Your browser does not support JavaScript!

Current Landscape and Competitive Dynamics in AI Chip Manufacturing

GOOVER DAILY REPORT June 25, 2024
goover

TABLE OF CONTENTS

  1. Summary
  2. Samsung's Strategic Advances in AI Chip Technology
  3. Nvidia's Product Strategy and Market Position
  4. TSMC’s Market Leadership and Technological Advancements
  5. Intel's Comeback Strategy in AI and Semiconductor Market
  6. US-China Trade Sanctions and Their Impact on the Semiconductor Industry
  7. Competitive Dynamics in the Semiconductor Market
  8. Conclusion

1. Summary

  • The report, titled 'Current Landscape and Competitive Dynamics in AI Chip Manufacturing,' delves into the recent advancements, market strategies, and competitive dynamics among major semiconductor companies like Samsung, TSMC, Nvidia, and Intel. It provides a comprehensive overview of the technologies, strategic partnerships, and market positioning shaping the AI chip manufacturing industry. Key points include Samsung's innovations in power delivery and process development, Nvidia's new AI accelerators and market expansion, TSMC's leadership in 3-nanometer technology and advanced packaging, and Intel's introduction of new AI chips. The report also examines the impact of US-China trade sanctions on the semiconductor market and how these geopolitical factors are influencing global supply chains and company strategies.

2. Samsung's Strategic Advances in AI Chip Technology

  • 2-1. Backside Power Delivery Network Technology

  • Samsung has introduced backside power delivery network technology in its chip production process, which places power rails on the backside of a silicon wafer. This advancement enhances power, performance, and area efficiency, while significantly reducing the drop in voltage compared to its first-generation 2-nanometer (2nm) process. The company claims that this technology will be integral in winning orders for AI processors, which are among the highest-performing and most expensive chips in use.

  • 2-2. 3-Nanometer and 2-Nanometer Process Development

  • Samsung has made significant strides in its 3-nanometer (3nm) and 2-nanometer (2nm) process development. The company plans to mass-produce its second-generation 3nm process in the second half of this year, and continue enhancements with gate-all-around (GAA) technology. GAA is essential for AI products and will be integrated into Samsung's upcoming 2nm process. Furthermore, the company is preparing for 1.4nm production, with targets for performance and yield on track for mass production by 2027. In 2022, Samsung became the first in the industry to begin GAA-based 3nm mass production.

  • 2-3. AI Customer Base Expansion

  • Samsung aims to expand its AI customer base significantly. At its annual foundry forum, the company predicted that its AI-related customer list would expand fivefold, and revenue would grow nine times over current levels by 2028. Samsung's strategy includes offering logic, memory, and advanced packaging solutions, which it believes will expedite its progress in securing outsourced semiconductor manufacturing orders for AI-related chips. The company has unveiled several new types of production technology and a layout for future AI-related chips to attract more customers. However, Samsung has faced challenges in proving the reliability of its advanced memory chip production to demanding customers like Nvidia, and Intel has also emerged as a new competitor in the AI chip market.

3. Nvidia's Product Strategy and Market Position

  • 3-1. New AI Accelerators: Blackwell Ultra and Rubin Platform

  • Nvidia has announced plans to upgrade its AI accelerators yearly, introducing the Blackwell Ultra chip for 2025 and the Rubin platform for 2026. These new products are part of Nvidia's strategy to stay ahead in AI technology, especially as data processing demands continue to grow. Nvidia CEO Jensen Huang emphasized that traditional computing methods cannot keep up, and Nvidia's accelerated computing provides significant cost and energy savings. Nvidia's commitment to innovation is reflected in its continuous development of new platforms and tools, which attracted positive market responses, including stock gains for its suppliers like TSMC and Wistron Corp. The Rubin platform, in particular, will feature HBM4 memory from SK Hynix, addressing a crucial bottleneck in AI accelerator production.

  • 3-2. Expansion of Customer Base Beyond Cloud-Computing Giants

  • Nvidia, traditionally dominant in the gaming and cloud computing sectors, is broadening its customer base to include a wider array of industries. CEO Jensen Huang envisions a future where diverse sectors, from shipbuilding to drug development, embrace AI technologies. Nvidia's push to diversify its customer base comes as it aims to leverage the exponential growth in data processing needs, which traditional methods can't handle efficiently. The company's emphasis on AI tools and pretrained models underscores its intention to cater to new markets and applications, ensuring that businesses without AI capabilities do not fall behind.

  • 3-3. Collaboration with Microsoft for AI-Integrated Laptops

  • Nvidia is collaborating with Microsoft and their hardware partners to integrate AI functionalities into laptops, branded under Copilot+. These new laptops are designed to offer enhanced performance and energy efficiency, bolstered by Nvidia's graphics cards. The integration aims to bring advanced AI functionalities to personal computing devices, significantly boosting their capabilities. Nvidia's tools and pretrained AI models are set to help software developers optimize PC performance, managing tasks on the machine itself or offloading them to data centers when needed. This collaboration highlights Nvidia's strategy to extend its influence from data centers to consumer-grade devices.

4. TSMC’s Market Leadership and Technological Advancements

  • 4-1. 3-nanometer Technology and Advanced Packaging

  • TSMC has solidified its leadership in the semiconductor sector with the successful implementation of its 3-nanometer (3nm) technology. The company's 3nm family includes N3, N3E, N3P, N3X, and N3A, targeting a broad range of applications from AI accelerators and high-end smartphones to data centers and automotive clients. N3E began mass production in the fourth quarter of last year, and N3P is scheduled for mass production in the second half of this year. TSMC's advanced packaging technologies, such as Chip-on-Wafer-on-Substrate (CoWoS) and Integrated Fan-Out (InFO), have also contributed significantly to its competitive edge. These technologies are crucial for high-performance computing (HPC) and AI applications. CoWoS, in particular, has seen a surge in demand, with TSMC’s Zhunan advanced packaging plant (AP6) becoming Taiwan’s largest CoWoS base.

  • 4-2. Price Adjustments and Capacity Expansion

  • TSMC is planning significant price hikes due to overwhelming demand for its 3nm technology from major clients such as Apple and NVIDIA. The prices for 3nm and advanced packaging processes are expected to increase by approximately 10% to 20% next year. This move is supported by key partners like NVIDIA, which has agreed to price increases to secure more advanced packaging capacity. In response to escalating demand, TSMC is also doubling its CoWoS production capacity from 17,000 to 33,000 wafers per month by the third quarter. Additionally, TSMC's advanced packaging capacity remains scarce, with NVIDIA and AMD being principal customers. The company is adding more CoWoS-related equipment and requesting dispatch of more engineers to fully staff its Longtan AP3, Zhunan AP6, and Central Taiwan Science Park AP5 plants.

  • 4-3. Strategic Partnerships with Major Tech Firms

  • TSMC's market leadership is bolstered further by its strategic partnerships with major tech firms. Apple is one of TSMC's most significant partners, accounting for about 25% of TSMC’s revenue in 2023. The company also produces high-performance chips for AMD and Nvidia, which make up 7% and 11% of TSMC's revenue respectively for the same year. These partnerships not only provide substantial revenue but also affirm TSMC's ability to meet the high standards of leading technology companies. NVIDIA has highlighted TSMC’s role not only in manufacturing but also in handling numerous supply chain issues. Furthermore, other companies like Broadcom, Amazon, and Marvell have shown strong interest in TSMC’s advanced packaging processes, further securing TSMC’s market dominance.

5. Intel's Comeback Strategy in AI and Semiconductor Market

  • 5-1. Introduction of Falcon Shores Chip

  • Intel is set to introduce the Falcon Shores chip, combining Gaudi's technology with its own Xe GPU cores. Scheduled for release in late 2025 or 2026, this new chip aims to compete directly with Nvidia's CUDA by utilizing Intel's OneAPI software platform. Expected to consume 1500 watts of power, potentially higher than Nvidia's B200, it will offer a new competitive edge in the AI GPU accelerator market. Remarkably, Falcon Shores production will move back to Intel's fabs, promising higher margins.

  • 5-2. OneAPI Software Platform

  • To break Nvidia's stronghold with its CUDA software, Intel has focused on expanding its OneAPI software platform, introduced in 2018. This platform aims to program virtually any kind of AI processor, encouraging tech companies to choose hardware based on their needs rather than being locked into Nvidia's ecosystem. Collaborating with seven other technology giants, Intel aspires to make this open-source platform a compelling alternative for developers who seek flexibility and cost efficiency.

  • 5-3. High-NA EUV Machine Investments

  • A critical aspect of Intel's strategy involves significant investment in high-NA EUV machines. In 2024, Intel purchased all five high-NA EUV machines produced by ASML, ahead of its competitors such as TSMC and Samsung. CEO Pat Gelsinger's ambitious plan to progress through 'five nodes in four years' remains on track, anticipating the production of 1.8nm transistors by 2025. This strategic investment aims to establish Intel at the forefront of semiconductor manufacturing technology, potentially reshaping the competitive landscape.

6. US-China Trade Sanctions and Their Impact on the Semiconductor Industry

  • 6-1. China's Efforts Towards Self-Sufficiency

  • In response to US trade sanctions, China has been intensifying its efforts to achieve self-sufficiency in semiconductor manufacturing. The country has increased investments in domestic semiconductor companies and technology development to reduce reliance on foreign technology. These efforts include funding for research and development, subsidies for local firms, and the establishment of national semiconductor projects.

  • 6-2. Impact of US Trade Sanctions

  • US trade sanctions have significantly affected Chinese semiconductor companies, particularly those reliant on US technology and components. The sanctions have led to supply chain disruptions and difficulties in accessing advanced technologies. Companies such as Huawei have faced significant challenges due to restrictions on purchasing semiconductor components from US-based manufacturers. This has compelled Chinese companies to seek alternative suppliers and explore in-house development of critical technologies.

  • 6-3. Global Supply Chain Disruptions

  • The US-China trade sanctions have caused considerable disruptions in the global semiconductor supply chain. Many companies worldwide are experiencing delays and shortages of critical components due to the geopolitical tensions. For example, firms dependent on Chinese manufacturing or US technology have had to navigate complex trade restrictions and regulatory hurdles. These disruptions have further accelerated the trend of diversifying supply chains and seeking resilient and reliable sources of semiconductor components.

7. Conclusion

  • The report underscores the significant progress and competitive strategies employed by leading companies like Samsung, TSMC, Nvidia, and Intel in the AI chip market. Samsung's advancements in backside power delivery and process nodes position it as a rising competitor to TSMC. Nvidia maintains its dominance with continuous innovations and market expansion efforts. Intel is staging a comeback with its Falcon Shores chip and OneAPI platform. TSMC remains the market leader, driven by its technological advancements and strong partnerships. However, US-China trade sanctions pose challenges, influencing companies to diversify supply chains and seek self-sufficiency. To navigate this dynamic landscape, stakeholders must stay informed about technological trends and geopolitical developments that impact the semiconductor industry. Future prospects indicate sustained growth and intense competition, with a focus on innovation and strategic partnerships driving the next wave of advancements in AI chip manufacturing. Practical applications of these advancements are expected to permeate various industries, enhancing capabilities in areas like AI, high-performance computing, and consumer electronics.

8. Competitive Dynamics in the Semiconductor Market

  • 8-1. Market Share and Revenue Comparisons

  • TSMC is the world's largest dedicated independent semiconductor foundry with over 60% of the market share as of Q4 2023. In contrast, Samsung's foundry holds a 7.5% market share for the same period. In terms of revenue, TSMC reported $57.39 billion in 2021, $73.86 billion in 2022, and $70.59 billion in 2023. Conversely, Samsung Electronics, including its diverse business units, recorded $240.71 billion in 2021, $233.13 billion in 2022, and $194 billion in 2023. Both companies display significant contributions from their semiconductor divisions.

  • 8-2. Technological Innovations and Their Market Implications

  • TSMC advanced significantly with its 3nm process technology, offering 10-15% speed improvement and 25-30% power reduction over its 5nm predecessor. Samsung's notable technological leap includes adopting Gate-All-Around (GAA) technology for its 3nm process. This innovation enhances performance and power efficiency. TSMC’s advanced packaging technologies, like Chip-on-Wafer-on-Substrate (CoWoS) and Integrated Fan-Out (InFO), meet high demands in HPC and AI applications. Samsung, on the other hand, continues to innovate in memory technology, remaining a leader in V-NAND and DRAM. These advancements influence applications across AI, data centers, and consumer electronics.

  • 8-3. Investment and Growth Strategies of Leading Firms

  • TSMC announced a $100 billion investment plan spanning over three years to expand globally, particularly in the US and Japan, mitigating geopolitical risks. Similarly, Samsung is scaling up its investments, projecting $150 billion towards system semiconductors by 2030. Both companies' strategies to ramp up production capabilities signify a robust competitive stance in the semiconductor market. The aggressive investment plans facilitate advancements in manufacturing technologies and support the growing demands in various industries, including automotive and IoT.

9. Source Documents