According to SEMI's latest report, the global semiconductor industry is expected to invest more than $500 billion in 84 large-scale chip manufacturing plants to be built between 2021 and 2023, with segments including automotive and high-performance computing driving spending growth. Growth expectations include 33 new plants to start construction this year and 28 new plants expected to be added in 2023.   SEMI President and CEO Ajit Manocha said: "The latest update to the World Fab Forecast Report reflects the growing strategic importance of semiconductors to countries around the world and many industries. The report highlights the significant impact of government incentives in expanding production capacity and strengthening the supply chain. Given the industry's long-term prospects, increased investment in semiconductor manufacturing is critical to laying the foundation for long-term growth driven by a variety of emerging applications. ” Build 41 wafer fabs in four years The data provided earlier by Taiwan's Industrial Technology Research Institute International said that 41 wafer fabs will be built worldwide from 2022 to 2025, equivalent to the current total number of 12-inch fabs in Taiwan. Among them, the United States has the largest number of nine new plants, mainly because TSMC, Samsung, and Intel have invested heavily in building plants in the United States, which may not only cause the industry to face pressure from supply and demand imbalance, but also bring challenges such as high power consumption and high carbon emissions. Taiwanese plants will be tested by geopolitics. When releasing the above data, Taiwan's Industrial Technology Research Institute (ITRI) called on semiconductor giants to pay attention to the many challenges that will be brought about by the crazy construction of new plants around the world. ITRI analyzed that in terms of the single country where new wafer fabs will be located in the future, with TSMC, Samsung, Intel, Micron, and Texas Instruments investing heavily in expanding production in the United States, the total number of new wafer fabs in the United States in the next three years will be the largest, reaching nine, including eight 12-inch plants and one 8-inch plant. Huang Huixiu, an industry analyst at the Semiconductor Research Department of ITRI, pointed out that due to the growth of semiconductor application demand and geopolitical factors, the global wafer fab expansion wave is expected to start construction of 41 new plants from 2022 to 2025, but it will also generate more carbon footprints. The semiconductor industry must improve problems such as high power consumption and high carbon emissions. Huang Huixiu believes that geopolitics continues to drive the semiconductor industry from globalization to regionalization. Among them, the US chip bill has pushed large companies to produce in the United States and reduce their dependence on non-US companies. It is estimated that the proportion of production capacity in the United States will increase. In addition, the upgraded US export control has restricted the development of mainland semiconductors. We have seen that Taiwan's wafer foundries continue to reduce their dependence on mainland customers as the proportion of customers outside the United States and mainland China increases. Strong demand for chip manufacturingAccording to Yole, the global pandemic initially diverted investments in wafer fab transactions, but the resilient semiconductor industry, which is struggling with supply chain constraints and a shortage of skilled labor, has regained interest due to potential financial support from U.S. and European chip legislation.

Stephen Rothrock, CEO of ATREG, and Jean-Christophe, CEO of Yole, discussed emerging business models that could enable the global semiconductor industry to migrate to the forefront of manufacturing capabilities for both cutting-edge and mature technologies.

During COVID, constrained supply chains led to significant price increases across all areas of the semiconductor world, except for the memory business. Yole estimates that this segment of the business grew by 25% in semiconductor sales in 2021 but will drop to 11% this year, partly due to price increases. The automotive industry has seen particularly notable growth due to the rapid development of electric vehicle technology, but strong demand has also driven up prices in the industrial and data center markets. Another factor contributing to the slowdown in growth is inflation, which has led to a reduction in orders for smartphones and other consumer products. The semiconductor supply chain has also experienced disruptions due to the war in Ukraine, coupled with the growth of hyperscale companies and startups.

Advertisement

Undoubtedly, the winds of recession are blowing, forcing chip companies to take this into account when considering increasing wafer production. However, a lower growth rate does not necessarily mean a decline in manufacturing demand for many companies' products. Demand remains strong, and both IDMs and foundries need to find ways to increase wafer fab capacity and address current and future challenges.

Although the semiconductor industry has been able to grow rapidly, the demand for more capacity and the associated significant investment is still challenging under limited budgets. The resilience of the chip manufacturer's supply chain will be tested as it tries to quickly adapt to the current economic and geopolitical environment. In the coming months, it will be crucial to monitor and identify weak links and for the industry to develop strategies to address these weaknesses.

The significant construction cost dilemma of wafer fabs

An increasing number of companies are interested in end-to-end design and manufacturing capabilities. Cutting-edge technologies will benefit from extensive R&D programs, but the key issue is the lack of wafer fabs at this level. One significant reason for the global shortage of wafer fabs is the substantial cost required to build them and bring them online.

Considering that the cost of a cutting-edge wafer fab at 5nm is around $5 billion or more, and the cost at 10nm is slightly less than $2 billion, the cost difference is enormous. Government incentives will undoubtedly help, but despite China having invested about $73 billion in semiconductor subsidies to date (not including government grants, equity investments, and low-interest loans exceeding $50 billion), it seems the effect is also very limited, let alone the United States, Europe, and Japan, which have not been able to come close to matching that figure. The upcoming European CHIPS Act and the recently passed U.S. legislation mean we may see more cutting-edge capabilities developing worldwide. South Korea and Japan are also enacting their own financial support legislation.

Even with all these initiatives, it may take us three years before we start to see manufacturing capacity entering the market.There have been no significant investment consolidations in the past three quarters, and the delays have led everyone to catch up and invest. The need for more cutting-edge equipment necessitates further expansion of production.

Currently, global supply is limited because there are only about 150 300mm wafer fabs worldwide, with 42 in Taiwan, 33 in mainland China, 19 in the United States, and only 12 in Europe and the Middle East. While the leading edge is important and attracts a lot of investment, a significant portion of the activity is still at 200mm, with a large demand for different nodes of the 90-nanometer to 180-nanometer process.

There are about 230 200mm wafer fabs globally, with 51 located in the United States and 49 in Europe and the Middle East.

Trade-offs in building a wafer fab

The new wafer fabs being built today need to strike a balance between producing at less advanced nodes and meeting long-term demand. Texas Instruments is a typical example, as the company announced its goal of building a new 300mm wafer fab to meet its analog production needs in traditional processes. This is also an important trend for the industry, as it means a major reorganization of production due to the difficulty of accessing 200mm equipment.

After Intel announced the construction of a factory in Ohio and upgrading its fabs in Phoenix, Israel, and Ireland, many large companies mainly turned to the leading edge. The level of investment announcements is very important—Micron's $100 billion plan in central New York, Samsung and Texas Instruments' greenfield in Texas, and Bosch, Infineon Technologies, and STMicroelectronics have 300mm investments in Europe. However, it is worth noting that not all of these 300mm wafer fabs are leading edge—many are slightly behind the process of 14 nanometers and above.

There are also geopolitical factors at play. Taiwan has witnessed the rise of TSMC, which now controls more than 55% of global foundry manufacturing. This means that many small companies are now wondering what will happen to them, as large companies like Apple and Qualcomm dominate demand within the foundries. This concern has led them to look for their own wafer fab manufacturing opportunities to reduce dependence on others and ensure the security of the supply chain.

In addition, the trend of power electronic technology using compound semiconductors provides opportunities for the semiconductor industry. We see investment shifting towards GaAs, SiC, and GaN manufacturing facilities. For example, Wolfspeed has established a leading silicon carbide factory in the Mohawk Valley of New York, and they recently announced the establishment of a new material factory in North Carolina to support the SiC business. Osram also announced an investment of $850 million for an 8-inch facility for microLEDs. Such compound semiconductor investments will significantly expand the available capacity of these products to serve end markets such as mobile phones, 5G networks, and electric vehicles.A Series of Challenges in Wafer Fab Operations

It typically takes 18 to 24 months to construct the theme of a wafer fab, install equipment, and then spend another 18 months to get the tool line running and make the products meet mass production requirements. As companies seek turnkey facilities to meet the growing demand, this has led to an unprecedented demand for site locations. Some chip manufacturers adopting the fab-light strategy are selling facilities to other users to ensure long-term supply, rather than considering alternative uses. This is the case with onsemi, who helped sell three operating facilities last year, including the sale of its 150mm wafer fab located in Oudenaarde, Belgium, to BelGaN Group BV, which will use an 8-inch conversion kit that can be used for 70% of the tools.

Other companies choose to renovate existing premises, transforming them to include wafer fab spaces or adding cleanrooms as extensions to existing buildings.

SkyWater Technology, a wafer fab headquartered in Minnesota, expanded its 200mm Bloomington wafer fab with a Department of Defense grant that covered most of the construction costs. There are various factors that compel these and other large, medium, and small chip manufacturers to consider renovating existing facilities or building new ones.

In the United States, the typical cost of a cleanroom building was $5,500 per square foot three years ago, but it is now close to $8,500 per square foot. Cleanrooms usually account for one-third of the total wafer fab floor space and need to be efficient to see a return on investment within five years.

In Albany, New York, a nanocenter has been built with about 75,000 square feet of cleanroom space available for smaller users for R&D and small-batch production. While some companies are very focused on control and intellectual property in this nanocenter model, others may find it a viable alternative to building their own smaller on-site facilities. In today's context, it is crucial to continue building the existing semiconductor ecosystem and clusters to provide a stronger supply chain and ensure additional support infrastructure critical to wafer fab operations.

Chip legislation in the United States and Europe also aims to provide subsidies for new companies and small players, not just large manufacturers. It remains to be seen how many smaller projects will be funded, considering the four-year funding allocation is equivalent to China's two-year expenditure. With other plans such as the U.S. Infrastructure Act and the Science Act competing for funds, construction engineering companies and skilled trade workers will face fierce competition. These human capital issues pose a significant additional challenge for the semiconductor industry.

The current situation forces chip manufacturers to work smarter and join forces to maximize the use of limited resources to reduce manufacturing costs. The recent announcements from ST Microelectronics and GlobalFoundries represent an interesting business model. The collaboration to establish a jointly operated 300mm wafer fab next to ST's factory in Crolles, France, indicates that companies are increasingly comfortable with how to handle and protect their intellectual property, and they can create modules within wafer fabs that run in parallel.

In the near future, we may see more of these partnerships as a way to optimize wafer fab operations in terms of loading, efficiency, time to market, and cost, especially in the compound semiconductor field where two or three companies can jointly operate a single facility.Just as important as the design of a wafer fab, it is equally crucial for a company to control manufacturing costs and balance them with a sound strategic R&D investment plan. Moreover, product cycles can be very short—a process may take two years to establish, but it becomes obsolete when competitors are already ahead in the market. This explains why the semiconductor industry is cautious in building wafer fabs and focuses on clusters.

Wafer fabs need to operate at 75% or higher capacity to be cost-effective. Given the high costs of operating a wafer fab, chip manufacturers need to be able to manage processes effectively, which has led to so much consolidation over the past decade. Companies can manufacture a single product in a streamlined process. Texas Instruments is a great example, focusing on analog and 300mm, and now dominating the market. The same consolidation strategy has also been successful for Intel, Qualcomm, and Nvidia.

The enthusiasm of the financial market for leading companies with gross margins of 40% to 60% needs to be addressed, as it does not necessarily synchronize with the need to hold inventory. As we learned from COVID-19, lean inventory can cause problems in the supply chain. Production investment is a solution to improve profit margins, so that wafer fabs can process more equipment and reduce costs.

The nature of the global semiconductor industry means that plans made today and their results will only emerge in four to five years. IDMs and foundries are both developing their technologies, manufacturing expectations, and ideas on how to meet the growing demand for more microchips. We expect them to find new sources of funding, new collaboration models, and new ways to adjust their business practices to adapt to the new global semiconductor manufacturing landscape.