Strategic Thinking for Long-Tail Semiconductors
In 1990, NHK hailed Japan as an “Electronic Superpower,” praising its semiconductor industry. Now, after 30 years, the spotlight has returned to this sector. While cutting-edge manufacturing technology is at the center of attention, this article focuses on semiconductor design.
• December 3, 2023: “The Evolution of PDKs: Past and Present”
• August 1, 2024: “A Qualitative Cost Analysis of the Semiconductor Business”
• December 23, 2024: “Semiconductors We Want to Create, Semiconductors We Want to Use”
• January 17, 2025: “Strategic Thinking for Open-Source PDKs”
Thanks to your support, the articles I have posted in the past continue to receive steady access from many readers. I am truly pleased if they have contributed to a better understanding of semiconductors.
This time, under the title “Strategic Thinking for Long-Tail Semiconductors,” I will explain in a clear and accessible manner the development strategies for long-tail semiconductors, assuming their integration into long-tail products. This discussion is for semiconductor designers, foundries, and those involved in planning semiconductor-related businesses. I have expanded the content to be helpful for a broader audience, and I hope you find it informative and enjoyable to read.
Please note that the content presented here is based on my past professional experience and represents my personal opinions and views. It does not reflect the official stance of any organization.
What is MOQ (Minimum Order Quantity)?
The semiconductor industry is filled with three-letter acronyms, and in this discussion, I will start by explaining MOQ (Minimum Order Quantity). MOQ refers to the minimum order quantity required when placing an order. It is specified in quotations from foundries and typically corresponds to the minimum order unit measured in lots. Generally, one lot consists of 25 (or sometimes 24) wafers.
In addition to this MOQ, orders for the upcoming period (semi-annually or quarterly) must be presented in advance when negotiating production line allocation. Usually, orders exceeding this reserved quantity are not accepted. Conversely, if actual orders fall below the reserved amount, securing the desired production slots in the next negotiation may become difficult, and price negotiations may also become less favorable. The more advanced the process node, the longer the turnaround time from order placement to delivery. Misjudging the production line allocation can lead to product shortages or excess inventory.
From the foundry’s perspective, high-volume customers and those with consistent order quantities are preferred clients. As a result, foundries are often reluctant to accommodate the requests of low-volume customers or those with irregular orders, such as adjusting order quantities. In the case of mega-foundries, when signing manufacturing contracts with new customers (such as fabless startups), they may require compliance with MOQ and a forecast of order quantities for the next three to five years. For products with low total order volumes, the foundry may decline the contract altogether or, in some cases, even refuse to disclose the PDK (Process Design Kit) under an NDA (Non-Disclosure Agreement).
What Are Long-Tail Semiconductors?
In most cases, long-tail products refer to those that meet either of the following criteria:
1. Low annual production volumes, such as military, defense, and aerospace products.
2. Long product lifecycle (10–20 years), as seen in automotive, infrastructure, and industrial machinery applications.
For this discussion, long-tail products are defined as those that exhibit either (1) alone or a combination of (1) and (2). Furthermore, semiconductors used in these products are referred to as long-tail semiconductors.
On the other hand, products that are mass-produced but still have a long production lifespan (10–20 years) can typically be supported within the existing semiconductor manufacturing ecosystem. Therefore, such products are outside the scope of this discussion.
Since long-tail products have low annual production volumes, the semiconductors incorporated into them — referred to as long-tail semiconductors — also tend to have low annual production quantities. In many cases, the annual production volume is below one million units, and in extreme cases, it may not even reach 1,000 units per year.
As mentioned earlier, commissioning a custom semiconductor design and placing orders with a foundry requires a substantial annual production volume to justify the costs. For instance, assuming a chip size of 5×5 mm² and a yield rate of 95% (with defect density D = 0.002/mm²), then…
The following graph illustrates the relationship between chip size and the number of units delivered per lot.
For semiconductors with an annual demand of tens of thousands of units, a single production lot per year is sufficient to meet the requirements. Furthermore, if the annual demand is only a few thousand units, a single production lot could produce more than the total required over the product’s lifespan.
From these figures, it becomes clear why long-tail semiconductors with annual production volumes below one million units are not attractive business opportunities for large-scale semiconductor foundries (mega-foundries). The scale of demand simply does not align with their high-volume manufacturing model.
Miniaturization ≠ Minimum Cost
A detailed explanation of semiconductor cost structures is provided in “Qualitative Cost Analysis,” but in the case of long-tail semiconductors, the cost dynamics differ from those of high-volume production. Specifically, the initial cost (design + mask cost) outweighs the wafer cost (manufacturing cost).
As a result, adopting more advanced process nodes to reduce chip size can significantly increase the initial cost, ultimately leading to higher per-unit chip costs. Instead of pursuing further miniaturization, choosing an appropriate chip size that ensures a sufficient yield (considering defect density) while integrating the required functionality can optimize overall chip costs.
The following graph illustrates how the cost of a 50 mm² chip (fabricated at 180 nm) varies depending on the selected process node. The calculation is based on the wafer cost formula used in the “Qualitative Cost Analysis” and is shown for different lifetime production volumes.
For simplicity, the calculations assume that chip size decreases proportionally to the square of the process node scaling. However, in reality, certain design constraints — such as PAD areas and analog circuit blocks — do not scale linearly with process miniaturization. As a result, chip size does not always shrink as expected. Nonetheless, the purpose of this analysis is to understand the overall trend. In practice, process miniaturization offers benefits such as higher circuit speeds (i.e., increased operating clock frequencies) and reduced power consumption. However, it also comes with challenges, such as an increase in optional process steps (resulting in more mask layers), making direct cost comparisons more complex.
Case: Lifetime Production Volume of 10,000 Units
As the process node becomes more advanced, chip costs increase. The only exception is 40 nm > 28 nm, where costs decrease because, at 40 nm and below, a single wafer can produce more than the required lifetime volume, effectively distributing the cost over a more significant number of chips.
Case: Lifetime Production Volume of 100,000 Units
The chip cost reaches its minimum around the 180 nm process node, forming a lower convex point.
Case: Lifetime Production Volume of 1,000,000 Units
The chip cost reaches its minimum around the 90 nm process node, forming a lower convex point.
As process nodes shrink, mask costs increase, which means that for long-tail semiconductors with low total production volumes, smaller process nodes do not necessarily lead to lower overall costs. Even a simplified model confirms this conclusion.
When selecting a process for long-tail semiconductors, it is essential to consider not only wafer cost calculations but also the initial costs, including mask and design costs. These factors play a crucial role in determining the most cost-effective manufacturing approach.
Low-Volume Production Foundries
Beyond mega-fabs that focus their investments on cutting-edge processes and high-volume production, legacy foundries exist that cater to different market needs. The following summarizes the current state of these legacy fabs.
SkyWater Technology (US)
SkyWater Technology, the semiconductor foundry chosen by Google as a partner for Open Source Silicon, was originally a Cypress Semiconductor facility. It was acquired by a private investment firm in 2017 and now operates as an independent foundry.
Strategically, SkyWater positions itself as a foundry for companies “overlooked by TSMC.” Its target customers include the consumer electronics, industrial, military/defense, and automotive sectors. The company is also a trusted supplier for the U.S. Department of Defense (DoD), contributing to the DoD’s efforts to secure a domestic semiconductor supply chain. In other words, SkyWater serves as a supplier of long-tail semiconductors for national security applications.
SkyWater primarily manufactures at the 90 nm node on 8-inch wafers, though in Google’s OpenMPW program, it has open-sourced a 130 nm PDK. Additionally, in 2020, the company received funding from the U.S. government through the CHIPS Act and invested in a new 90 nm FDSOI production facility.
According to multiple sources, most of SkyWater’s customers are government-related, primarily in military and defense. FDSOI technology, known for its radiation resistance, further suggests applications in defense and space.
iHP Microelectronics (Germany)
The Leibniz Institute for High-Performance Microelectronics (IHP) in Germany provides low-volume manufacturing services for academic institutions and Europe’s long-tail industries through its commercial arm, iHP Solutions. The foundry operates at 130 and 250 nm nodes using 8-inch wafers. Given its cleanroom area of only 1,500 m², its production capacity is estimated to be just a few thousand wafers per month.
Prototype fabrication services using iHP Solutions are also accessible from the University of Tokyo’s VDEC (VLSI Design and Education Center), a System Design Research Center division within the Graduate School of Engineering.
In 2023, iHP announced its commitment to OpenPDK, and in 2024, it presented technical details at Latch-Up 2024, an event organized by Free and Open Source Silicon (FOSSi), a European open hardware community. The presentation video can be found here.
Tokai Rika (Japan)
For over 30 years, Tokai Rika has maintained an in-house semiconductor fabrication facility to ensure a stable supply of automotive semiconductors, which require strict quality control and flexible production planning.
In 2021, the company launched a foundry service specializing in low-volume, short lead-time manufacturing, including on-demand wafer deliveries and prototype fabrication starting from just a single wafer — making it well-suited for long-tail semiconductor production.
In its 2023 mid-term business plan, Tokai Rika announced an investment in a 0.35 μm process in addition to its existing one μm (6-inch) process. The target customers for its foundry services include the automotive, industrial, aerospace, and academic sectors.
While SkyWater Technology, iHP Solutions, and Tokai Rika have entirely different origins, they all focus on long-tail semiconductors, providing low-volume production (LVP) capabilities tailored to niche markets.
Long-Tail Semiconductor Supply Chain
Legacy Semiconductors as Strategic Materials
Before the COVID-19 pandemic, legacy semiconductors were primarily associated with Chinese manufacturing. I, too, have obtained quotes from SMIC (Shanhai Manufacturing International Corporation) and considered prototype production there multiple times. Domestic turnkey service providers also routinely outsourced manufacturing to SMIC. However, the global semiconductor ecosystem has shifted fundamentally in response to the pandemic.
In recent years, major economies have ramped up investments in domestic manufacturing facilities for cutting-edge semiconductor processes. The United States, Europe, and Japan are actively working to attract TSMC to establish new fabrication plants. Based on publicly available reports, these new facilities appear to be mega-foundries designed for large-scale production, with monthly wafer output exceeding 50,000 units.
On the other hand, there has been no significant movement at the national level — either in the legacy foundry sector or in long-tail semiconductor manufacturing investments — anywhere in the world. Unlike the push for leading-edge semiconductor manufacturing, government-backed initiatives for legacy semiconductor production remain largely absent.
Since Russia invaded Ukraine, semiconductors have gained increasing recognition as strategic resources. This is primarily driven by the belief that securing cutting-edge semiconductor processes is essential for developing AI technologies critical to military and defense applications. However, as demonstrated by the case of SkyWater Technology, not all semiconductors used in military and defense applications require the most advanced process nodes. Similarly, with its existing domestic legacy foundries, Japan has a unique advantage that should be leveraged through a strategic approach to long-tail semiconductors. Rather than solely focusing on leading-edge manufacturing, a well-thought-out long-tail semiconductor strategy could be crucial in Japan’s semiconductor industry and national security.
EOL and Discontinued Products
Here, we encounter another three-letter acronym, EOL (End of Life), which refers to the discontinuation of semiconductor production. General-purpose semiconductors designed and manufactured by semiconductor companies may be discontinued for various reasons, such as equipment retirement, design modifications, or business closure. Such discontinued semiconductor products are commonly referred to as “discontinued products” (Discon, short for “Discontinued”).
Before a semiconductor reaches EOL, manufacturers notify users in advance, providing alternative chip suggestions or last-buy (final purchase) opportunities. However, for industries that rely on long-tail products — particularly those requiring long-term manufacturing guarantees, such as industrial equipment manufacturers, automotive, infrastructure, medical, military, and defense industries — EOL can pose a critical business risk.
In consumer electronics, where product lifecycles are short, companies can mitigate EOL risks by placing a last-buy order and stockpiling enough semiconductor inventory to cover the product’s lifetime. However, proactive EOL management is essential in long-tail industries, where manufacturing may continue for over 20 years or where support and spare parts must be maintained even after production ends.
Recognizing this challenge, some semiconductor distributors in Japan, such as Toshiba Information Systems and NSW, provide redesign and contract manufacturing services for discontinued semiconductors used in long-tail products. The existence of these services underscores the critical need for strategic planning in long-tail semiconductors, particularly for industries that rely on long-term product support.
The Urban Myth of FPGA
Around the year 2000, Japan’s major IDM (Integrated Device Manufacturer) companies began withdrawing from DRAM production and shifting from ASIC (Application-Specific Integrated Circuits) to SoC (System on Chip) development. In 2003, Xilinx introduced the 90 nm Spartan series to replace ASICs. This was followed by 45 nm Spartan-6 and 28 nm Spartan-7, gradually expanding the adoption of low-cost Spartan FPGAs in long-tail products. Over time, this led to the widespread belief that “long-tail semiconductors = FPGA,” forming an urban myth in the industry.
However, the COVID-19 pandemic disrupted this assumption. Global supply chain disruptions led to FPGA shortages and price surges. While FPGAs had been perceived as easily replaceable due to their reprogrammable nature, the reality proved more complex:
• Alternative FPGA models often require different pin configurations,
• Some replacements had different packaging,
• In many cases, PCB redesign was necessary to accommodate substitutes.
Additionally, FPGA manufacturers, who once assured that their products would never face EOL (End of Life), have shown shifts in their business strategies, raising concerns about the long-term availability of FPGA devices.
Over the past 20 years, as Japan’s major IDM companies retreated from ASIC development, FPGAs filled the gap. Meanwhile, ASIC manufacturing evolved into an ecosystem dominated by turnkey service providers and mega-foundries. However, mega-fabs prioritize high-volume production and are not inclined to accommodate low-volume long-tail semiconductors. As a result, only high-volume ASICs remain viable within this ecosystem. To sustain long-tail semiconductor development in Japan, establishing a robust domestic ecosystem is critical to leverage the country’s remaining legacy fabs. Japan’s legacy fabrication capabilities offer a strategic advantage, but a new mindset and strategic approach are essential to utilize them fully.
Strategic Recommendations
1. Integrating Low-Volume Production (LVP) Fabs into the Semiconductor Ecosystem Alongside Mega-Fab Investments
As an LVP (Low-Volume Production) foundry, adopting a clear business strategy is crucial, such as “focusing on companies that are not prioritized by mega-foundries.” By deliberately targeting this underserved market, LVP foundries can establish a unique and sustainable position in the semiconductor industry.
Furthermore, fostering a group of long-tail product companies that actively leverage long-tail semiconductors for product differentiation and anti-counterfeiting measures is essential. Strengthening these industries will not only enhance the competitiveness of individual companies but also contribute to the resilience of Japan’s industrial structure in the face of global supply chain uncertainties.
2. Establishing LVP Capabilities in Domestic Legacy Fabs (8-inch, Domestic Capital) as a National Advantage
To support Japan’s long-tail product industries, it is essential to position domestic legacy foundries (8-inch, domestically owned) as LVP (Low-Volume Production) foundries that can guarantee long-term, stable manufacturing under a framework of “security and reliability.” Achieving this will require a structured support system for these foundries, including funding mechanisms and customer guidance to ensure their sustainability. Furthermore, since LVP foundries are crucial in underpinning Japan’s national security and industrial infrastructure, deeper discussions on their strategic importance and long-term viability are necessary. Strengthening this sector is not just about semiconductor manufacturing — it is about reinforcing the backbone of Japan’s economic and technological resilience.
3. Gathering Industry Consensus on the Development of LVP Fabs for 12-inch (65 –28 nm) Process Nodes
Given that domestically funded manufacturers have withdrawn from mass production at the 65–28nm process nodes (12-inch wafers), it is essential to thoroughly investigate the demand for long-tail semiconductors and the feasibility of target processes in this range. Semiconductor manufacturing equipment for 65–28nm nodes is designed for 12-inch wafer production. This means that operational innovations require transforming these fabs into LVP foundries with small MOQ capabilities. Unlike legacy 8-inch facilities, adapting a 12-inch fab for low-volume manufacturing presents additional challenges.
Moreover, introducing 12-inch semiconductor manufacturing equipment requires an investment of several tens of billions of yen. To ensure the economic viability of such a venture, it is crucial to gather input from Japan’s long-tail industries and establish a stable baseline demand to support fab operations. Building consensus among key stakeholders in the domestic semiconductor and long-tail product industries will be a critical step in shaping the future of Japan’s semiconductor ecosystem.
4. Recognizing the Need to Secure Not Only Cutting-Edge Semiconductor Manufacturing but Also Legacy Process Manufacturing for Long-Tail Industries
There is no room for doubt regarding securing cutting-edge semiconductor manufacturing facilities, as they are critical strategic assets. However, Japan also possesses legacy manufacturing facilities nurtured by major domestic IDMs, which should not be overlooked.
These legacy fabs represent valuable domestic assets that can play a crucial role as suppliers of long-tail semiconductors, another category of strategic resources. Rather than allowing these facilities to decline, efforts should be made to maximize their value and ensure their continued contribution to Japan’s industrial and national security interests.
Conclusion
It has been roughly 25 years since Japan’s major IDMs ceased supporting the domestic long-tail industries. Over this period, the knowledge, experience, cost awareness, and risk management skills related to ASIC development have largely been lost within the long-tail industry. At the same time, many 8-inch and 12-inch foundries that were spun off from major IDMs have been acquired by foreign entities and are now operated as mere extensions of global manufacturing networks. As a result, they often cannot prioritize the needs of Japan’s long-tail industries, making domestic supply chain support increasingly tricky.
Personally, I do not doubt the importance of long-tail semiconductors. However, ensuring the sustainability of the long-tail semiconductor ecosystem — including LVP (Low-Volume Production) foundries — requires active engagement from the long-tail industry itself. Given the nature of this ecosystem, it is crucial to foster a user community that can consolidate opinions, articulate concrete needs, and actively participate in shaping a sustainable support framework. Maintaining a viable ecosystem for long-tail semiconductors will be challenging without long-tail industries' direct involvement and commitment.
Having experienced the early days of EDA adoption in semiconductor design and worked hands-on with PDK implementation at pioneering domestic fabless companies, I remain committed to sharing well-balanced insights that bridge semiconductor manufacturing and design. By continuing to publish and share these discussions, I hope to contribute to the open-source silicon design community and Japan’s long-tail industries, legacy foundries, and the broader semiconductor ecosystem. I aim to support and strengthen these key players by providing informed perspectives that foster sustainable development in the industry.
