Gather global wisdom and talk about the future of "core". On June 28, 2024, the 8th Jiwei Semiconductor Conference was held at the Xiamen International Conference Center Hotel. As one of the core forums of the conference, the 4th Jiwei Semiconductor Analyst Conference and the 2nd Global Semiconductor Industry Strategy Forum gathered the world's top analysts and industry experts to discuss the changes in global semiconductor development, the latest trends in the semiconductor supply chain, and how artificial intelligence can revolutionize the semiconductor industry.
Grace Wang, General Manager of Hetao International Consulting Co., Ltd., was invited by Aijiwei to serve as the moderator of this analyst conference. In her opening remarks, she said, "As we all know, we are entering the era of artificial intelligence while always looking for the next key growth area in the next 5 to 10 years. According to the latest forecast data from JW insights, the global semiconductor market revenue will reach $8,000 to $1 trillion by 2028, driven by artificial intelligence and automotive applications. Until then, the problem of oversupply of mature processes may be more severe in 2025, and many uncertainties will be changed. We're looking forward to identifying more opportunities for growth, which is why we're all here now. ”
Artificial intelligence is revolutionizing the semiconductor supply chain and consumer markets
With the rapid iteration and application of AI technology, the semiconductor consumer market is undergoing significant changes. Neil Shah, Vice President of Counterpoint Research, said in a keynote speech titled "How AI Will Revolutionize the Semiconductor Consumer Market from Cloud to Edge," Neil Shah, Vice President of Counterpoint Research, said, "As AI technology continues to evolve and chip processes become more advanced, semiconductors have more and more capabilities to drive and train models, while the overall AI ecosystem continues to evolve. At present, AI drives servers to occupy the main source of semiconductor applications, and in the future, smartphones, PCs, and automobiles will become the top three sources of semiconductor consumption for edge device applications. ”
At present, the cloud part of the entire large model ecosystem has achieved good development and is showing a trend of shifting to the edge, which makes edge device computing more and more important. Neil Shah believes that under the huge demand for computing power, the three key categories of computing semiconductors in servers, smartphones and laptops have reached about $447 billion this year, and it is expected to consume $1 trillion in computing semiconductors by 2030. As investment in edge computing continues to increase, this number will grow rapidly. Technological upgrades, such as advanced computing and memory, are key factors driving the development of most applications and services.
At the same time, AI technology is profoundly impacting the development of the semiconductor supply chain. Eric Bouche, Managing Director of the Silicon Valley Research Initiative (SVRI), said in a keynote speech titled "The Age of Artificial Intelligence: How to Improve the Critical Needs for Semiconductors for Equipment and Materials," "Industry 4.0 or the Fourth Industrial Revolution is characterized by the integration of digital technologies into the manufacturing process. For AI semiconductors, this means they must be designed to integrate seamlessly with these digital systems, enabling smarter and more efficient manufacturing processes. Semiconductor manufacturing equipment that leverages machine learning and automation can improve efficiency and shorten lead times, while semiconductor Industry 4.0 is characterized by the convergence of several related cutting-edge technologies that can synergistically build a seamlessly connected manufacturing ecosystem. ”
These include customization and prototyping: AI applications often require custom chip designs. Reducing the time from design to production is critical. Advanced manufacturing technologies: Utilizing advanced technologies such as extreme ultraviolet (EUV) lithography and 3D integration can significantly speed up production. Flexible development process: Flexibility that supports design, manufacturing, and testing processes to quickly adapt to market demands.
By 2028, the adoption of AI-powered scheduling software could reduce lead times by 8%, while implementing agile manufacturing principles such as just-in-time inventory management could reduce lead times by a further 5%. Eric Bouche said, "In the future, the world economy and technology will need more and better AISoC, and the semiconductor manufacturing industry will now need better and cheaper Industry 4.0 materials and equipment. Among them, through the synergistic effect of artificial intelligence, China's material equipment industry can gain a considerable market share. ”
In addition, Kyn Min Cho, Managing Director of IAAN Corporation, in his presentation entitled "How to Revolutionize the Semiconductor Supply Chain with AI Digital Twin Factory", also mentioned the concept of establishing a virtual fab through 3D simulation technology, that is, by establishing an industrial digital twin platform that combines 2D/3D design, AI, XR and ICT technologies, to achieve integrated KPIs throughout the lifecycle of industrial facilities, which can greatly improve efficiency and reduce costs. He also pointed out that the digital twin market is expected to reach $35.8 billion in 2025, with a compound annual growth rate of 37.87%, and the resulting economic value will reach nearly $4-11 trillion.
The growth of the global semiconductor industry may accelerate significantly in the short term
After a low ebb in 2023, the global semiconductor industry is now entering a recovery trend. Malcolm Penn, President and CEO of Future Horizons Ltd, said in his keynote speech entitled "Looking Forward to the Development and Changes of Global Semiconductors in 2024 and Beyond" that as an industry with strong cyclicality, semiconductors will have a "golden cross" and a "hope cross" in cyclical ups and downs, so it is necessary to pay close attention to data, rising and falling trends, and always be vigilant. In the long-term trend line, the annual growth rate of semiconductors has remained around 8% over the past 40 years.
Malcolm Penn noted that although the semiconductor industry is now above the long-term trend line, production capacity has far outpaced industry growth. Even if market demand returns to where it should be, there is currently too much capacity and it will become a big problem for some time to come. "In any case, the rise in average selling prices has laid the foundation for a recovery in the semiconductor industry, and we expect the global semiconductor market to grow by around 5% this year, but the performance in the first quarter was very weak, resulting in a three-fold increase for the rest of the year. In the long run, the semiconductor industry has good growth prospects because there are always many new creations and inventions that will make other products smarter, more beneficial, and more useful, but the factors that drive and change the market include technology, legislation, structure, diversification, and so on. ”
At present, it is extremely difficult for the industry to fully grasp the cyclical fluctuations of semiconductors, and most of the existing indicators such as the Philadelphia Semiconductor Index only focus on the secondary investment market. In this regard, Zhao Yi, deputy general manager of the consulting business of Aijiwei Consulting (Xiamen) Co., Ltd., said in the keynote speech of "Introduction to Jiwei Guolianan Global Semiconductor Prosperity Index" that "Jiwei Guolianan Global Semiconductor Prosperity Index" is committed to building an "investment anchor" and "professional companionship" of semiconductor ETF, and provides cutting-edge insight research and prosperity measurement tools to fill the gap in the industry by building a semiconductor industry prosperity exchange and service platform, so as to help investors better adapt to market trends and industry rhythms.
Zhao Yi said that Jiwei Guolianan Global Semiconductor Prosperity Index is an index that comprehensively reflects a series of semiconductor industry prosperity indicators such as "global + China" and "fundamentals + sentiment", and has four significant characteristics.
First, comprehensiveness. At the same time, the index reflects the fundamentals and secondary market sentiment, and the industrial chain covers a complete range, including the entire industrial chain from design, manufacturing to packaging and testing, from terminal sales to equipment procurement. Second, science. The index reflects both short-term month-on-month and year-on-year changes in the cycle, and more fully demonstrates the seasonal fluctuations of semiconductors. There are not only public data, but also offline research and analyst research. Third, intuitive. The index takes 50 as the line of prosperity and decline, which more intuitively reflects the changes in the prosperity of the industry. Fourth, abundance. In addition to the index, a detailed analysis report is also included; Not only monthly regular reports, but also quarterly detailed reports.
Paying attention to the evolution of the global semiconductor market, the development trend of the downstream consumer market is very important. After a stumbling 6.2% decline in 2023, the global portable communications equipment market will grow by a flat 1.5% this year, but headwinds remain due to uncertainty in North America and Western Europe, Elliot Lin, GfK's Supply Chain & Industry Market Research Analyst, said in his keynote speech on "Global Market Demand and Changing Consumer Behavior Analysis of Portable Communication Devices".
In terms of industry trends, according to the survey, more than 70% of consumers have been using smartphones for more than 2 years, and the extension of the replacement cycle has adversely affected the global demand for smartphones, but the development trend of high-end specialty products continues. Among them, the recovery of Chinese consumers' purchasing power will continue, while competition in the high-end market will continue to intensify. In emerging regions, the market share of 5G smartphones is expanding, and the penetration rate will reach 46% in the fourth quarter of 2024, while configuration upgrades and premiumization trends will drive ASP growth.
Elliot Lin further said, "Artificial intelligence is boosting the smartphone market, but integrating AI technology into smartphones requires a complete upgrade of smartphone hardware. In 2024, AI-powered smartphones will hold 19% of the market share, with global sales reaching 217.1 million units. At the same time, the global average DRAM will reach 7.3GB. Due to the popularity of 12GB/16GB in mid-to-high-end models, the growth rate is expected to expand further this year. In addition, the OLED penetration rate will reach 53% in 2024. The high-end market has driven the growth of OLED penetration over the past few years, and the mid-range market will be the main driver this year. ”
The "evolution" of the semiconductor industry in China, the United States and India
There is no doubt that AI and automobiles are becoming the key drivers of growth in the semiconductor market, and the Chinese market is crucial. By 2030, battery electric vehicles (BEVs) will contain about 1.7 times more semiconductors than conventional gasoline or diesel vehicles, said Kevin Li, director of market research at Techlnsights China, in a keynote speech titled "2024 02 China Market Latest Forecasts: Automotive Semiconductor Demand, Cockpit Integration and Generative AI". From 2023 to 2028, BEVs are expected to be the fastest-growing models in demand for automotive semiconductors (CAGR 23%), followed by plug-in hybrids (10%) and mild hybrids (8.4%).
In addition, generative AI has evolved in-vehicle human-computer interaction from the original era of rule-based mechanical interaction to the era of AI-based intelligent interaction. While OEMs are adopting large-scale models to improve the user experience of in-vehicle interaction through independent R&D or external collaboration, OEMs' large-scale model product capabilities are mainly focused on conversational dialogue, personalized recommendations, content generation, and multi-modal interaction. Kevin Li pointed out that the areas that need to be improved include the smoothness of interactions, the usefulness of functions, the accuracy of personalized services, and the quality of generated content.
As for the opportunities and challenges of China's new energy vehicle market, Kevin Li said, "First of all, in the de-globalization new energy vehicle market, global automakers have launched new energy models for the Chinese market, such as Volkswagen, General Motors, Toyota, etc., while Chinese models face some challenges with additional requirements in the US and European markets. Second, OEMs are becoming Tier 1S. Among them, some domestic automakers have become Tier 1 suppliers of their global partners, providing vehicle platforms, software and hardware integration. For example, Xpeng and Volkswagen, Leapmotor and Stellantis, SAIC and Audi and General Motors, etc. Under the framework of the HarmonyOS Intelligent Mobility Alliance (HIMA), some Chinese automakers are providing OEM services to Huawei. In addition, as for the future growth space, China's auto market is facing challenges such as reducing costs while maintaining competitiveness, balancing fuel vehicles and new energy vehicles, and de-globalization. “
In the field of AI chips, some American companies are leading the industry. Jay Goldberg, CEO of D2D Advisory Consulting, said in a keynote speech titled "Future Expectations for Nvidia GB200, Apple and AIGC" that the rapid development of artificial intelligence is profoundly changing the technology industry, such as software companies are using large AI models and various different neural network-based systems to upgrade their software to reduce customer service or overall costs, as well as improve recommendation engines, search engines and ad matching engines.
At the chip enterprise level, Nvidia has chosen to bet on artificial intelligence in recent years, making unprecedented achievements, while building a solid moat through the CUDA software platform. "Nvidia releases a new GPU every year, and it's hard to maintain that speed, but they've done a great job of not only capturing more than 80% of the market share in AI chips, but also getting a lot of adoption of the GB200 and B-series AI chips built with the new Blackwell architecture, and the demand is outstripping supply," said Jay Goldberg. Despite the challenges from companies such as Google, Nvidia's dominance is expected to continue in the near term. ”
As for Apple, although it is aggressively increasing its AI to provide related products and services, it cannot find a way to charge users for AI, which is also an important problem for the entire industry. In fact, Jay Goldberg said, it took a long time for the industry to find a way to use AI and confirm that it is a very important tool, but there is still a lot of experimentation to understand and validate its functionality to create more innovative, fun and fun products in the future.
As an emerging market, the rapid development of India's semiconductor industry is attracting widespread attention. Arvind Consultancy CEO Sanjeev Keskar gave a presentation entitled "India – From Semiconductor Chip Design to Chip Manufacturing". According to Sanjeev Keskar, the Indian semiconductor market is expected to reach $640 billion by 2026. The electronic systems design and manufacturing (ESDM) market is expected to reach $400 billion by 2028, and the related market opportunity will reach $5 trillion.
"India will maintain a high GDP growth rate (7%) and will become the world's leading labor market (1 billion) by 2030," said Sanjeev Keskar. At the same time, the Indian government has launched a series of incentive and subsidy policies to promote the construction of semiconductor wafer factories in the fields of CMOS, display, compound semiconductors, etc., with the Indian government subsidizing 50% and the state government subsidizing 20% of the funds in the above fields. ”
The semiconductor supply chain is being reshaped in a complex evolution
Under the interweaving of technology, capital, market, inventory and international geopolitical factors, the global semiconductor supply chain is undergoing a new reshaping. Doug Sparks, founder & CEO of M2N Technology LLC, said in a keynote speech entitled "The Latest Status of the Semiconductor Supply Chain in 2024" that in order to improve semiconductor infrastructure, China pledged to withdraw $27 billion last year, more than $52 billion of the U.S. Chips Act began to allocate funds, the European Union is expected to allocate $47 billion, India recently announced an investment of $15 billion, while Japan, South Korea and other countries are increasing investment, And it's all about the semiconductor supply chain.
"Despite increased investment in semiconductors, global supply chains are fragmenting into smaller regional or national supply chains, with increasing trade restrictions, bans, regulations and tariffs being used to force supply chain companies to make new sourcing plans," said Doug Sparks. At least in 2022, about 92% of the global semiconductor supply chain will be effectively controlled by the United States and its political allies, such as the United States, Japan, the Netherlands and other countries, whether it is IC design, wafer factory equipment, materials, front-end, manufacturing, or SAT or back-end assembly, manufacturing, etc. ”
Currently, a combination of nearshoring, friend-shoring, and allied outsourcing is being used to build a new second semiconductor supply chain around the world, while Japan, the U.S., Europe, and Singapore are heavily using high-end fab manufacturing capabilities from Taiwanese and South Korean companies, while countries and regions such as China, the U.S., Europe, Japan, South Korea, India, and Taiwan are investing billions or billions of dollars to enhance their semiconductor infrastructure. In the future, this trend will continue, and further nationalization and regionalization in the supply chain will be sought.
Further, in the field of semiconductor materials, Andy Tuan, president and general manager of ESG Flagship/Linx-consulting, said in a keynote speech entitled "The Latest Trends in the Global Semiconductor Material Supply Chain" that the global semiconductor supply chain inventory index will continue to decline in 2023 to reach a healthy level in Q4, and the wafer production capacity is expected to continue to rise in the next two or three years. In order to cope with the growing demand, it has become an inevitable trend to create high-performance system chips, which also means that more new materials will be used in chips.
But the average utilization rate of wafer manufacturers is likely to be lower globally compared to past years, and if they do put all of their announced capacity into production, many fabs are either idle or underutilized. Therefore, from the demand side, there will be some adjustment in the global materials market. Andy Tuan said that at the moment, the production of semiconductor materials is still low, while more industry consolidation is taking place, which requires the supply chain to balance its technical capability infrastructure in terms of new investments, and there is a "winner-takes-all" opportunity for leading players.
Andy Tuan further said, "The challenge of the global semiconductor supply chain is that it is becoming less and less efficient, and the cost is getting higher and higher, as the world tries to expand its capacity for self-sufficiency, which leads to overcapacity, and at the same time, environmental control regulations will become more stringent and some of the current international geopolitical risks will also bring certain constraints to the materials industry. However, since the beginning of this year, there have indeed been some opportunities for recovery in the semiconductor materials industry, and high-performance computers and artificial intelligence will drive some new growth demand and related investment and technology iteration. In the future, this growth trend is expected to continue through strong collaboration across the supply chain. ”
In addition, Mike Walden, senior head of research at TECHCET, said in a speech entitled "The Key to Supporting Technology Growth: Semiconductor Materials and Markets" that the economic trend in 2024 will achieve a soft landing and slow recovery, and the global semiconductor market outlook is optimistic, which will also kick off the prelude to huge investment in semiconductors, with global capital expenditures totaling $532 billion from 2023 to 2028, of which semiconductor investment from the United States, China and South Korea will account for 80%.
Mike Walden said, "In terms of semiconductor materials, sustainable growth will be achieved after the market recovery, and the expansion of chip fabs is a key factor driving the materials market forward, and the semiconductor materials market is expected to exceed $72 billion in 2024 and reach $87 billion in 20208, with a compound growth rate of 5% for the overall market and more than 10% for the advanced materials market from 2023 to 2028." ”
Chiplets will become a key technology to lead the industry's growth
In the current global semiconductor recovery, the development of memory chips is particularly important. Marco Mezger, Founder & CEO of Neumonda/APIS4, gave a presentation entitled "Memory Market Dynamics: The Future of DRAM Testing". According to Marco Mezger, the global semiconductor market will double in size over the next decade.
However, as an important part of the semiconductor industry, the investment in related equipment has decreased due to the decrease in investment by memory manufacturers. Although the emergence of generative AI has increased the demand for storage to a certain extent, it is more reflected in the HBM field, which is expected to decrease by 17% year-on-year in 2024. When it comes to DRAM test equipment, Marco Mezger believes that the challenges of weight, cost, ease of operation, power consumption, and flexibility need to be addressed.
Obviously, semiconductor equipment is an indispensable "weapon" for the development of the entire industry. Marco Mezger, Co-Founder & Chief Business Development Officer of D-SIMLAB Technologies, gave a presentation entitled "Capacity Planning and Logistics Optimization Keys in Semiconductor Front-End Manufacturing", sharing his views on optimizing semiconductor manufacturing processes and improving efficiency from the aspects of semiconductor manufacturing and digital twins, improving predictability in the production process, and AI empowerment.
"Due to the increasing challenges of today's semiconductor wafer manufacturing due to cost, complexity, cycle time, etc., it is necessary to simulate the operation of the fab, including the flow of materials, and the need for a virtual fab based on a digital twin," said Marco Mezger. With the development of AI and the effective combination of digital twins, a large amount of data in the wafer manufacturing process will be very valuable for the training of large AI models, which can significantly shorten the manufacturing cycle (more than 30%), reduce costs, and increase revenue (10%). ”
At the same time, chiplets will become one of the key technology paths leading the future growth of semiconductors. Jan Vardaman, president of TechSearch, said in a keynote speech titled "The Role of Chiplets in Advanced Packaging" that the need for performance means more transistors, bigger chips, higher silicon costs, but the cost of manufacturing a large mold becomes uneconomical. To address the challenges of power, performance, area, and cost, chiplets are an important solution. Unlike SiP or traditional MCM, chiplet is a new design concept that shifts from "silicon-centric thinking" to "system-level planning" and "IC and package co-design".
Chiplet features include: open-source chip design; Customizable, package-level integrated chip-to-chip interconnects and protocol connectivity from an interoperable, multi-vendor ecosystem; hierarchical protocols; Support for different data ranges, widths, convex pitches, and channel ranges to ensure the widest possible interoperability; It can be used with UCIe's proprietary interface by merging the protocol translation layer between the proprietary interface and the UCIe interface. At present, AMD, Intel and other companies are vigorously promoting the adoption of chiplet technology in advanced packaging, such as AMD's Instinct MI300 for AI, which uses TSMC's SoIC process and integrates 5nm GPU and CPU logic stacking in a 3D hybrid bonding structure.
According to Jan Vardaman, "The era of chiplets has arrived, and it will be critical to the co-design of IC design and packaging. Among them, advanced packages for high-performance computing with heterogeneous integration will include chiplets, laminated (or multilayered) substrates, silicon interposers (2.5D), sector-output RDLs with bridging options, and embedded bridging in laminates. But no single solution is the answer, and all options need to be grounded. In addition, there will be a need for continued work in thermal, test and metrology, as well as the adoption of third-party chiplet standards. “
The conference ended successfully with on-site questions and enthusiastic exchanges among the guests, and the 4th Jiwei Semiconductor Analyst Conference showed a wonderful international exchange event that transcended national borders, so that we look forward to meeting again next year.