation, they will be eligible to receive a full-time technician role at Samsung. Engineering students also have the opportunity to intern as a semiconductor engineering intern, a 3-month internship they are assigned a project in an array of areas such as analytics, automation, diffusion, etch, photolithography, etc. with the opportunity to transition to a full-time engineering position as well.

A particular focus of WFD programs also consist of boot-camp courses offered by various community colleges to train future technicians, such as the community college system of Arizona [25], one of the key states that is projected to become the next technological hub with the ongoing construction of fabs by both Taiwan Semiconductor Manufacturing Company(TSMC) and Intel. The routine operation of these fabs relies on the hands of clean-room technicians, who facilitate production manufacturing processes, oversee the quality and flow of equipment functions, and handle general maintenance. Candidates don’t require advanced degrees to take on these jobs, merely needing a short-term training course to meet the position qualifications. Such programs are targeted towards skilling and hiring equipment maintenance technicians. Particularly, this includes graduates of technical Associate degree programs, veterans entering the workforce, and skilled workers from other sectors of manufacturing (with transferable skills), as all are credible sources of talent in for such positions.

A case is The Registered Apprenticeship at GF, an apprenticeship program that utilizes the learn-and-earn approach through hands-on training and classroom instructions for full-time technician roles. The training is free, targeted for all backgrounds with a high school diploma. Additionally, program graduates have the option to pursue an Associate degree afterwards. Through this program, all three stakeholders benefit within the partnership: the industry receives a constant flow of employees, each program graduate gets immediate employment with new skills, and the training program receives a pipeline of student participants to optimize new content delivery modalities for adult learners [26].

To further expand educational and training initiatives, Purdue University established the Semiconductor Degrees Program’s interdisciplinary track to offer undergraduate, masters’, and postgraduate courses that encompass all key steps in chip production, such as chemical and materials processing, manufacturing, packaging, and supply chain management, as well as associates degrees through the nearby community college.

In brief, the CHIPS Act has compelled key pathway agreements between 2-year and 4-year institutions to increase the flow of student transfers, new multidisciplinary educational and training programs (internships/apprenticeships) to expand and diversify the workforce, and strategic industry partnerships with leading semiconductor companies and organizations [27].

TSMC, a global leading manufacturer of semiconductor chips, is spending $40 billion to build two major semiconductor foundries in Arizona, based on the 3nm and 4nm process, with plans to produce the most advanced chips in building a resilient U.S. supply chain that meets annual chip demands. To do so, it will need approximately 4,500 new hires to operate the two fabs, but has already earned a poor reputation amongst both current and prospective employees. On the company’s Glassdoor page, employees anonymously reveal their experiences about the brutal working environment: cultural clashes, high-stress fab work, misaligned management, 12-hour work days, and a cultural rift [29]. Meanwhile, U.S. workers are accustomed to standard working conditions with expectations of a competitive salary and reasonable compensation. As a part of the onboarding process, U.S. staff are required to train at the company’s main hub in Taiwan for 12-18 months, which is a considerable downside for potential graduates entering the workforce. The next issue is salary, which is comparatively low compared to big tech and even rivaling companies in the industry, notably Intel, Micron, and GlobalFoundries—both of which still experience similar challenges in staffing their news plants despite a well-established reputation within the U.S. technological industry.

Decades of offshoring has not only eroded the domestic supply chain and experienced workers, but the entire talent ecosystem and expertise as well, including instructors, consultants, and other support systems needed to sustain the semiconductor industry and even extends to general high tech manufacturing. Regardless of the current manufacturing gap and staffing challenges, doubled by the hurdles in initiating new plant operations, the potential geopolitical and economic gains far outweigh the risks for leading companies [29].

To summarize, rebuilding a successful workforce with continuous output of talent while improving workplace desirability for the next generation is a part of CHIPS Act’s objectives coming to fruition. To keep U.S. competitive in the manufacturing race calls for measures in investing in two innovative solutions to complement WFD efforts: Automation and AR/VR. To revitalize domestic chip production, the semiconductor industry immensely benefits from leveraging digitization, realized through Deep Learning (DL) (a subset of ML), along with Computer Vision(CV), in broader areas of manufacturing, packaging, and failure analysis/detection [30, 31, 32, 33, 34, 35, 36, 37, 38, 39, 40, 41]. This is coupled with operation processes becoming more sophisticated with heterogeneous integration in the context of advanced packaging techniques. The following section introduces various opportunities in automation and AR/VR to streamline various fab roles to support worker productivity and maximize workplace desirability.

The process of semiconductor manufacturing is broadly split across three mains areas: wafer manufacturing, chip manufacturing (verification of internal circuitry), and product manufacturing (assembling and testing of the full chip) [53]. Automation within the semiconductor value-chain spans across a continuum, as the high-volume manufacturing environment is highly dynamic and errors of any degree within any step of production can drastically affect the quality and yield.

Historically, front-end operations within 200mm fabs were manually handled by clean room technicians, such as material transportation, which affected production throughput due to more human error and higher risk of contamination during wafer processing—this was replaced by the AMHS, which reasonably reduced human action during manufacturing, as well as advanced process control, and production planning. As a result, current fabs are highly automated up to Industry 3.0 level.

Despite the evolved techniques of Industry 4.0, the semiconductor sectors, specifically back-end manufacturing ATP, do not fully harness the power of smart manufacturing, which restricts the valuable time of existing workers spent towards manual and repetitive tasks. Amid the talent shortage, automation releases the pressure the industry is facing by optimizing workplace desirability through automating repetitive and/or hazardous tasks and increasing worker productivity, as well as maximizing profitability in parallel, as automation tools streamline time-consuming processing, assembly, and quality control flow [8].

The SEMI Foundation [54] provides a spectrum that presents the progressing stages of automation for fabs, from level 0 (zero automation) to level 5 (full automation). Level 1 and 2 consists of Industry 3.0 techniques. For level 1, it consists of AMHS, defect control for yield tracking, and limited preventative maintenance, all of which aid the operator at a subsystem level. Advanced Process Control (APC) is introduced at level 2, which involves Run-to-Run (R2R) processing and Fault Defect Classification (FDC). Broader Industry 4.0 applications, such as AI/ML, are introduced in level 3 to provide conditional automation, in providing preventative maintenance techniques and digital twin modeling on cloud-based platforms. Level 4 and 5 further advance level 3 as with the move towards full automation, where AI/ML provides high-level cognitive abilities for operational decision-making.

In addition, back-end semiconductor manufacturing ATP, which is currently present across US-based foundries and IDMs, remains relatively labor-intensive as it has not fully embraced advanced automation technologies for its existing tasks. With the growing use of advanced packaging techniques, there is a call for more Industry 4.0 automation tools, since more manual tasks increase error rates and reduce worker productivity during the execution of complex procedures. The amount of time a worker spends handling various materials and machinery accounts for 30 to 50 percent of all labor, in addition to time spent idle for production cycle completion [55].

While the construction of new fabs across the U.S. will open up numerous potential jobs, Accenture presents that certain lower-level fab roles, extensively involved in laborious tasks, will be fading away by 2030 due to automation superseding manual labor across all segments of the industry, as forecasted by the U.S. Bureau of Labor [8]. A series of case studies will be presented that realize the potential in automating fab roles such as: equipment engineering and technician, quality engineering, characterization, reliability tests engineer, and process engineering.

Smart manufacturing, within the context of Industry 4.0, is a broad amalgamation of the following innovations and their role: IoT provides data collection platforms and communication mediums, cyber-physical systems programmed with predictive algorithms allow the physical and virtual systems to interact through a IoT platform which provides predictive equipment maintenance capabilities, and cloud computing serves as a network storage base for cloud manufacturing services such as product design and testing simulations [56]. Large datasets of critical processing information is collected and analyzed by an ML algorithm to derive hidden patterns [57])

A particular use case within front-end manufacturing is Automated Visual Inspection (AVI) of wafers aims to reduce manual labor spent on quality control while increasing production yield. The general procedure of AVI consists of inspecting the device by sensors and processing the collected data through multitude of CV or ML techniques, which provides feedback throughout manufacturing flow for further optimization [58]. Any detected abnormalities notify the quality or process engineer in-charge for further fault assessment, only if a back-up Out-of-Control Action Plan is not implemented to automatically halt the operation to prevent potential yield loss [59]. As a result, this eliminated the need for manual data tracking and analysis.