Monthly Report March 2024

TAIWAN AND
THE GLOBAL SEMICONDUCTOR
SUPPLY CHAIN

Edited by:
Chen-Yuan Tung, Ph.D.
Representative
Taipei Representative Office in Singapore

1 | 25 Taipei Representative Office in Singapore

Please feel free to reach out to the Economic Division of the Taipei
Representative Office in Singapore should you have any enquiries or are
seeking partnership opportunities of investment or collaboration in the field of
semiconductors in Taiwan.

Email: [email protected]
Telephone: +65 6500-0128

Published: Taipei Representative Office in Singapore

Address: 460 Alexandra Road, #23-00 mTower, Singapore 119963
Email: [email protected]
Telephone: +65 6500-0100

The Taipei Representative Office in Singapore provides monthly reports on

Taiwan and global semiconductor supply chains. We welcome you and your
friends to join our WhatsApp community, “Taiwan Semiconductor
Reports”.

https://chat.whatsapp.com/BqwdAMgi1sUIGsujDx3YDk

2 | 25 Taipei Representative Office in Singapore

IN THE SPOTLIGHT
Artificial Intelligence and the Semiconductor Industry
 Advancements in Artificial Intelligence (AI) and its widespread adoption
are opening up new frontiers in nearly all sectors.
 Semiconductor technology is at the core of all present and future AI-
related innovations.
 TSMC, Samsung and Intel are major players in AI chip fabrication.
 The opportunities and challenges posed by AI and the semiconductor
industry are of great interest to governments and policymakers
worldwide.

Source: Gorodenkoff, Shutterstock.com

OVERVIEW

Artificial Intelligence (AI) and semiconductors are two deeply

interconnected fields that are driving technology and innovation.
Semiconductors, the tiny chips that power electronic devices, play a crucial
role in the functioning of AI systems. In recent years, progress in
semiconductor technology, including the creation of smaller and more
powerful chips and the advent of advanced semiconductor packaging
techniques, has enabled tremendous technical progress in AI. In turn, AI is
revolutionizing and creating new opportunities in the semiconductor industry,
particularly in chip design, manufacturing technology and the supply chain.1

1
Tim Culpan, “AI is helping create the chips that design AI chips,” Bloomberg, May 25, 2023.

3 | 25 Taipei Representative Office in Singapore

 By stretching the limits of what is possible and opening up new frontiers
in nearly all sectors, AI is shaping the lives of people across the globe, including
the way we work, live and play. Indeed, AI and semiconductors are not only
integral to our lives but are also shaping our future. Due to their
transformative potential and the challenges that they pose, AI and by
extension, semiconductors are an expanding area of research and a policy
concern. 2

WHAT IS ARTIFICIAL INTELLIGENCE?

The United Nations Educational, Scientific and Cultural Organization

(UNESCO)’s Recommendation on the ethics of artificial intelligence notes that
“Artificial Intelligence (AI) systems are information-processing technologies
that integrate models and algorithms that produce a capacity to learn and to
perform cognitive tasks leading to outcomes such as prediction and decision-
making in material and virtual environments. AI systems are designed to
operate with varying degrees of autonomy by means of knowledge modelling
and representation and by exploiting data and calculating correlations.” 3

AI can be used both on the cloud and on the edge. Cloud AI involves
running AI applications in a cloud computing environment, where AI algorithms
and models are executed, and AI services are delivered using internet
technologies. Examples of Cloud AI services include Google Cloud AI, Amazon
AI Services, Microsoft Azure AI, and IBM Watson. 4 AI on the edge, on the other
hand, operates on local devices such as wearables, Internet of Things (IoT)
devices, or edge computing servers. 5 The use of AI both on the cloud and on
the edge makes AI more prevalent and accessible.

Technologies such as machine learning, deep learning, and natural

language processing are all part of the larger AI landscape. 6 These technologies
are used in applications in our daily lives, including autonomous driving, virtual

2
The White House, “Remarks by President Biden on Artificial Intelligence,” July 21, 2023.
3
UNESCO, Recommendation on the ethics of artificial intelligence, 2021, https://en.unesco.org/artificial-
intelligence/ethics.
4
Gartner Peer Insights, “Cloud AI Developer Services Reviews and Ratings, 2024,
https://www.gartner.com/reviews/market/cloud-ai-developer-services.
5
Nacho Palou, “Cloud AI vs. Edge AI: know their differences and choose the right approach for your AI
project,” Telefonica Tech, June 14, 2023.
6
Artificial Intelligence Board of America, “AI Strategies: Using NLP, ML, and DL for Businesses,” July 7, 2021.

4 | 25 Taipei Representative Office in Singapore

assistants, financial trading, spam filtering and language translation. AI
advances are also providing great benefits to our social wellbeing in areas such
as precision medicine, environmental sustainability, education, and public
welfare.7

A popular class of AI systems is deep neural networks (DNNs), which is

capable of learning and recognizing patterns in large amounts of data. DNNs
use algorithms, or models, to mimic neurons in the brain to identify complex
patterns. The computational power required by DNNs for its complex
mathematical modeling is provided by semiconductors. 8 DNNs involve two
main phases: training and inference. During training, the AI model is fed data
to identify patterns. In the inference stage, the trained model is used to enable
predictions and guide decisions, such as autonomous driving systems detecting
and avoiding obstacles on the road. In general, the accuracy of AI models
increases with training on larger amounts of data, which in turn requires more
computational power.

Traditional AI and Generative AI

Traditional AI and generative AI are two different approaches to artificial

intelligence, each with its own strengths and applications. Traditional AI
systems usually perform a specific task and include examples such as spam
filters, voice assistants like Siri or Alexa, recommendation engines on Netflix or
Amazon, or Google's search algorithm.9 In contrast, generative AI is usually
broader and can create new content, partly because generative AI tools are
trained on larger and more diverse data sets than traditional AI. Furthermore,
unlike traditional AI which is usually trained using supervised learning
techniques, generative AI is trained without continually needing human
intervention.10 Examples of generative AI include chatbots like OpenAI’s
ChatGPT, Microsoft's Copilot and Google's Gemini.

While traditional AI is used in finance, healthcare, and manufacturing

industries to automate processes and improve efficiency, generative AI opens

7
U.S. Department of State, “Artificial Intelligence (AI),” 2020, https://www.state.gov/artificial-intelligence/.
8
Congressional Research Service, “Semiconductors and Artificial Intelligence,” September 18, 2023.
9
Emily Heaslip, “What's the Difference Between Traditional and Generative AI?” CO by US Chamber of
Commerce, October 16, 2023.
10
Greg Pavlik, “What Is Generative AI? How Does It Work?” Oracle Singapore, September 15, 2023.

5 | 25 Taipei Representative Office in Singapore

new doors of creativity and adaptability, making it a powerful tool in various
industries. Industries relying most heavily on knowledge work are likely to see
more disruption—and potentially reap more value from generative AI. 11 As AI
research advances, both traditional AI and generative AI will complement each
other, contributing to a more robust and versatile AI ecosystem.

Popular large language models, such as Generative Pre-trained

Transformer 4 (GPT-4), are trained on trillions of text data to process and
generate text. 12 AI models are trained on specialized hardware like graphics
processor units, which draw more power and are generally more expensive
than traditional central processing units. Given the large data sets associated
with AI models, some of the largest AI models can take weeks or months to
train, using thousands of chips and costing millions of dollars. 13 As powering
and training of artificial intelligence models is energy intensive, the electricity
required to operate the hardware contributes to the high costs, and there are
also concerns on the environmental impact of the growing use of AI. 14

AI models are becoming ever more expensive to develop, in terms of

raw computational power and the training on massive datasets. According to
OpenAI, the research organization responsible for developing GPT-3, the
project’s total cost is estimated to be around US$ 4.6 million. 15 Meanwhile, the
newer GPT-4 model costs more than US$ 100 million to train during its initial
development and requires about US$ 700,000 a day to run. 16 The typical cost
of developing a large AI model may soon be in the billions of dollars. 17 As the
financial cost of AI development rises, only a few select labs can afford to do it,
and they will be the ones to set the agenda for what kinds of AI models get
developed.

11
McKinsey and Company, “The state of AI in 2023: Generative AI’s breakout year,” August 1, 2023.
12
Ernestas Naprys, “Scientists to make their own trillion parameter GPTs with ethics and trust,” Cybernews,
November 28, 2023.
13
Congressional Research Service, “Semiconductors and Artificial Intelligence,” September 18, 2023.
14
Lauren Leffer, “The AI Boom Could Use a Shocking Amount of Electricity,” Scientific American, October 13,
2023.
15
Kevin Pocock, “How much did GPT-3 cost?” PC Guide, August 11, 2023.
16
Erik Brynjolfsson and Gabriel Unger, “The Macroeconomics of Artificial Intelligence,” International Monetary
Fund, December 2023.
17
Craig Smith, “What Large Models Cost You – There Is No Free AI Lunch,” Forbes, September 8, 2023.

6 | 25 Taipei Representative Office in Singapore

SEMICONDUCTOR USE IN AI-BASED APPLICATIONS

AI models employ different types of chips. General chips such as logic

chips and memory chips respectively process and store large amounts of data,
making it possible to create and use AI. 18 In recent years, advanced
semiconductor packaging techniques, by enabling the integration of various
components from different wafers into a single unit, have played a crucial role
in the development of AI-specific chips such as Application-Specific Integrated
Circuits (ASICs) and Systems-on-a-Chip (SoCs).19

ASICs are single-purpose chips used for performing repetitive processing

routines such as scanning a barcode. SoCs, on the other hand, are essentially
integrator chips that combine many chips and circuits in a single chip. As both
ASICs and SoCs are designed from ground up for AI, they can perform more
computations per unit of energy, resulting in faster processing speeds and
lower energy consumption compared to general-purpose chips. 20

By delivering new functionalities, better performance and lower cost

with each generation, advances in chips have spawned new products and
transformed industries. The global semiconductor market is forecasted to grow
from 2022 to 2027, albeit with some fluctuations in between years, as demand
for consumer goods with AI applications amplifies across the globe. The global
semiconductor market size which was valued at US$ 599.6 billion in 2022 is
projected to fall to US$ 525.9 billion in 2023. From 2024 onwards, the market
is expected to grow from US$ 633.0 billion to US$ 762.9 billion in 2026, before
falling to US$ 751.6 billion in 2027 (See Figure 1).

2023 has been described as generative AI’s breakout year. 21 Following

its launch in November of 2022, OpenAI’s ChatGPT set the record as the
fastest-growing consumer application of all time and ushered generative AI to
the mainstream. Other generative AI tools such as Google’s Gemini and
Microsoft’s Copilot AI came quickly on the heels of ChatGPT’s success. As the
year progressed, numerous startups also emerged to innovate with generative
18
Congressional Research Service, “Semiconductors and Artificial Intelligence,” September 18, 2023.
19
Ondrej Burkacky, Taeyoung Kim, and Inji Yeom, “Advanced chip packaging: How manufacturers can play to
win,” McKinsey, May 24, 2023.
20
Saif M. Khan and Alexander Mann, Policy Brief: “AI Chips: What They Are and Why They Matter, An AI Chips
Reference”, Center for Security and Emerging Technology, April 2020.
21
McKinsey, “The state of AI in 2023: Generative AI’s breakout year,” August 2023.

7 | 25 Taipei Representative Office in Singapore

AI. 22 Consequently, AI semiconductor revenue is increased by the growing use
of generative AI in data centers in the near term, and by the deployment of AI
into PCs, smartphones, edge and endpoint devices in the longer term. 23

Figure 1: Global AI Semiconductor Market Trend for 2022 to 2027

Source: Hsuan Chih Wang, "The AI Semiconductor Market Shows Strong Growth, with Emerging Consumer AI
Applica�ons," IEK, ITRI, October 2, 2023, p. 1.

Table 1: Global AI Semiconductor Market Growth Trend: 2023-2027

2023 2024 2025 2026 2027
Global Semiconductor Market -12.3% 20.4% 14.3% 5.4% -1.5%
Global AI Semiconductor Market 20.9% 25.6% 23.2% 19.8% 20.5%
Source: Hsuan Chih Wang, "The AI Semiconductor Market Shows Strong Growth, with Emerging Consumer AI
Applica�ons," IEK, ITRI, October 2, 2023, p. 1.

Despite rapid technological progress in AI making predictions of the AI

semiconductor market challenging, the Al semiconductor market is expected
to show a rapid growth trend compared with the global semiconductor

22
Lori Perri, “What’s New in Artificial Intelligence from the 2023 Gartner Hype Cycle,” Gartner, August 17,
2023.
23
Alan Priestley, “Forecast: AI Semiconductors, Worldwide, 2021-2027, 3Q23 Update,” Gartner Research,
October 17, 2023.

8 | 25 Taipei Representative Office in Singapore

market. When the global semiconductor market declined 12.3% to US$ 525.9
billion in 2023, the Al semiconductor market was forecasted to grow from
US$ 44.2 billion in 2022 to US$ 53.4 billion in 2023 (see Figure 1), an increase
of 20.9% from 2022, according to Gartner (see Table 1). 24

With more users experimenting directly with generative AI applications

and the increasing use of a wide range AI-based applications in data centers,
edge infrastructure and endpoint devices, the demand for high-performance
graphics processing units (GPUs) and optimized semiconductor devices is
soaring. In fact, worldwide shipments of AI PCs and generative AI smartphones
are projected to total 295 million units by the end of 2024, up from 29 million
units in 2023, according to a forecast from Gartner, Inc. 25 Gartner estimates
240 million generative AI smartphones and 54.5 million AI PCs will be shipped
by the end of 2024 (see Figure 2). This will represent 22% of basic and
premium smartphones and 22% of all PCs in 2024. Hence, AI semiconductor
revenue is expected to continue to experience double-digit growth from 2023 to
2027 (See Table 1). By 2027, AI chips revenue is expected to be more than double
the size of the market in 2023, reaching US$ 119.4 billion (See Figure 1).

In a nutshell, Al semiconductors’ share in the global semiconductor market

will see a steady increase – from 7.4% in 2022 to 15.9% in 2027. This increase of
8.5 percentage points from 2022 to 2027 reflects the growing adoption of Al
technology. It also means that in the next few years, the Al semiconductor market
will account for significantly more than 10% of the global semiconductor market.
Clearly, the Al semiconductor market is becoming an integral part of the global
semiconductor market (see Figure 1).

24
Gartner, Press Release “Gartner Forecasts Worldwide AI Chips Revenue to Reach $53 Billion in 2023 by 2027,
AI Chips Revenue Is Forecast to More Than Double,” August 22, 2023.
25
Gartner, Press Release: “Gartner Predicts Worldwide Shipments of AI PCs and GenAI Smartphones to Total
295 Million Units in 2024,” February 7, 2024.

9 | 25 Taipei Representative Office in Singapore

Figure 2: AI PCs and Generative AI Smartphones Market Share, Worldwide,
2023-2025
AI PC Share of All PCs
Generative AI Smartphones
AI PCs Generative AI Share
of Basic and
Premium
Smartphones

Source: Gartner, Press Release: “Gartner Predicts Worldwide Shipments of AI PCs and GenAI Smartphones to
Total 295 Million Units in 2024,” February 7, 2024.

KEY PLAYERS IN AI CHIP FABRICATION

The semiconductor industry's big three – Taiwan Semiconductor

Manufacturing Company (TSMC), Samsung and Intel – are key players in
fabricating chips, especially essential ones for AI development. According to
TrendForce, TSMC had 57.9% of the market share of the third-party chip-
manufacturing business, followed by Samsung with 12.4% and Intel Foundry
Services with 1.0% in the third quarter of 2023. 26 All three are racing to
produce 2 nanometer (nm) processor chips that will power the next generation
of smartphones, data centers, and artificial intelligence by 2025. 27 Several
media outlets have reported Samsung is targeting mass production at the 1.4

26
Chang Chien-chung and Frances Huang, TSMC cements global lead as market share rises to 57.9% in Q3,
Focus Taiwan, December 9, 2023.
27
Yifan Yu, “Intel joins 1.4-nanometer chip race against TSMC and Samsung,” Nikkei Asia, February 22, 2024.

10 | 25 Taipei Representative Office in Singapore

nm level in 2027 while TSMC is aiming for 2027 to 2028. Of the Big Three,
TSMC remains the analysts’ favorite to maintain its global supremacy in the
sector. 28

Besides reducing nanometer size to yield more powerful and efficient

chips, today’s most powerful AI chips require advanced packaging methods to
enable the greatest performance. 29 TSMC, with a cache of 2,946 advanced
packaging patents, has developed the most expansive arsenal of patents
surrounding advanced chip packaging, followed by Samsung Electronics (2,404)
and then Intel (1,434), according to LexisNexis. 30

TSMC’s Chip-on-Wafer-on-Substrate (CoWoS®), Intel’s Embedded Multi-

Die Interconnect Bridge (EMIB), and Samsung’s Interposer-Cube (I-Cube) are
mature advanced semiconductor packaging technologies that are widely
utilized in high-performance chips.31 By 2024, suppliers are expected to focus
on ramping up 2.5D packaging capacity to meet the rising demand for high
computational power in applications like AI. Additionally, both TSMC and
Samsung have also taken steps to establish 3D integrated circuit research
centers in Japan, highlighting the critical role of packaging in the evolution of
semiconductor technology. 32

With their ability to innovate and adapt to the rapidly evolving AI

landscape, these companies are poised to ride the AI boom and continue their
lead in the years to come.

TSMC

TSMC is a major provider of logic chips for AI. As a leader in advanced

chip packaging technology and the world’s largest contract chipmaker, TSMC is
a key partner to both established and startup AI chip designers. TSMC

28
Christian Davies, Song Jung-a, Kathrin Hille and Qianer Liu, “Semiconductor giants race to make next
generation of cutting-edge chips,” Financial Times, December 11, 2023.
29
Refer to the January issue of “Taiwan and the Global Semiconductor Supply Chain” for more information.
30
Max A. Cherney, “TSMC leads in advanced chip packaging wars, LexisNexis patent data says,” Reuters,
August 2, 2023.
31
Trendforce, Press Release: “TrendForce 2024: Riding the Wave of Revolutionary Tech Trends,” October 17,
2023.
32
TSMC, Press Release: “TSMC Japan 3DIC R&D Center Completes Clean Room Construction in AIST Tsukuba
Center,” June 24, 2022; Jin Eun-Soo, “Samsung to build $280 million chip research center in Japan,” Korea
JoongAng Daily, December 21, 2023.

11 | 25 Taipei Representative Office in Singapore

produces chips designed by tech giants such as NVIDIA, the frontrunner in the
AI chip race; Advanced Micro Devices (AMD), NVIDIA’s rival; and Arm Holdings,
which has a significant processor share across Google's Android and Apple's
iOS devices. 33 Notable AI chip design startups such as US-headquartered
SambaNova and UK-based Graphcore, which both specialize in AI chips for
machine learning, are also among TSMC’s customers. 34 Graphcore, for
example, has used TSMC’s wafer-on-wafer 3D integration technology to speed
up AI processing by 40%. 35 TSMC also produces customized AI chips for data
centers such as Amazon Web Services (AWS), Microsoft and Google, and other
companies such as Apple.36 TSMC’s partnerships with these companies allow
them to focus on designing innovative AI chips, while TSMC handles the
manufacturing process.

At an investor conference on January 18, 2024, TSMC projected a year-

on-year sales increase of 21-26% in 2024 due to the booming demand for high-
end chips used in AI applications. In view of the company's lead over its rivals
in advanced technology development, TSMC’s 2024 sales projections are
higher than an expected 20% in the global pure play wafer foundry industry. 37

Additionally, TSMC has already disclosed the test results of its 2 nm

prototype process to major clients such as Apple and NVIDIA, and Apple is set
to become TSMC’s inaugural customer for the 2 nm process when the mass
production of its 2nm chips commences in 2025. 38

Samsung

Samsung is the world’s largest memory chipmaker. A surge in demand

for high bandwidth memory (HBM), a critical component in all AI chips, has
meant new growth for memory-chip makers. According to Morgan Stanley, the
global HBM market, which grew from US$ 1.9 billion in 2022 to an estimated

33
Aditya Soni and Roshan Abraham, “Arm's shares get helping hand as IPO banks weigh in with ratings,”
Reuters, October 9, 2023.
34
SambaNova, Press Release: “SambaNova Unveils New AI Chip, the SN40L, Powering its Full Stack AI
Platform,” September 19, 2023.
35
Samuel Moore, “Graphcore Uses TSMC 3D Chip Tech to Speed AI by 40%,” IEEE Spectrum, March 3, 2022.
36
Jai Vipra & Sarah Myers West, AI Now Institute, Computational Power and AI,” September 27, 2023.
37
Chang Chien-chung and Frances Huang, “TSMC forecasts sales to grow by over 20% in 2024,” Focus Taiwan,
January 18, 2024.
38
Trendforce, Press Release: “TSMC’s Primary Client for 2nm Chips Expected to be Apple, Set to Debut with
iPhone 17 Lineup Next Year,” January 26, 2024.

12 | 25 Taipei Representative Office in Singapore

US$ 4 billion in 2023, is forecasted to more than double in 2024. The strong
demand for HBM memory chips is a boon to Samsung, which is expected to
command a 47-49% share of the global HBM market this year. 39

Besides memory chips, Samsung also produces its own AI chips, such as
Exynos and ISOCELL.40 Samsung has also innovated new types of solid-state
drives (SSDs) optimized for AI and machine learning applications. 41

Several media have reported that Samsung is trying to seize the market
for its 2 nm process by offering lower prices. 42 The first beneficiary of this
discount is Preferred Networks (PFN), which has reportedly chosen Samsung
over TSMC. 43 PFN is a Japanese startup known for its expertise in AI deep
learning development and has attracted large-scale investments from major
companies, including Toyota Motor, NTT and FANUC, a Japanese robot system
maker. According to Korean media, PFN decided to forgo TSMC as Samsung
can provide a comprehensive chip-making service on a turnkey basis, from chip
design to production and advanced chip packaging.44

Intel

Intel has set itself a goal of ensuring that 50% of the world’s
semiconductors are built in the United States and Europe in a decade,
compared with 20% today.45 As part of its ‘IDM 2.0’ strategy, Intel is
reorganizing into two units — Intel Foundry, which manufactures
semiconductors, and Intel Products, involved in their design. 46 Launched on
February 21, 2024 as the world’s first systems foundry for the AI era, Intel
Foundry positions itself as a ‘systems foundry’, which brings system-level
39
Jeong-Soo Hwang, “Samsung eyes $8.8 billion 2024 chip profit: sources, The Korea Economic Daily, January
2, 2024.
40
Jeong-Soo Hwang, “Samsung Electronics to up AI chip foundry sales to 50% by 2028,” November 20, 2023.
41
Samsung, “How Samsung Memory Is Powering the AI Revolution,” May 16, 2023,
https://semiconductor.samsung.com/us/news-events/tech-blog/how-samsung-memory-is-powering-the-ai-
revolution/
42
Christian Davies, Song Jung-a in Seoul, Kathrin Hille in Taipei and Qianer Liu, “Semiconductor giants race to
make next generation of cutting-edge chips,” Financial Times, December 11, 2023.
43
Trendforce, Press Release: “Samsung Secures 2-Nanometer Order from Japanese AI Startup Preferred
Networks,” February 16, 2024.
44
Jeong-Soo Hwang, “Samsung wins first 2 nm AI chip order from Japan’s PFN; a snub to TSMC,” The Korea
Economic Daily, February 16, 2024.
45
Michael Acton, “Intel to manufacture chips for Microsoft as AI drives demand,” Financial Times, February
22, 2024.
46
Ibid.

13 | 25 Taipei Representative Office in Singapore

design, technology development, supply chain capabilities, packaging,
connectivity fabric solutions, and cooling technology under one umbrella. 47
Intel has roped in support from ecosystem partners – including Synopsys,
Cadence, Siemens and Ansys – to expedite chip designs for Intel Foundry
customers using validated tools, design flows, and IP portfolios for Intel’s
advanced packaging and 18A process technologies. Under a US$ 15+ billion
deal, Microsoft will be leveraging Intel's 18A process technology to
manufacture an undisclosed AI chip that Microsoft itself designed. 48

Serving both external and internal customers, Intel Foundry aims to be

the second-largest foundry globally by 2030 and expects to beat an internal
deadline of 2025 to overtake its biggest rival, TSMC, in advanced chip
manufacturing. 49

IMPACT OF AI ON THE SEMICONDUCTOR INDUSTRY

The rise of AI has led to a surge in demand for semiconductors, driving

innovation and growth in the semiconductor industry. AI applications require
chips that are faster, more power-efficient, and capable of handling vast
amounts of data. This has led to the development of advanced semiconductor
technologies, such as the 3 nanometer process nodes currently used by TSMC
and Samsung,50 the exploration of new materials beyond silicon,51 and
advanced packaging techniques.52

AI is increasingly prevalent in the semiconductor industry. One area

where AI is playing a significant role is in the design of semiconductor chips.
Advanced AI tools can test human designs by finding placement errors that
increase power consumption, impede performance, or use space inefficiently;
suggesting improvements; and then simulating and testing those. Additionally,
AI tools are more efficient in defect detection. For example, AI can be used to

47
Intel, Press Release: “Intel Launches World’s First Systems Foundry Designed for the AI Era,” February 21,
2024.
48
Stephen Nellis and Max A. Cherney, “Intel signs Microsoft as foundry customer, says on track to overtake
TSMC,” Reuters, February 22, 2024.
49
Ibid.
50
Alan Patterson, “Samsung Seen Stumbling at Silicon’s Leading Edge,” EE Times Asia, December 29, 2023.
51
Milton D'Silva, “Beyond Silicon – What Will Replace the Wonder Material?” EMEA Industry, February 28,
2023.
52
Ondrej Burkacky, Taeyoung Kim, and Inji Yeom, “Advanced chip packaging: How manufacturers can play to
win,” McKinsey & Company, May 24, 2023.

14 | 25 Taipei Representative Office in Singapore

improve fault detection by visual inspection of wafers by almost nine times. 53
Compared to traditional methods, advanced AI tools can enable faster and
more efficient design of new chips.

Moreover, AI is increasingly being used in the semiconductor supply

chain.54 Chip companies can use AI to address supply chain challenges such as
managing a network of outsourced semiconductor assembly and test (OSAT)
providers. AI-based solutions also provide near-real-time insights into pricing
and demand fluctuations, helping companies build resilience into their supply
chains.55 According to Gartner, more than 80% of new supply chain
management technology applications will use AI and data science in some way
by 2025. 56 Those who don’t invest now risk falling behind.

TSMC, which projected upwards of 20% growth in 2024 driven by

demand for artificial intelligence, leverages AI to enhance its operations.
Besides using AI for chip design, demand forecasting and streamlining its
operations, TSMC also uses AI for its trade secret management. As of January
2021, over 100,000 trade secrets have been registered and recorded by 30,000
employees in its trade secret management system. 57 In 2023, TSMC developed
a customized AI talent scouting system called ‘Trade Secret Intelligent
Management Version 2.0: Innovation Talent Scouting Online Merge Offline
Service (“ITS OMO”)’ to develop a greater pool of prospective talent and
further encourage innovations. 58

Like TSMC, Samsung also uses AI to improve its chip design and
production processes. According to South Korean media reports, Samsung
Electronics is planning to fully automate its semiconductor factories by 2030,
with “smart sensors” set to control the manufacturing process. In doing so,

53
Tobias Schlosser et al., “Improving automated visual fault inspection for semiconductor manufacturing using
a hybrid multistage system of deep neural networks,” Journal of Intelligent Manufacturing 33: pp. 1099–
1123, 2022.
54
Institute of Electrical and Electronics Engineers (IEEE), “International Roadmap for Devices and Systems,
2020 Edition: Semiconductors and Artificial Intelligence”, 2020, https://irds.ieee.org/topics/semiconductors-
and-artificial-intelligence.
55
Francisco Caudillo, Michael Gordon, Wim Gysegom and Juan Sarmiento, “Chip hunting: The semiconductor
procurement solution when other options fail,” McKinsey & Company, April 4, 2023.
56
Nick Carey, “Focus: Startups apply artificial intelligence to supply chain disruptions,” Reuters, May 3, 2022.
57
Yu-Chieh Wu, “TSMC Continues to Use Artificial Intelligence (AI) for Trade Secret Management to Sustain
Innovation and Strengthen Competitiveness,” TSMC ESG, January 21, 2021.
58
Fortune Shieh, “TSMC Trade Secret Intelligent Management 2.0: Radical Innovation for Talent
Development,” TSMC ESG, September 19, 2023.

15 | 25 Taipei Representative Office in Singapore

Samsung aims to create an “artificial intelligence fab” that operates without
human labor. 59

Intel has been harnessing AI for critical tasks such as yield optimization,
defect detection, and process control for almost two decades. 60 Data on Intel’s
supply chain, which spans order taking, resource procurement, manufacturing,
testing, and final delivery of products are analyzed using AI to optimize
business processes, drive operational excellence, and provide higher levels of
customer satisfaction. 61 Intel’s warehouses, for example, are highly
automated, with RFID and barcode scanners tracking shipments and inventory
levels. Through predictive analytics, Intel can anticipate customer needs and
optimize its supply chain operations. 62

POLICIES ON AI CHIPS AND ARTIFICIAL INTELLIGENCE

US Exports Controls on AI Chips

The United States’ goal of export controls on AI chips was to restrict

China’s ability to both purchase and manufacture certain high-end chips that
could fuel breakthroughs in artificial intelligence and give China a military
advantage. 63 This is especially important in the face of China’s military-civil
fusion strategy that targets technologies such as quantum computing,
semiconductors, 5G, nuclear technology, aerospace technology, gene editing
and artificial intelligence to achieve military dominance. 64 Furthermore, China
has set 2030 as its target date to become a global leader in artificial
intelligence, with the subsequent goal of putting the People’s Liberation Army

59
Scott Foster, “Samsung to build all-AI, no-human chip factories,” Asia Times, January 16, 2024. Also refer to
the November issue of “Taiwan and the Global Semiconductor Supply Chain for more information on the
U.S. – China conflict over semiconductors.
60
Rao Desineni and Eugene Tuv, White Paper: “High-Value AI in Intel’s Semiconductor Manufacturing
Environment,” 2022, https://www.intel.com/content/dam/www/central-libraries/us/en/documents/ai-in-
semiconductor-manufacturing-paper.pdf.
61
Craig Chvatal and Anil Varhadkar, White Paper: “Transforming Intel’s Supply Chain with Real-Time
Analytics,” Intel, September 2017.
62
Ibid.
63
U.S. Department of Commerce’s Bureau of Industry and Security (BIS), Press Release: “Commerce
Implements New Export Controls on Advanced Computing and Semiconductor Manufacturing Items to the
People’s Republic of China (PRC),” October 7, 2022.
64
Nicholas R. Licata, “China’s Military-Civil Fusion Strategy: A Blueprint for Technological Superiority,” Foreign
Policy Research Institute, December 19, 2023.

16 | 25 Taipei Representative Office in Singapore

on par with the U.S. military by 2035 — a goal the export controls intend to
complicate.65

The U.S. Department of Commerce’s Bureau of Industry and Security

(BIS) implemented a series of export controls on AI chips in October 2022 and
October 2023. 66 The rules in October 2023 extended export controls to chips
with fewer capabilities than previously restricted in October 2022, and
included additional types of chip manufacturing equipment. Chipmakers
including NVIDIA, AMD, and Intel, which make the most popular chips for the
AI industry, were among those affected by the rules. 67

NVIDIA’s A800 and H800 chips, which can perform complex AI tasks and
are in high demand in China after the U.S.’s October 2022 export controls,
were hit by the updated rules. To preserve its market share in China, NVIDIA is
planning to launch the H20, L20 and L2 chips, which include most of its newest
features for AI work but with computing power cut back to comply with the
new rules, in the second quarter of 2024. 68

Additionally, the BIS’s “Advanced Computing Foreign Direct Product

Rule” means that chip manufacturing facilities globally were restricted from
manufacturing certain advanced chips for Chinese-headquartered chip design
firms without a license if the manufacturer uses U.S.-origin technology or
software. As the United States is a global leader in the production of
semiconductor manufacturing equipment, this rule would apply to most
foreign chip manufacturing firms, including TSMC, which previously produced
advanced chips for Chinese AI chip design companies such as Biren Technology
and Cambricon Technologies. 69

65
Manya Koetse, “In the race for AI supremacy, China and the US are travelling on entirely different tracks,”
The Guardian, January 9, 2024.
66
U.S. Department of Commerce’s Bureau of Industry and Security (BIS), Press Release: “Commerce
Implements New Export Controls on Advanced Computing and Semiconductor Manufacturing Items to the
People’s Republic of China (PRC),” October 7, 2022.; U.S. Department of Commerce’s Bureau of Industry and
Security (BIS), Press Release: “Commerce Strengthens Restrictions on Advanced Computing Semiconductors,
Semiconductor Manufacturing Equipment, and Supercomputing Items to Countries of Concern,” October 17,
2023.
67
Kif Leswing, “U.S. curbs export of more AI chips, including Nvidia H800, to China,” CNBC, October 17, 2023.
68
Yelin Mo and Fanny Potkin, “Nvidia to launch China-focused AI chip in Q2 2024 – sources,” Reuters, January
9, 2024.
69
Debby Wu, “TSMC Suspends Work for Chinese Chip Startup Amid US Curbs,” Bloomberg, October 23, 2022.

17 | 25 Taipei Representative Office in Singapore

 Both Biren Technology and Cambricon Technologies, which positioned
themselves as challengers to industry giants like NVIDIA and AMD, faced
significant challenges, including supply chain disruptions due to U.S. export
controls. Being blacklisted by the U.S. government, they no longer have access
to TSMC's advanced process technologies and must rely on Chinese foundries
that are years behind those of TSMC. As a result, Biren Technology and
Cambricon Technologies’ products have become less competitive and relevant
than before. Although Biren Technology recently received a US$ 280 million
cash infusion from the Guangzhou government-backed investors, a series of
founder departures has raised concerns about the company’s stability and
future direction. 70 Meanwhile, Cambricon Technologies is anticipating a net
loss ranging from RMB 756 million (US$ 105.1 million) to RMB 924 million yuan
(US$ 128.5 million) for 2023, marking the seventh consecutive year of losses
for the company. Its problems got worse when Huawei, a major client, started
designing its own AI processors, cutting it off. After undergoing multiple
rounds of layoffs in 2023, Cambricon Technologies is encountering significant
obstacles in promoting the adoption of Chinese-made AI chips, and profitability
remains a distant prospect. 71

Biren Technology and Cambricon Technologies are not the only AI GPU
developers in China, but their struggles reflect that of the whole Chinese AI
hardware in general. To survive the new odds, Biren Technology and
Cambricon Technologies will now have to redouble their efforts to seek self-
sufficiency in building chip-making tools and equipment. 72

Meanwhile, the Chinese government is pushing for semiconductor self-

sufficiency as a strategic move aimed at reducing dependence on imported
semiconductors, fostering domestic innovation, and positioning China as a
global leader in the semiconductor industry. 73 The Chinese government has
been working to convince domestic technology firms to source their inputs

70
Anton Shilov, “China's largest AI GPU maker loses its CEO despite generous government funding —
development calls into question China's GPU advances,” Tom’s Hardware, January 25, 2024.
71
China Money Network, “China’s Nvidia Contender Cambricon Tech Struggles to Push Chinese-Made AI
Chips as It Marks 7th Consecutive Year of Losses,” February 9, 2024.
72
Ann Chao, “Tech war: US sanctions on Biren and Moore Threads strike strong blow to China’s GPU
champions,” October 18, 2023.
73
Refer to the December issue of “Taiwan and the Global Semiconductor Supply Chain” for more information
on China’s semiconductor industry.

18 | 25 Taipei Representative Office in Singapore

domestically rather than from U.S. suppliers. 74 According to a report from The
Wall Street Journal in October 2023, the Chinese semiconductor industry’s
reaction to NVIDIA’s downgraded chips was cool as major Chinese players like
Alibaba and Tencent are redirecting some advanced semiconductor orders to
domestic companies and relying more on internally developed chips. This trend
is also observed with two other major chip buyers, Baidu and ByteDance. 75

Despite impressive progress to date, indigenous Chinese chips will likely

lag in performance compared to chips from the United States and its allies for
years to come. There also exists the possibility of the U.S. further tightening AI
chips export controls in the near future. So far, the U.S. controls targeted at AI
chips are missing a key workaround component: cloud computing. 76 A gap in
export control policy allows Chinese firms to access highly advanced
computers physically located in different countries through the cloud.
Additionally, one early study, from the Center for a New American Security
(CNAS), a Washington-based non-profit organization, said in October 2023 that
thousands of controlled AI chips could have been smuggled into China in
2023. 77 It was reported in July that BIS was considering a rule change, and its
solicitation for public comment suggests that an additional update to current
rules may be forthcoming.78

CONCLUSION

The relationship between semiconductors and AI is symbiotic. While

semiconductors provide the necessary hardware for AI systems, the increasing
use of AI is driving the growth and innovation in the semiconductor industry.
As AI continues to evolve, we can expect to see further advancements in
semiconductor technology, paving the way for more powerful, efficient, and
intelligent systems.

74
Brenda Goh and Katerina Ang, “China rushes to swap Western tech with domestic options as U.S. cracks
down,” Reuters, October 27, 2023.
75
Trendforce, “Cooling Response to NVIDIA’s Exclusive Chips for China, Lack of Interest in Downgraded Models
by Customers,” January 8, 2024.
76
William Alan Reinsch, Matthew Schleich and Thibault Denamiel, “Insight into the U.S. Semiconductor Export
Controls Update,” Center for Strategic and International Studies, October 20, 2023.
77
Tim Fist and Erich Grunewald, A Working Paper: “Preventing AI Chip Smuggling to China,” Center for a New
American Security, October 24, 2023.
78
Yuka Hayashi and John D. McKinnon, “U.S. Looks to Restrict China’s Access to Cloud Computing to Protect
Advanced Technology,” The Wall Street Journal, July 4, 2023.

19 | 25 Taipei Representative Office in Singapore

 The global semiconductor market is forecasted to grow from 2022 to
2027. 2023 marked a pivotal moment in the growth trajectory of AI, setting
the stage for rapid advancements and widespread adoption in the years to
come. Consequently, compared to the global semiconductor market, the Al
semiconductor market is expected to grow even more rapidly, with chips
designed to accelerate generative AI workloads constituting the fastest-
growing market in semiconductor sales.

In addition, tech supremacy is closely intertwined with advancements in

semiconductors and artificial intelligence. The on-going race for tech
supremacy between the United States and China means the prospect of U.S.
further tightening its export restrictions on high-performance computing chips
to China and select locations is still hovering in the horizon, and that chip
company revenues and growth could be negatively impacted. Furthermore,
responses from restricted countries could also adversely affect companies,
industries, and the global economy.

The export controls imposed by the United States on AI chips will have
far-reaching implications for the Chinese advanced semiconductor sector.
Among the most affected will be the Chinese AI data center companies, which
rely heavily on cutting-edge chips for their operations. While these export
controls will significantly harm Chinese AI data center companies in the short
term, they will add new impetus for China to strengthen its domestic chip
design ecosystem in the long run.

In a nutshell, the integration of AI and semiconductors has profound

implications on both economic and military development, and the
opportunities and challenges posed by AI and the semiconductor industry are
of great interest to governments and policymakers worldwide. AI is not only
transforming chip design and manufacturing but also reshaping the entire
semiconductor industry landscape and unlocking new possibilities and
economic value.

*****

20 | 25 Taipei Representative Office in Singapore

TaiwanPlus
• Curious about Taiwan?
• Planning a visit?
• Want to stay up to date on
regional politics?

TaiwanPlus is the premier international media platform providing English-

language news and entertainment offering independent and impartial daily
news from Taiwan. From its unique position at the nexus of geopolitics and
international trade, TaiwanPlus also provides an inside look at Taiwan-China
relations, with in-depth reporting and analysis, and showcases Taiwan’s unique
culture and lifestyle, offering some of the nation's best food, music, and travel
programming. For more information, visit https://www.taiwanplus.com/

TaiwanPlus Videos on Semiconductors

2024/02/06
Taiwan’s Mega
Factories

Inside Micron
Taiwan’s
Semiconductor
Factory - Taiwan's
Mega Factories
EP1

In the first episode of Taiwan’s Mega Factories, join us for a tour of

Micron Technology’s Taiwan chip manufacturing facili�es to discover
how chips are produced and how innova�ons in the semiconductor
industry impact people and businesses all over the world. Micron
leaders share how they are expanding their innova�ve approach into
the area of resource management, enabling them to create more
sustainable products.

*****

21 | 25 Taipei Representative Office in Singapore

SEMICONDUCTOR STATISTICS AT A GLANCE

GLOBAL TRENDS

End-User Products and Semiconductor Application

Figure 3: Worldwide Shipment Volume and

Growth Rate of End-User Electronics Products: 2023

Consumption,

Source: Hui-Hsiu Huang, “Advanced Process Technologies Driving End- Applica�on Innova�on and Market
Compe��on,” IEK, ITRI, November 10, 2023, p. 1.

Semiconductor chips are deployed across a wide range of applica�ons,

transforming various sectors and industries. Figure 3 shows the worldwide
shipment volume and growth rate of end-user electronics products in 2023.

According to Gartner’s forecast in October 2023, the global shipment of

general electronics products was the largest, reaching 26.4 billion units in 2023,
represen�ng an annual growth rate 3.4%. Computing electronic products at
10.7 billion units, accounted for the largest shipment, followed by the category
of industrial/military/aviation/other electronic products, with shipments
reaching 9.68 billion units in 2023.

22 | 25 Taipei Representative Office in Singapore

 In terms of annual growth rate of end-user electronics products in 2023,
the category of industrial, military, aviation/other electronic products had the
highest annual growth rate, reaching 8.9%. Such electronic products require
advanced semiconductors that provide both high performance and high
reliability.

Coming in second to the category of industrial/military/aviation/other

electronic products was automotive electronic products with an annual growth
rate of 7.4%. Semiconductors are used in automo�ve electronic products in
areas such as automobile safety, control and entertainment. With the
improvement of automobile intelligence and electrifica�on, the demand for
automo�ve electronic products, and by extension, semiconductors, will
con�nue to grow.

Other end-user electronic products, including communications,

computing, and consumer electronics products witnessed posi�ve annual
growth rates in 2023. The category of storage electronic products was the only
excep�on, with worldwide shipment of storage electronic products shrinking
by 7.9% compared to the previous year. Likewise, the memory semiconductors
industry, also faced a significant downturn in 2023.

Overall, the global end-user electronics market s�ll maintained growth in

2023. In general, although the global semiconductor market was affected by
factors such as the global economic downturn in 2023, it s�ll showed posi�ve
growth. With the con�nued growth of high-end electronic products and the
automo�ve semiconductor markets, the long-term growth trend of the global
semiconductor market looks promising.

23 | 25 Taipei Representative Office in Singapore

 Figure 4: Global Market Size and Growth Rate of Semiconductor Application:
2023

Source: Hui-Hsiu Huang, “Advanced Process Technologies Driving End- Applica�on Innova�on and Market
Compe��on,” IEK, ITRI, November 10, 2023, p. 6.

Figure 4 shows the global market size and growth rate of semiconductor
applica�ons. According to data from Gartner in October 2023, the global
semiconductor market size was es�mated to be US$ 534.5 billion in 2023, a
decrease of 10.9% compared to 2022. Of the products where semiconductors
are deployed, the category of communication products is the largest market at
US$ 154.6 billion, followed by computing products at US$ 142 billion,
accoun�ng respec�vely for 28.9% and 26.6% share of the global market.

In terms of market size, the automo�ve semiconductor market, valued

at US$ 79 billion, is the largest. In fact, major players including NXP
Semiconductors, which provide microcontroller and analog chips for various
automo�ve applica�ons, and Tesla, which is known for its electric vehicles and
self-driving technology, have launched automo�ve processors using advanced
processes to meet the increasingly complex automo�ve applica�on needs.
These advanced process automo�ve processors have higher performance,
lower power consump�on, and can provide safer and more advanced func�ons
for cars.

24 | 25 Taipei Representative Office in Singapore

 It is also worth no�ng that the compound annual growth rate (CAGR) of
the automo�ve semiconductor market from 2022 to 2027 is 12.7%, which is
higher than the CAGR of the overall semiconductor market of 5.0%. It is the
only market in the semiconductor applica�on field to show double-digit growth
in 2023, making it a significant driver of technological advancement in the
automo�ve industry. The integra�on of semiconductors with AI and other
cu�ng-edge technologies is fueling this growth, leading to innova�ons in
safety, efficiency, and vehicle electrifica�on. As the automo�ve sector
con�nues to evolve, semiconductors play a pivotal role in shaping the future of
mobility. With the improvement of automo�ve electronics and advances in
ar�ficial intelligence, it is foreseeable that the demand for automo�ve
semiconductors will con�nue to grow in the future.

*****

25 | 25 Taipei Representative Office in Singapore