The "Creators" behind the chips usher in a brand new era! Wall Street senses the scent of a "bull market".

After a hiatus of four years, Taiwan Semiconductor, known as the "king of chip manufacturing," has returned to the top ten global market capitalization list. Its US-listed ADR has surged by as much as 37% this year, significantly outperforming the Philadelphia Semiconductor Index. Against the backdrop of the surging demand for AI chips, Taiwan Semiconductor, as the global leader in AI chips, serving as the sole foundry for NVIDIA and AMD's AI chips, as well as the sole foundry for self-developed AI chips by cloud giants like Microsoft and Amazon, is bound to continue benefiting. With its top-notch chip manufacturing technology and cutting-edge Chiplet advanced packaging, Taiwan Semiconductor's "NVIDIA moment" - the moment when stock price and performance synchronize for a substantial increase, may already be here.

At the same time, the "masters" who jointly create AI chips, namely the key semiconductor equipment suppliers of Taiwan Semiconductor, are expected to see their sales continuously expand starting in 2024. This is mainly due to chip manufacturers such as Taiwan Semiconductor, Samsung Electronics, and Intel continuously expanding their production capacity based on advanced processes of 5nm and below for AI chips. Additionally, Taiwan Semiconductor's large chip manufacturing plants in the United States, Japan, and Germany are expected to be completed this year, leading to large-scale purchases of high-end semiconductor equipment such as lithography machines, etching equipment, and film deposition equipment required for chip manufacturing. These semiconductor equipment suppliers mainly include ASML, Applied Materials, Tokyo Electron, and BE Semiconductor, among other top-tier equipment suppliers in the chip industry.

In 2023, ChatGPT swept the globe, followed by the grand debut of Sora's text-to-video large model in 2024, and NVIDIA, the "shovel seller" in the AI field, delivering unparalleled performance for four consecutive quarters. This may signify that human society will gradually enter the AI era starting in 2024. As for Taiwan Semiconductor, ASML, and Applied Materials, the "creators" of the most crucial underlying hardware for human technological development - chips, after experiencing the "golden age" of the PC era and the smartphone era, starting from 2024, they may usher in a new "golden age" in the global AI wave. The latest data released by the Semiconductor Industry Association (SIA) shows that global semiconductor industry sales in January reached a total of $476 billion, a 15.2% increase compared to January 2023. In fact, semiconductor industry sales have been significantly rebounding since the end of 2023 due to strong demand for AI chips. SIA's statistics show that sales in the fourth quarter of 2023 were approximately $1.46 trillion, an 11.6% increase from the fourth quarter of 2022, and an 8.4% increase from the third quarter of 2023. Regarding the 2024 semiconductor industry sales expectations, SIA President and CEO John Neuffer stated in the data report that overall sales in 2024 are expected to achieve a double-digit increase compared to 2023.

SEMI, in its latest annual silicon shipment forecast report, pointed out that with the recovery of silicon wafers and the entire semiconductor industry demand, inventory levels normalization, and a significant increase in silicon wafer demand to support areas such as artificial intelligence (AI), high-performance computing (HPC), 5G, automotive, and industrial applications, global silicon wafer shipments are expected to rebound comprehensively in 2024, with even stronger growth expected in 2025. SEMI's silicon wafer data includes polished silicon wafers and epitaxial silicon wafers shipped by silicon wafer manufacturers to end-users, excluding unpolished or reclaimed wafers. These silicon wafers will be sent to chip manufacturers such as Taiwan Semiconductor for further processing, where they will be transformed into integrated circuits (ICs) through an extremely complex manufacturing process on the silicon wafers.

TECHCET recently released a significant forecast report, predicting that the continued strong demand for AI chips in 2024 will lead the global semiconductor market into a new upward cycle. TECHCET expects a substantial 12% increase in semiconductor industry sales in 2024, followed by a more intense growth in 2025 driven by AI chips and broader chip demand, reaching 21%. TECHCET anticipates that the global semiconductor market size is likely to surpass the trillion-dollar mark in 2031 or 2032.

Taiwan Semiconductor, with Top-notch Chip Manufacturing Technology and "Chiplet" Packaging Expertise

Taiwan Semiconductor (TSM.US) has been dominating the chip manufacturing field for decades with its accumulated expertise in chip manufacturing technology, and has long been at the forefront of global chip manufacturing technology improvement and innovation (ushering in the FinFET era, leading the 2nm GAA era). With its advanced processes and ultra-high yield, Taiwan Semiconductor has long held the majority of global chip foundry orders, especially for advanced processes of 5nm and below.

The most in-demand AI training/inference high-performance AI chips, such as NVIDIA H100 and AMD MI300 series, are widely used in servers in major data centers worldwide. Taiwan Semiconductor, single-handedly, holds NVIDIA and AMD by the neck.

Both NVIDIA (NVDA.US) and AMD (AMD.US), the two dominant forces in the AI chip field, predominantly utilize Taiwan Semiconductor's 5nm process. The upcoming AI chips are expected to adopt the advanced packaging of 3nm chiplets paired with a 4nm process from Taiwan Semiconductor. Therefore, with unparalleled 3nm, 4nm, and 5nm advanced processes, Taiwan Semiconductor's impact on the global AI chip supply chain is significant. Executives at Taiwan Semiconductor anticipate a 50% compound annual growth rate in revenue for AI-related chips in the coming years, highlighting its importance and revenue growth potential in the AI field.

Currently, Taiwan Semiconductor is the sole chip foundry for NVIDIA's highly demanded AI chips - A100/H100, while NVIDIA is the absolute leader in the global AI chip market, holding a market share of 80%-90%. Citigroup on Wall Street predicts that the AI chip market size will reach around $750 billion in 2024, and expects AMD's latest flagship AI chip, MI300X, to capture approximately 10% of the market share, with Taiwan Semiconductor also being the exclusive chip foundry for AMD. These factors underscore the unparalleled importance of Taiwan Semiconductor in providing foundry services to top-tier chip companies.

The current supply of AI chips falls far short of demand. During NVIDIA's earnings conference in February, CEO Jensen Huang stated that for the remainder of the year, NVIDIA's latest products will continue to be in short supply. Huang emphasized that despite the growing supply, there are no signs of demand slowing down in any way. At the "Advancing AI" conference, NVIDIA's strongest competitor AMD unexpectedly raised its global AI chip market size forecast for the period up to 2027 from the previous $150 billion to $400 billion. The expected AI market size for 2023 is only around $30 billion.

It is reported that in the AMD MI300 series, the Instinct MI300A is a data center APU that combines GPU cores, Zen 4 CPU cores, 128GB HBM3, and a total of 146 billion transistors, utilizing Taiwan Semiconductor's 3D chiplet advanced packaging integrating 5nm and 6nm processes.

The AMD Instinct MI300X GPU is based on the new AMD CDNA 3 architecture, with up to 8 XCD cores, 304 CU units, 8 HBM3 cores, over 150 billion transistors, and integrates 5nm and some 6nm process chiplets. In terms of packaging, the MI300X uses Taiwan Semiconductor's new 3.5D chiplet advanced packaging with a 2.5D silicon interposer and 3D hybrid bonding. The maximum memory capacity of MI300X can reach 192GB.

In addition, there are reports that NVIDIA will announce a new AI chip, the B100, at the GTC conference on March 18. The B100 will also use advanced packaging from Taiwan Semiconductor and may integrate CoWoS and 3D SoIC packaging into a chiplet advanced packaging. As the successor to the Hopper series, the B100 is highly anticipated and considered the most powerful hardware for large language model (LLM) training and inference. Vivek Arya, an analyst at a U.S. bank bullish on NVIDIA, expects the pricing of the upcoming B100 to be at least 10% to 30% higher than the H100.

Looking at the products of the two AI chip leaders, NVIDIA and AMD, it is evident that the ultimate production capacity of their designed AI chips relies entirely on Taiwan Semiconductor's "Chiplet advanced packaging capacity."

In the Post-Moore Era, breakthroughs in advanced chip processes face great challenges (such as quantum tunneling effects and exponentially increasing development costs). Coupled with the gradual entry into the AI era and the increasingly obvious trend of the Internet of Things, the demand for computing power from various tasks may surge, such as deep learning tasks, training/inference, AI-driven image rendering, recognition, etc. These tasks have very high hardware performance requirements, indicating that standalone CPU or GPU integration like PCs can no longer meet the computing power demands. Therefore, the emergence of advanced Chiplet packaging technology allows different "chip processing units" to be integrated together, meeting diverse computing needs and optimizing performance. Chiplet packaging technology enables different GPU modules, or CPU, FPGA, ASIC chips to work together in the same system, efficiently scheduling the computing power of various types of chips to provide greater parallel computing capabilities.

Currently, Taiwan Semiconductor dominates the market for high-end chip packaging orders of 5nm and below with its leading 2.5D/3D advanced packaging technology. The advanced packaging capacity is far from meeting the demand, as evidenced by the shortage of NVIDIA H100. It is reported that due to the overwhelming orders for NVIDIA H100 and B100, as well as the high demand for AMD orders, the advanced packaging capacity based on Taiwan Semiconductor's 3nm, 4nm, and 5nm processes is fully utilized and may need to borrow capacity from packaging and testing factories like Amkor.

IDC predicts that by the second half of 2024, Taiwan Semiconductor's CoWoS capacity is expected to increase by about 130%. Another research firm, Markets And Markets, latest research shows that the total market size of Chiplet advanced packaging products covering GPUs, CPUs, FPGAs, and advanced packaging technologies (2.5D/3D, SiP, WLCSP, FCBGA, and Fan-Out, etc.) is expected to reach around $148 billion by 2028, with an astonishingly high compound annual growth rate (CAGR) of 86.7%. According to their estimates, the total Chiplet market size in 2023 may only be $6.5 billion. At the opening ceremony of Taiwan Semiconductor's first chip manufacturing plant in Japan, known as the "Father of Taiwan Semiconductor" Zhang Zhongmou recently stated that, based on his discussions with top executives of global chip companies, there is an immense global demand for chip production capacity, including AI chips. Zhang Zhongmou said, "These top chip companies are almost all discussing the need for us to establish more chip manufacturing plants globally, not just tens of thousands or hundreds of thousands of wafers, but three, five, or even ten large chip factories."

Regarding the stock price expectations of Taiwan Semiconductor, Wall Street's Morgan Stanley recently raised Taiwan Semiconductor's target price on the Taiwan Stock Exchange from 770 New Taiwan Dollars to 850 New Taiwan Dollars (the latest closing price being 784 New Taiwan Dollars). Morgan Stanley expects that in the next three to four years, Taiwan Semiconductor will maintain over 90% market share in AI-related chip manufacturing. Morgan Stanley emphasized that Taiwan Semiconductor possesses leading-edge advanced packaging technology, top-notch process technology, and the semiconductor industry's most extensive IP and design service ecosystem support. Taiwan Semiconductor's moat in the AI chip field seems wider than in previous product cycles.

13F disclosure data shows that Goldman Sachs, Nomura, Tiger Global, and other Wall Street investment institutions have significantly increased their holdings of Taiwan Semiconductor's American Depositary Receipts (ADRs) in the fourth quarter. According to TipRanks' compilation of Wall Street analysts' target prices, the highest target price for Taiwan Semiconductor ADR is as high as $185 (the latest closing price being $142.14); renowned investment firm Susquehanna recently raised Taiwan Semiconductor ADR's 12-month target price from $130 to $160, maintaining an "outperform" rating; HSBC Research recently raised Taiwan Semiconductor ADR's target price from $119 to $164, maintaining a "buy" rating.

Semiconductor Equipment Giants Holding the "Chipmaking Lifeline"

With Taiwan Semiconductor's large chip manufacturing plants in the United States, Japan, and Germany expected to be completed successively this year, and another chip manufacturing giant Samsung Electronics' U.S. factory expected to be completed in 2024 and start producing advanced process chips below 5nm in 2025, Taiwan Semiconductor and Samsung may begin to procure a large number of high-end semiconductor equipment such as lithography machines, etching equipment, and thin film deposition required for chip manufacturing from 2024 onwards. Moreover, as mentioned by the founder of Taiwan Semiconductor, Zhang Zhongmou, the demand for AI chips is strong. Chip manufacturers such as Taiwan Semiconductor, Samsung, and Intel will comprehensively expand their production capacity. In addition, storage giants like SK Hynix and Micron are expanding HBM capacity, all of which require large-scale procurement of semiconductor equipment for chip manufacturing and advanced packaging. Some core equipment even needs to be upgraded. After all, AI chips have higher logic density, more complex circuit designs, and higher power and precision requirements for devices. This may lead to higher technical requirements in processes such as lithography, etching, thin film deposition, multi-layer interconnects, and thermal management, necessitating customized manufacturing and testing equipment to meet these demands. Therefore, semiconductor equipment giants such as ASML and Applied Materials can be said to hold the "lifeline of chip manufacturing."

In the view of some analysts, only a small portion of Taiwan Semiconductor's smartphone and PC chip production capacity may be transferred to data center AI chip production lines. After all, with smartphones and PCs entering the "AI consumer electronics era" in 2024, AMD, Broadcom, and Qualcomm's demand for AI smartphones and AI PC chips will occupy Taiwan Semiconductor's production capacity.

For a long time, nearly half of Taiwan Semiconductor's mobile chip production capacity has been dominated by its largest customer, Apple, and there are signs of a recovery in smartphone demand. According to the latest statistics from TechInsights, global smartphone shipments rebounded by 7.1% YoY in Q4 2023, reaching 317 million units, ending nine consecutive quarters of decline. Apple leads the global smartphone market with a 23% share. After a 4% decline in shipments in 2023, Canalys predicts a moderate recovery in the global smartphone market in 2024, with shipments expected to grow by 4% to reach 1.17 billion units.

SEMI emphasizes in its forecast report that starting in 2024, the majority of silicon wafer demand will be driven by AI and HPC needs. Research firm Gartner also predicts that the development of generative AI and LLM will fully drive the deployment of high-performance servers based on AI chips in data centers. This also means that chip manufacturers like Taiwan Semiconductor need to expand their production capacity to manufacture more AI chips and need to update or procure semiconductor manufacturing equipment in large quantities. For example, upgrading to manufacturing equipment with higher precision or more complex technical requirements, as well as procuring a large number of manufacturing and testing equipment required for chiplet advanced packaging, 2.5D and 3D advanced packaging systems are much more complex than traditional packaging systems, especially in terms of alignment and assembly. Therefore, the expected expansion of silicon wafers in 2024 implies that the anticipated demand for semiconductor equipment required to create AI chips will also expand. ASML from the Netherlands is the world's largest manufacturer of lithography systems, and its lithography equipment plays a crucial role in the chip manufacturing process. ASML is the sole supplier of advanced Extreme Ultraviolet (EUV) lithography machines used by Taiwan Semiconductor, Samsung, and Intel to manufacture high-end chips.

If chips are the "pearl in the palm" of modern human industry, then lithography machines are the essential tools needed to produce this "pearl." ASML is the global sole supplier of EUV lithography equipment for manufacturing the most advanced chips, such as 3nm, 5nm, and 7nm chips.

For Taiwan Semiconductor's research on 2nm and below node technology, ASML's high-NA EUV lithography machine is crucial. Compared to the standard EUV lithography machines currently produced by ASML, the main difference lies in the use of a larger numerical aperture. High-NA EUV technology uses a 0.55 NA lens, achieving an 8nm-level resolution, while standard EUV technology uses a 0.33 NA lens. Therefore, this new NA technology can print smaller feature sizes on chips, which is vital for the development of process technology for 2nm and below chips.

In chip factories, Applied Materials' presence is ubiquitous. Unlike ASML, which has always focused on the lithography field, Applied Materials, headquartered in the United States, provides high-end equipment that plays a crucial role in almost every step of chip manufacturing. Its products cover key chip-making processes such as Atomic Layer Deposition (ALD), Chemical Vapor Deposition (CVD), Physical Vapor Deposition (PVD), Rapid Thermal Processing (RTP), Chemical Mechanical Polishing (CMP), wafer etching, ion implantation, and more. Applied Materials has high-precision manufacturing equipment and customized solutions for chiplet advanced packaging steps like Wafer Hybrid Bonding and Through Silicon Via, which are crucial for 2.5D and 3D advanced packaging.

Equipment from ASM International in the Netherlands leads the world in the ALD field. In the process of manufacturing AI chips, ALD plays a crucial role, especially for producing AI chips with high integration and more complex structures. Additionally, ALD is essential for transitioning to Gate-All-Around (GAA) technology and the demand for precise threshold voltage tuning.

In the crucial Hybrid Bonding field for chiplet advanced packaging, BE Semiconductor, a semiconductor equipment company from the Netherlands, has its unique "Hybrid Bonding" advanced packaging technology. This innovative high-end bonding technology connects different "chiplets" to enhance their performance, transitioning from the adoption stage to capacity expansion. Goldman Sachs predicts that by 2027, BE Semiconductor's revenue from "hybrid bonding" advanced packaging will reach at least €500 million, compared to only about €50 million in 2022. They also mentioned that the company has received a large order for capacity expansion from chip manufacturer Taiwan Semiconductor.

Other semiconductor equipment companies focusing on multiple steps in the chip manufacturing process include the American company Lam Research and Tokyo Electron from Japan. Compared to Applied Materials and Lam Research, Tokyo Electron has a very high market share in the Coater/Developer field. Tokyo Electron is the strongest competitor to Applied Materials in areas such as ALD, CVD, PVD, RTP, CMP, etching, and ion implantation equipment.

Wall Street giants have long been investing in these semiconductor equipment companies, with the presence of industry giants like Applied Materials visible in the 13F holdings of top institutions such as Goldman Sachs, Morgan Stanley, and JPMorgan. In terms of stock price expectations, semiconductor equipment leader Applied Materials is highly regarded on Wall Street. Despite its stock price hitting record highs since 2024, Sanford C. Bernstein recently raised Applied Materials' 12-month target price from $230 to $240 (latest closing price at $200.56), while UBS Group significantly increased the target price from $185 to $235.

Tokyo Electron, a competitor of Applied Materials, is also favored by Wall Street. Jefferies recently raised Tokyo Electron's 12-month target price from ¥30,000 to ¥42,000 (latest closing price at ¥37,220). Tokyo Electron's stock price has repeatedly hit new highs this year, currently hovering near its all-time high. Jefferies also has a positive outlook on ASM International, a leader in the ALD field benefiting from the AI trend, giving it a target price of €740 (latest closing price at €566).

BE Semiconductor, focusing on the "hybrid bonding" field, is one of the biggest beneficiaries in the European stock market amid the AI investment boom. Its stock price surged by 140% in 2023 and has repeatedly hit new highs this year. Goldman Sachs predicts that BE Semiconductor still has room for growth in 2024, mainly due to the strong demand for the company's advanced packaging equipment in the market. The Goldman Sachs analysis team expects the stock to rise to €170 in the next 12 months (latest closing price at €141). Based on the Wall Street analysts' target price compiled by TipRanks, BE Semiconductor has a bullish target price of up to 200 euros.

China Strives to Solve the "Bottleneck" Issue of Semiconductor Equipment

In recent years, the United States has continuously escalated sanctions on the Chinese chip industry, focusing on semiconductor equipment, raw materials, and chip manufacturing. Therefore, to achieve comprehensive localization in the chip manufacturing sector, the development of various high-end semiconductor equipment needed for chip manufacturing is essentially starting from scratch, becoming the core focus of government and private funding at all levels.

Analysts from Barclays and Sanford C. Bernstein believe that companies like Naura Technology Group (North Huachuang) in the semiconductor equipment sector may one day become globally renowned, just like Applied Materials and Lam Research in the United States. In particular, analysts from Bernstein directly compare North Huachuang to the semiconductor equipment giant Applied Materials.

The reason for these Wall Street analysts is simple: the U.S. efforts to restrict China's access to cutting-edge semiconductor equipment technology will promote the vigorous development of the domestic semiconductor equipment sector and create huge opportunities for local companies, with investments in these companies likely to pay off in the coming years.

In response to the U.S. Department of Commerce's comprehensive ban on Applied Materials and ASML, among others, from entering China and the prohibition of NVIDIA's high-end AI chips from entering the Chinese market, China has invested over $100 billion in substantial funds to create domestic suppliers that can fill the gap. Bernstein analysts recently wrote in a research report, "The U.S. sanctions are a double-edged sword. While they may suppress China's development in cutting-edge semiconductor equipment, they will also force China to rapidly develop its own supply chain and strive for self-sufficiency."

Barclays analysts, on the other hand, point out that China's plan to achieve a 70% self-sufficiency rate in domestic chips by 2025 faces challenges, as reducing dependence on imports will take time. However, with hundreds of billions of dollars in government funding pouring into local chip companies, the roles played by semiconductor equipment suppliers and chip manufacturers in China's chip industry chain will become increasingly important, with domestic chip production expected to double in the next 5 to 7 years.