Smarter Ways To Manufacture Chips

Early successes are spurring further investments, with a concentration on high ROI projects.


OSAT and wafer fabs are beginning to invest in Industry 4.0 solutions in order to improve efficiency and reduce operating costs, but it’s a complicated process that involves setting up frameworks to evaluate different options and goals.

Semiconductor manufacturing facilities have relied on dedicated automation teams for decades. These teams track and schedule chip production, respond to equipment or network issues, and they move data files to and from central locations. In addition, they interpret data output from station controllers.

But IC fabs also are capital-intensive operations. They have high operating costs due to the complex manufacturing challenges they need to address. On top of that, they must deal with frequent changes in customer and market demands. And because they are run 24/7 when demand is high — these days, many fabs are running at full capacity — they also consume large amounts of energy.

This is where Industry 4.0 fits in. Done right, it can significantly boost production efficiency and agility with lower operating costs.

“Smart manufacturing, or Industry 4.0, is a collection of technologies that have the potential to increase the efficiency of traditional semiconductor manufacturing facilities,” said Mark da Silva, senior director of smart manufacturing initiative at SEMI. “By far the most dramatic change that has happened in the last decade has been the rise of AI/ML tools and techniques that are slowly working to ‘close the loop,’ i.e., enable actionable predictions in each step.”

Enormous potential
The amount of data generated by fabs, and to a lesser extent assembly and test facilities, has exploded in recent years due to the implementation of internet of things (IoT) technologies. That enables engineering teams to monitor every aspect of factory operations, to analyze the data, and to act upon it, often with full automation. But these projects do take time to implement, and they require investments in technology and the selection of projects with the highest impact potential.

Factory managers cite IoT, AI/ML, cloud infrastructure, and big data analytics as the top four technologies that can make an impact. And the ability to respond to manufacturing changes has significantly heightened the interest in deploying these technologies.

“We describe smart manufacturing as integrated, collaborative manufacturing systems that enable us to respond rapidly to changing demands and conditions in our factory, in the supply chain, and for customer needs,” said Won Lee, vice president of smart factory engineering at Amkor. “Automated manufacturing simply refers to robotic devices that are programmed with the sole purpose of performing a specific action. Most often, these repetitive tasks occur in concert with other machines and humans in an environment such as an assembly line. This automated system is optimized through the harmonization of people’s actions, roles, and responsibilities within the factory. In addition, safety during harmonized work between machine and people is secured through thorough risk assessment before implementation.”

Smart manufacturing solutions also can help address the chip industry’s talent shortage.

“The rapid growth of IC manufacturing capacity has created a severe shortfall of experts to run these modern factories. Smart manufacturing is expected to automate decision making to compensate for this shortfall,” noted Anjaneya Thakar, senior director of product marketing in Synopsys’ Silicon Engineering Group. “Software solutions that supplement human expertise in enabling accurate decision-making are necessary for efficient operations.”

First steps
Industry 4.0 begins with gathering data specific to equipment/workstation controllers, and tracking all the information related to production workflows. Combining this information with automation techniques and algorithms can result in lower operational costs and increased production efficiency. Such approaches can be applied in the design of new factories, as well as for improvements to existing factories.

A wide array of technologies can be considered to enable the data collection and subsequent translation into actionable tasks, including automated data storage and transportation, cybersecurity solutions, robotics, IoT devices and networks, cloud computing, and software for machine learning and artificial intelligence.

An assessment tool can be used to evaluate a factory’s current state and the desired next state, including possible paths, implementations, and available technologies. In addition, a successful assessment requires collaboration across various departments — production, maintenance, engineering, logistics, quality management, planning, and facilities.

Having a consistent evaluation framework is important for driving discussions among partner organizations and upper management regarding the next level of automation adoption and expected benefits.

As part of SEMI’s Smart Manufacturing Initiative, the Smart Manufacturing Guidelines sub-committee developed the Industry 4.0 Readiness Assessment Model (IRAM) as an evaluation tool. [1] The sensing, connecting and predicting pillars all have four levels of implementation. For instance, predicting level 1 is reactive to results, whereas level 4 features autonomous predictions. Engineering teams can use the tool to identify the current factory status and envision a future status. The foundational needs in this model consider cybersecurity, identity and trusted traceability, and workforce skill set.

Fig. 1: Assessing smart manufacturing ROI by evaluating a facility’s sensing, connecting predicting capabilities. Source: SEMI

Fig. 1: Assessing smart manufacturing ROI by evaluating a facility’s sensing, connecting predicting capabilities. Source: SEMI

To develop IRAM, Andrew Seward of TEL and David Gross of Siemens Digital Industries Software co-chaired the subcommittee which included manufacturers, equipment suppliers, and automation suppliers in development of the model. Feedback from reviewers of the beta model resulted in version 1.5, which is available and is being shared to gain additional feedback from users.

Other organizations are using internally developed assessment frameworks.

“We are using a goal and scorecard for each manufacturing line/process based on the predefined I4.0 pillars. This enables us to track and systematically improve our goals and achievements for I4.0 implementation in each line and process,” said Joon Ahn, vice president and group manager for Amkor’s IT Division.

Ahn described graduated steps toward a fully automated factory, which starts with switching to paperless operations, then enabling smart mobility, followed by an increasing reliance upon IoT-cloud analysis for real-time production analysis, and increasing the use of robotics that work alongside operators.

Fig. 2: A four-step approach to higher levels of automation.  Source: Amkor

Fig. 2: A four-step approach to higher levels of automation.  Source: Amkor

“It is important to conduct an analysis to identify the areas and processes where smart factory implementation can be most effective,” said Amkor’s Lee. “This can be achieved through analyzing production data and conducting job analyses. After the implementation of automation, specific metrics such as productivity, work time, and quality level can be quantified and analyzed. This allows for the validation of effective systems and the ability to advance or expand into other areas. It is also important to conduct a risk assessment to identify potential risks during the process of change and prepare appropriate contingency and backup plans in advance.”

Alongside the physical process of manufacturing is the flow of information, which can move both upstream and downstream.

“From the standpoint of solutions that enable smart manufacturing, we are at a very nice confluence of available technologies and innovations in the industry,” said Synopsys’ Thakar. “For effective automation solutions, which are required for smart manufacturing, it is critical to have feed-forward and feedback mechanisms during this entire process. The decisions made early on during process development impact manufacturability and yield in HVM. Our roadmap for smart manufacturing solutions comprehends this need and builds on linking insights from IC design, test, process modeling, lithography effects, and manufacturing data to ensure fast and accurate results in the context of a specific process technology and product design. We will deliver solutions in process control that shift the paradigm from reactive to predictive and, eventually, autonomous control.”

As an evaluation tool, IRAM will be most helpful with upgrades to 200mm fabs, in which gains from investments could be quite significant compared to state-of-the-art 300mm fabs manufacturing CMOS at the 10/7/5nm process nodes. “The 300mm fabs are the most advanced in terms of deployment of Industry 4.0. The 200mm fabs lag behind the 300mm fabs in terms of Industry 4.0 maturity,” said SEMI’s da Silva. “With the IRAM model you could see the difference in the scoring, as well. Most 200mm fabs require updates to deploy Industry 4.0 solutions, and the challenge is to quantify the benefits.”

Success stories
Sharing industry success stories enables others to comprehend the business impact of moving up the ramp to a smart factory. The World Economic Forum and McKinsey created the Global Lighthouse Network (GLN) to recognize manufacturing sites and value chains leading in the adoption of Industrial 4.0 technologies. [2] Semiconductor facilities that have joined this network showcase the achievable benefits.

In January of 2023, ASE’s bumping factory in Kaohsiung was added to the GLN. [3] “The increasing complexity of semiconductor chip manufacturing processes characterized by market disruptions in supply and demand had caused unprecedented challenges for ASE Kaohsiung’s bumping factory,” the company said. “In the bumping operation, there are more than 100 process steps compared with traditional IC packaging operations. To streamline the manufacturing processes and optimize production, ASE strategically planned and deployed ID4.0 technologies across its operations. In particular, AI-enabled processes helped ASE to improve manufacturing yields and accuracy, resulting in an increase in output by 67% and a reduction of order lead time by 39%.”

Production efficiency also can lead to product quality improvements. As Amkor’s Lee noted, “Along with an established roadmap over the next five years to smart manufacturing with ID 4.0, Amkor foresees the following key benefits in our factories:

  • Productivity and efficiency: The use of predictive analytics and big data analysis allows for optimized processes to be identified and put in place. Just-in-time inventory management, accurate demand forecasting, and faster time-to market are some of the efficiency benefits that smart factories deliver.
  • Product quality and customer experience: The connectivity and end-to-end visibility in smart manufacturing brings real-time insights and recommendations to all tiers of the manufacturing process. Advanced analysis of system data quickly spots weaknesses or areas for improvement. This leads to improved competitiveness in the market, better product reviews, and fewer costly returns or recalls.”

A recent webinar from the leaders for the IRAM effort highlighted several case studies that illustrate the successes in deploying smart manufacturing approaches. For example, the integration of 48k facility sensors into an IoT platform at a Micron DRAM factory resulted in $5 million in savings. The platform enabled early drift detection and a correlation to the responsible manufacturing tool and product performance impact. This reduced product downgrades from facility events by 5%. The same platform was used to optimize parameter settings for energy consumption, resulting in a 15% power savings.

UMC shared another implementation that reduced energy consumption involving a combination of sensor data and advanced analytics. [4] “Through analysis of big data collected by sensors, we built models of equipment operation and established a power management platform,” the company said. “The platform enables operators to take advantage of data visualization and to gain a real-time, complete picture of energy usage across all the equipment in a fab. This is helpful for analyzing the differences in power consumption and finding ways to improve energy efficiency of the facility.”

Part of the challenge in modern fabs is the depth and diversity of sensor data. Often several multi-variant analyses are required to determine an optimal state between the operation of equipment parameters and energy consumption. One example from UMC demonstrated the benefits of installing i-Chiller, a smart chiller water system that includes ancillary equipment including a chilled water pump, cooling tower, and heat exchanger. The energy consumption of each component correlates to changes in temperature and other external factors. Finding an ‘optimal energy efficiency’ is challenging because there are 30+ factors that affect the i-Chiller’s energy usage.

The team deployed sensors, established a huge database and used data analytics to arrive at a dynamic optimized solution. “Through deviation analysis and algorithms, as well as the repeated tests for the correlation and weighting between parameters, we were able to overcome limitations and made it so every chiller will adjust to the optimal status based on external climate or on-site loading conditions,” UMC said. “Under the same enthalpy value, the operation efficiency of the i-Chiller is increased by 3.3% compared with the traditional chiller, reducing electricity consumption by as much as 12.232 GWh, and saving up to NT$29.65 million/year (~ $1 million USD/year).”

Improving a tool’s throughput also supports production efficiency, another key performance metric. In their IRAM 1.5 presentation, Seward and Gross shared an example from a joint effort between a 200mm fab and equipment supplier that resulted in a 15% improvement in tool throughput efficiency. The specific goals driving that improvement included reducing process variability, extending life of aging tools, and overcoming fab capacity constraints. They achieved this result by upgrading the equipment controller and software for higher data rates, faster wafer processing, and enabling remote access and control.

The promise of Industry 4.0 is to effectively deploy automation in order to lower operational costs, decrease the need for humans in the decision loop, and to enable agile responses to production needs or factory disruptions. Evaluation frameworks such as IRAM assist in determining which technologies to deploy, and recent success stories describe the cost and production benefits to spur further investment.

“Over the next five years, it is expected that the characteristics of smart manufacturing will be further enhanced. In particular, robots capable of performing human-like actions, and the computer vision field, will be much more developed,” said Amkor’s Won Lee. “In addition, wearable devices that can be used in manufacturing environments will become lighter and more affordable. The data processing field will become more high-capacity and high-dimensional, making realistic analysis and predictions possible.”

In the end, the cost of manufacturing products will be lower due to increased efficiency and decreased operational costs, with clear metrics to validate those benefits.


  1. “The Industry 4.0 Readiness Assessment Model (IRAM),” SEMI,
  2. “Transforming advanced manufacturing through Industry 4.0,” McKinsey & Company, web article on dated June 27, 2022,
  3. “ASE’s Bumping Factory in Kaohsiung, Taiwan inducted into the World Economic Forum’s Global Lighthouse Network,” ASE, web article on dated January 14, 2023,
  4. “Making the Impossible Possible: Using AI, Big Data, and Green Operations to Reduce Electricity Use by 15% by 2025,” UMC web article on dated January 05, 2022,

Leave a Reply

(Note: This name will be displayed publicly)