Source: Observation of the Semiconductor Industry

Image source: Generated by Wujie AI

Today, we are witnessing an unprecedented paradigm shift in the global market. Following the attention brought by OpenAI's ChatGPT to consumers and investors, enterprises in various industries are competing to integrate artificial intelligence capabilities. Among the giants in the US stock market with a market value exceeding $1 trillion, Apple ranks first with a market value of $3.08 trillion, followed closely by Microsoft ($2.51 trillion), Google's parent company Alphabet ($1.67 trillion), Amazon ($1.35 trillion), and Nvidia ($1.15 trillion). Apart from Apple, which relies on consumer devices such as the iPhone, the other four tech giants are all vigorously promoting integration with the field of AI.

For example, Microsoft recently announced that its enterprise AI software tools will start charging monthly service fees to enterprise users, undoubtedly sounding the clarion call for large enterprises to successfully monetize AI business opportunities with clients. In addition, Alphabet is also integrating multiple products to introduce generative AI, thereby helping to expand the potential market. With these tech giants investing large amounts of capital, AI is stirring up a storm.

The wind of AI has also greatly benefited companies in the semiconductor supply chain. First, Nvidia dominates the entire generative AI chip market with its GPUs, while companies like SK Hynix and Samsung benefit from HBM. TSMC, responsible for packaging and foundry, is also in short supply, with production capacity in urgent demand. Assembly and testing factories such as ASE Technology Holding (ASE) and Siliconware Precision Industries (SPIL) have been able to secure packaging outsourcing orders from TSMC. Many AI chip players are also eyeing the market, and even IBM is promoting its AIU chip, which it has been researching for 5 years. The "gold rush" of generative artificial intelligence is leading some "shovel sellers" to become wealthy first.

Do you think that all the dividends of AI have been consumed by them? Actually, besides these direct beneficiaries, many device manufacturers and EDA/IP suppliers have unexpectedly become indirect beneficiaries. If AI chip manufacturers are the "shovel sellers," then they can be called the "shovel makers," and they are also enjoying the opportunities brought about by this transformation.

Device Manufacturers Unexpectedly Benefit

Most of the chips required in the generative AI field are produced using advanced processes. ASML, as the sole provider of EUV lithography machines for producing advanced process wafers, is definitely one of the winners. In the second quarter of 2023, ASML achieved a net sales of 6.9 billion euros, a gross margin of 51.3%, and a net profit of 1.9 billion euros. The company also has outstanding orders of 38 billion euros. ASML expects its sales to grow by 30% in 2023.

"Compared to traditional servers, advanced AI servers have significantly higher demand for cutting-edge logic, memory, and storage. For every 1% increase in the penetration rate of AI servers and data centers, an additional investment of $10-15 billion in (chip equipment) is expected. Currently, AI is still in its early stages, and more investment in factory and company equipment will be crucial in the coming years," said Tim Archer, CEO of Lam Research.

As of the quarter ending on June 25, 2023, Lam Research's revenue was $3.21 billion, with a net income of $803 million, exceeding expectations. In terms of regional sales, mainland China remains the largest revenue source for Lam Research, accounting for 26%, followed by South Korea at 24%, Taiwan at 20%, Japan at 10%, and the United States and Europe both at 8%, with Southeast Asia at 4%.

Archer stated that he expects the total market size of chip manufacturing equipment in the remaining time of 2023 to be approximately $70 billion. The demand for equipment and high-speed storage tools in China may drive the development of this industry. Since the implementation of export control restrictions by the United States in October 2022, Chinese companies have shifted their purchases to equipment for old logic and storage chips.

Testing equipment manufacturers are also beneficiaries. Many AI chips require 2.5D stacking, 3D stacking, and Chiplet technology, which increases the demand for testing equipment to ensure performance and quality. These chip tests require increasingly complex testing equipment to identify the location of manufacturing errors. Hideki Yasuda, an analyst at Daiwa Securities, stated, "Server chips will become larger and more complex, requiring more time for testing. There is no magic method to shorten the testing time. The only solution for chip manufacturers is to purchase more tools to test more chips simultaneously. Global revenue from high-performance GPU chip testing equipment may exceed that of smartphone chip testing equipment within a few years."

Benefiting from the growth in semiconductor demand driven by AI technology, in the second quarter of 2023, the revenue of the US chip testing equipment giant Teradyne was $684 million, with semiconductor testing business accounting for $475 million, system testing business for $94 million, wireless testing business for $44 million, and robotics business for $72 million.

Greg Smith, CEO of Teradyne, stated, "Our revenue reached the high end of the expected range, with semiconductor test shipments exceeding the soft demand for robotics within the quarter, and profits exceeding the plan, mainly due to higher gross margins. As we enter the third quarter, the demand for DRR5 and HBM memory device testing in data center applications remains strong, and the testing demand for SOC in automotive applications is gradually increasing. In the robotics field, we expect order rates to decline as customers respond to the slowdown in global industrial activities and macroeconomic pressures."

Mihashi, Co-Chief Strategy Officer of Advantest, a major testing equipment manufacturer in Japan, stated in a recent interview, "As industry leaders, we will benefit when high-performance computing applications such as ChatGPT expand. They also believe that the demand for AI will help revive the chip testing equipment.

Backend semiconductor equipment manufacturers have also enjoyed significant dividends. The shortage of chips behind generative AI has forced TSMC to repeatedly increase CoWos production capacity, even investing 90 billion yuan to build a new advanced packaging and testing factory in Taiwan. Therefore, equipment manufacturers have been pulled up. In order to meet the growing demand for CoWoS packaging, TSMC is collaborating with multiple global suppliers, including Rudolph Technologies from the United States, Disco from Japan, SUSS MicroTec from Germany, and Taiwanese experts Grand Process Technology (GPTC) and Scientech. According to DigiTimes, these suppliers have been asked to provide nearly 30 sets of tools by mid-2024.

EDA/IP Manufacturers Enjoy "Double Benefits"

In the past, during the downturn of the industry, the EDA and IP markets usually declined before the overall market slowdown, but recovered faster. However, this did not happen during this downturn. EDA vendors have shown strong performance throughout the entire pandemic and recovery period.

Specifically, looking at the financial reports of EDA vendors, Synopsys reported revenue of $1.395 billion for the second quarter of the 2023 fiscal year ending April 30, compared to $1.279 billion in the same period last year, a year-on-year increase of 9.07%, with a net profit of $273 million. For the third quarter, Synopsys estimates revenue to be between $1.465 billion and $1.495 billion, roughly better than market expectations. At the same time, Synopsys has raised its full-year revenue guidance for the 2023 fiscal year, expecting revenue to be between $5.79 billion and $5.83 billion.

Cadence achieved outstanding performance in the second quarter of 2023, with revenue of $977 million for the quarter ending June 30, compared to $858 million in the same period in 2022, and a net profit of $221 million. Anirudh Devgan, President and CEO of Cadence, stated, "With its unparalleled prospects, generative AI is beginning to have a significant impact globally. Our focus on AI over the past few years, combined with our expertise in computational software and valuable data on AI cores, puts us in a unique position to harness the enormous potential of this transformative technology." Cadence has also raised its full-year revenue forecast to slightly exceed Wall Street's expectations, expecting full-year revenue to be between $4.05 billion and $4.09 billion, a 14% year-on-year increase.

Regarding the impact of generative AI development on EDA vendors, unlike suppliers who only sell equipment and chips, EDA vendors can benefit from generative AI in at least two ways: by providing EDA tools for AI chip design, and by using generative AI to further help complete chip design by integrating it into their own software.

With more and more system vendors such as Google, Meta, and Alibaba developing their own AI chips, they have become one of the major buyers of EDA. Walden C. Rhines, initiator of the SEMI Electronic Design Market Data Report, stated, "The electronic design automation (EDA) industry continued to achieve double-digit growth in the first quarter of 2023, with all product categories and geographic regions experiencing growth. These product categories include computer-aided engineering, IC physical design and verification, printed circuit boards and multi-chip modules, and services, all showing double-digit growth."

The application of AI in EDA software is not new. The three EDA giants, Synopsys, Cadence, and Siemens, have all launched their own AI tools. Existing AI tools have already provided significant improvements in productivity and speed for chip manufacturers, gradually demonstrating their advantages. Therefore, the development of generative AI, for EDA vendors, in the long run, will be an added bonus.

In April 2023, Siemens and Microsoft announced that they are collaborating to use generative AI in the design, engineering, manufacturing, and operational lifecycle of industrial products to enhance innovation and efficiency. The two companies will integrate Siemens' product lifecycle management software Teamcenter with Microsoft's collaborative platform Teams, the language model in Azure OpenAI services, and other Azure AI capabilities.

Synopsys engineers are exploring how advanced large language models (LLMs) like ChatGPT can help simplify internal processes and enhance existing solutions.

KT Moore, Vice President of Enterprise Marketing at Cadence, stated at a seminar that generative AI can help build learning datasets. In turn, these datasets can be used to create other future designs.

However, while generative AI has shown excellent results in language and image aspects, its development is still in the early stages, and using it exclusively for chip design may have flaws. Actual chip design requires a high level of precision (9 nines), as even a tiny error could have significant consequences in terms of efficiency, yield, time to market, and other aspects.

Conclusion

The decline in sales of personal computers and smartphones has brought tremendous pressure to the industry. However, the rise of artificial intelligence, especially the strong momentum driven by generative AI tools such as ChatGPT and Stable Diffusion, has to some extent changed this situation. They have not only alleviated the impact of post-pandemic sales decline on the entire semiconductor industry but also opened up new business opportunities and possibilities for the entire industry. Looking ahead, one thing is clear: AI is not just a part of the technology industry; it is becoming the dominant force in the technology industry.

Akira Minamikawa, Senior Consulting Director at Omdia, pointed out in a seminar on semiconductor market trends, "Generative AI has already developed to account for about 20% of data center applications within three years and will increase processing power by about 10 times. To meet this demand, we will need to expand the number of centers to 1.7 times the current level, so there will be significant investments in the future."

The revenue obtained from generative AI markets such as ChatGPT will not be a short-term accidental windfall, but will be an important source of revenue in the semiconductor field for a long time to come.

免责声明：本文章仅代表作者个人观点，不代表本平台的立场和观点。本文章仅供信息分享，不构成对任何人的任何投资建议。用户与作者之间的任何争议，与本平台无关。如网页中刊载的文章或图片涉及侵权，请提供相关的权利证明和身份证明发送邮件到support@aicoin.com，本平台相关工作人员将会进行核查。