Apr W4｜Technology News Highlight : UMC Throws a Shock Bomb! Exit from UMC Board of Directors - Chipotle Industry Trend Research Institute

Apr W4｜Technology News Highlight : UMC throws a shockwave! Withdrawal from UMC's Board of Directors

Published On: 2024/05/02|Categories: 產業快訊(News)|

Apr W4｜UMC Throws a Shock Bomb! Withdrawal from UMC's Board of Directors

Driver IC major UMC will hold a regular shareholders' meeting on May 31st and completely reelect its directors. Recently, the company announced that the list of eight candidates for the board of directors are all natural persons, and UMC, the major shareholder and the current corporate director, is not even included in the list. This is the first time that UMC has withdrawn from the board of directors of UMC since the company split up and became independent from UMC in 1997, which has shocked the industry.

Ever since its establishment, UMC has long been categorized as the "UMC Army" and the "Golden Chicken" of the UMC Group. In recent years, UMC has earned more than one share capital every year, with net income per share of NT$63.87, NT$45.96, and NT$38.32 for the years from 2021 to 2023, making it a "Profitable Performer" in the Taiwan stock market. Recently, there have been murmurs about the recovery of the semiconductor industry, but UMC has withdrawn from the board of directors of UMC, letting go of this "big golden chicken" and shaking the market. According to the latest statistics announced in March this year, UMC holds about 16,400 shares of UMC, with a shareholding ratio of about 2.7%. After UMC's withdrawal from the board of directors of UMC, it has aroused concern as to whether UMC will divest itself of UMC's shareholding in the future.

Regarding UMC's withdrawal from the Board of Directors, UMC pointed out that the list of director candidates was an effort in the direction of corporate governance by the competent authorities to increase the proportion of independent directors and women. In the new term of UMC, the number of independent directors has increased from three to five, and the overall list of director candidates is evenly split between men and women, further strengthening the independence and diversity of the board. UMC responded that it respected the list proposed by UMC.

Ji-Pu's point of view.

The industry speculates that this move may be related to the fierce competition in the semiconductor market in recent years. Especially in the field of panel driver ICs and OLED driver ICs, UMC's main customers come from the mainland market. Facing the pressure of the US-China trade war, UMC, as a foundry, needs to make a cut with its Chinese customers and make its position clear. On the other hand, after UMC's withdrawal from the board of directors, UMC will be able to handle its shares in UMC more flexibly, which will be helpful for capital operations such as expanding plants or strengthening R&D investment, etc. Novatek, on the other hand, will be able to handle its shares more flexibly after its withdrawal from the board of directors. Novatek, as an IC design company, will also want to have more flexible foundry partners, and will think of different foundries to submit their products. If it keeps the name of UMC, it will have the concern of competition in the industry. In the series of changes, how to stabilize operation and innovate will be the focus of UMC's continuous attention in the future.

Apr W3｜SK Hynix and TSMC Collaborate to Develop Next-Generation HBMs

South Korean memory maker SK Hynix announced on the 18th that it has signed a Memorandum of Understanding (MOU) with TSMC to jointly develop the next generation of High Frequency Memory (HBM) chips.

SK Hynix is the leader in memory for AI applications and dominates the HBM chip market, which is one of the key components for AI technology. SK Hynix has announced a partnership with TSMC, the world's top logic foundry, to create a powerful alliance and to throw a shockwave into AI technology innovation.

Justin Kim, President and Head of AI Infrastructure at SK Hynix, said, "Through this partnership, we will further strengthen our market leadership as an AI memory supplier by enhancing our competitiveness in the area of customized memory platforms."

SK Hynix also said that the two companies will first focus on improving the performance of the Base Die, which is the bottom layer of the HBM package. HBM is made by stacking multiple DRAM chips (Core Die) on top of the Base Die and vertically connecting them through TSV technology. The Base Die is also connected to the graphics processing unit (GPU), which controls the HBM.

Ji-Pu's point of view.

Both parties will strengthen the integration of SK Hynix's HBM and TSMC's CoWoS advanced packaging technologies to meet the needs of related customers. Currently, the Base Die [Control IC or read/write IC, etc.] is OEM for TSMC, and then the HBM4 packaged by SK Hynix is shipped to TSMC; it has been rumored that in the future SK Hynix would like to directly ship the TSV stack to TSMC after the stacking is completed, but yield is still the biggest point of consideration [the current yield of the HBM3E is about 70~80%]. CoWoS is TSMC's world-leading technology, a packaging method that connects logic chips and high-bandwidth memories together, and is currently used to meet the demand for high volume computing.

Apr W2｜Intel's foundry suffers $220 billion loss

Intel's foundry division suffered an operating loss of US$7 billion (NT$224.4 billion) last year, partly because it did not use state-of-the-art equipment to reduce production costs. However, the company remains optimistic about the future, with plans to spend US$100 billion to build or expand fabs in four U.S. states.

Intel filed its 2023 financials with the U.S. Securities and Exchange Commission (SEC) on February 2, with total revenue slipping from $57 billion in 2022 to $47.7 billion in 2023. In terms of departments, the data center and artificial intelligence (AI) division's revenue was $12.64 billion in 2023, much lower than the $16.86 billion in 2022, while the foundry business last year's revenue was $18.9 billion, sharply lower than the $27.5 billion in 2022, a 31% contraction, and the operating loss reached $7 billion, a further widening of the $5.2 billion loss from the previous year. The operating loss was US$7 billion, widening again from the previous year's US$5.2 billion.

CEO Pat Gelsinger admits that this year will be the worst year for the division in terms of operating losses. With the current plan to 2030, the operation will gradually improve in the future, probably in the middle of this planning period will be able to make a profit and loss of two flat. Based on this estimate, it is expected that the point of no longer losing money will be in 2027.

Intel is optimistic about further growth in foundry revenue margins in the future, primarily through expanding the ratio of in-house wafers produced, expanding the advanced packaging business, expanding external foundry operations, increasing capacity utilization, cost savings, and scaling up production.

Intel 2021 announced the creation of a new independent foundry services (IFS) division in March, with the goal of providing investors with a better understanding of the value of its products in relation to each of its businesses, including foundry.2 The filing was the first time that it had reported sales results for this division.

Ji-Pu's point of view.

Kissinger said the division's operations have deteriorated because of a series of bad decisions, including opposing the use of an Extreme Ultraviolet Lithography (EUV) machine from Eskimo a year ago. In retrospect, although an EUV costs more than $150 million, it is more cost-effective than the current lithography and can produce the most advanced wafers. Similar bad decisions have led Intel to allocate about 30 percent of its overall wafer production to outside contractors such as TSMC. Tiesinger said Intel is now aiming to reduce the outsourcing ratio to about 20 percent. In line with that goal, it is already replacing older equipment with EUVs. But looking at it from another perspective, the current losses could well be the first step toward accelerating the breakup of IFS, and moving toward a separation of light and heavy assets would be a way for Intel to realize better shareholder equity, and would be the kind of script the U.S. government wants.

Related News : Intel's foundry suffers $220 billion loss

Apr W1｜SK Hynix plans to build a packaging plant in the U.S.

According to sources familiar with the matter, South Korea's SK Hynix plans to build an advanced chip packaging plant in West Lafayette, Indiana, and invest about $4 billion, a major boost to the U.S.'s goal of gaining semiconductor power.

It is expected to create about 800 to 1,000 jobs, and grants from the state and federal governments will be available to fund the project. According to people familiar with the matter, the plant is expected to be operational by 2028, and the SK Hynix board is expected to vote soon to finalize the decision to build the plant.

The Icheon, South Korea-based company's market capitalization has more than doubled in the past year to 12.9 trillion won ($96 billion), and SK Hynix CEO Kwak Noh-jung previously said the AI frenzy was expected to help the company's market value jump to 200 trillion won.

Ji-Pu's point of view.

SK Hynix is one of the world's top chipmakers, and it is also capitalizing on the recent artificial intelligence (AI) boom. The company dominates the high-bandwidth memory (HBM) market and is currently the exclusive partner of Phillip's state-of-the-art graphics processing units (GPUs). SemiAnalysis estimates that SK Hynix has a market share of about 73% of HBM, Samsung about 22%, and Micron about 5% in terms of billions of units, making it the leader among HBM vendors.

Mar W2｜Huang Yanxun Showcases New Technology at GTC

At this year's GTC stage, Pfizer CEO Jen-Hsun Huang emphasized that data centers will face a huge transformation, and appeared to announce four sensational technological highlights, "AI Super Chip GB200 Grace Blackwell," "AI In-vehicle Platform DRIVE Thor," "Humanoid Robot Technology," and "Operational Micrographics Platform Introduced to Semiconductor Manufacturing," and so on.

In the face of the trend of multimodal (supporting text, image, video, and sound simultaneously), the large language model parameter count is only going to get higher and higher, and even the general AI model training parameter count continues to go up. In order to cope with the next-generation trend of 10 megabytes of big speech model parameters, Jen-Hsun Huang showed off Pfizer's newest GPU architecture, Blackwell, at this year's GTC developer conference.

The Blackwell architecture GPUs are equipped with 208 billion transistors and TSMC's customized 4nm process, with two GPU dies interconnected to form a single, unified GPU, and built-in high-bandwidth memory of up to 192GB. Compared to the existing H100, the GB200, which is the highest-priced and the most powerful of the new Blackwell architecture products, delivers four times higher training performance, 30 times higher performance, and 25 times higher energy efficiency. The GB200 will be the first to use water cooling, which circulates two liters of water per second to reduce the AI server, which is running hot at high speed, to 45 degrees Celsius.

In addition, DRIVE Thor, the latest AI in-vehicle platform based on the Blackwell architecture, was introduced at the conference and has been adopted by a number of car manufacturers in China. It is reported that BYD will build its next-generation electric vehicle fleet on DRIVE Thor.

The humanoid robot project is predicted to set off a global robotics boom again. Jen-Hsun Huang emphasized that the next generation of robotics will probably be more humanoid-like. In addition to the general basic model of the humanoid robot, "Project GR00T," Pfizer also announced Jetson Thor, a new computer based on the Thor System SoC for use in humanoid robotics.

Jen-Hsun Huang also mentioned that the greatest value of generative AI is to create more possibilities, so that experts in various fields can use generative AI to innovate. These experts no longer need to be experts in programming language at the same time, but can use natural language to interact with the generative AI, and at the same time create more possibilities, not only in the entertainment and art, but also in programming, semiconductor design, energy exploration, drug research, etc. NVIDIA also announced a new NIM microservice to make it easier to customize. NVIDIA also announced a new NIM microservice that makes it easier to customize generative AI, and at the GTC, it also announced that cuLitho, a GPU-accelerated micrographics technology in collaboration with TSMC and Synopsys, will combine generative AI to take advantage of the stereotype-independent nature of generative AI to not only make micrographics faster, but also to help design more efficient graphics.

Especially in the smartphone market, in the past, WiFi 6 was only available in the Americas market, but after the iteration to WiFi 7, all Android smartphones need to be equipped with FEM. Although the penetration rate of cell phones is limited this year, the company still predicts that by 2025, it will reach 20% to 30%.

Ji-Pu's point of view.

The Blackwell system is equipped with a 2080 transistor and uses TSMC's customized 4nm process, which consumes 1/25th of the previous power consumption. Also announced is support for timely streaming of cloud computing results to Vision Pro, which provides a spatial computing experience as long as the device is connected to the network. The Isaac Manipulator is an AI designed for robotic arms, and there is also vision-enhancing AI that enables robots to see more clearly and increase efficiency. Generative micro AI services are also one of the key focuses of the conference, allowing companies to build customized AI on their own platforms, such as the breakthrough Phaidon Computational Micro Imaging platform that TSMC and Synopsys have formally announced will be put into production. According to TSMC President Chieh-Chia Wei, the company is working with Phaidon to integrate GPU-accelerated computing into the company's workflows to realize the need for dramatically higher performance, increased productivity, shorter cycle times, and reduced power consumption. TSMC is also currently putting cuLitho into production.

Mar W1｜Leaderboard WiFi 7 takes a big step forward

RF IC company Lapis held a press conference on March 1, its company has WiFi major main chip platform reference design, WiFi products accounted for 88% of the total revenue in the fourth quarter of last year, management said, WiFi 7 product line continues to expand, has been sent to the sample development and design in customers' products.

The year 2023 was the lowest year of operation for LAPIS, with annual consolidated revenue of NT$2.985 billion, a 13.0% annual decrease, marking a four-year low, and after-tax earnings per share of NT$2.46, a loss for five consecutive quarters. However, starting from the fourth quarter of last year, demand has gradually improved, with a rebound in demand from mainland customers and an increase in market share in North America, which has put the company back on the growth track. Looking ahead to 2024, the first quarter will be flat compared to the fourth quarter last year. The market expects strong growth in the second quarter, benefiting from the continued volume of WiFi 7. Currently, LAPIS has already passed the WiFi 7 Router certification of its customers, and its handsets have also been certified by Taiwan brand manufacturers.

Its VP of Business Marketing, Mr. Jack Huang, pointed out that the overall penetration rate of WiFi 7 will increase in the second half of the year, and is expected to reach 10%, and WiFi 7 will be the major part of the cooperation projects with customers. In addition, in terms of FEM (front-end radio frequency module), the unit price will grow from 60% to 100% of WiFi 6. In targeting the relatively high-end market, LAPIS is working closely with major chip platforms.

Ji-Pu's point of view.

This year, the growth of LAPIS will come from telecom operators and the demand is huge, FEM is estimated to reach 150 million units, the mainland market bidding, Europe and the United States, India market is also very good. However, WiFi 5/6 is still the main source of revenue, and WiFi 7 is expected to reach 10% at most. The company estimates that the second quarter will see a growth explosion, with a quarterly growth of over 20%, and the proportion will also increase quarter by quarter. Its WiFi 7 solution has been certified by Router, a major chip maker, and its handset has also been certified by a major international company, and it will start shipping Korean handset brands in the second half of the year.

Related News : A Big Step Forward for Lijia WiFi 7

Feb W3｜Cheng-Yi Sun creates AI chip company to rival Nvidia.

U.S. financial media revealed today that Masayoshi Son, CEO of Japan's Software Bank Group, intends to raise US$100 billion (about NT$3135 billion) to establish an artificial intelligence (AI) chip company to challenge Nvidia's leading position in this field.

Bloomberg quoted sources as reporting that Sun has named the project "Izanagi," the name given to the AI chip company. According to the report, the reason why Sun named the project Izanagi is because Izanagi is the god of creation and life in Japanese mythology, and Izanagi is written in English as Izanagi, and the last three letters of the name are AGI, which is the abbreviation of Artificial General Intelligence.

Sun has repeatedly predicted in recent years that AGI will be the future of human civilization, believing that a world full of AI machines will be safer, healthier and happier than human society.

The SoftBank Group may contribute $30 billion to the program, with the other $70 billion perhaps coming from some institutions in the Middle East, said the sources, who asked not to be named.

It's not clear what Sun's plans are for the AI company or where the money is going, people familiar with the matter said. Sources familiar with the matter said the plan is still under further study.

The source also said that although Sun has discussed with OpenAI CEO Sam Altman about building a semiconductor business and raising funds, it does not appear to be the same as the Izanagi project at this time. Sun has also approached another company to develop a set of basic AI models, but has been rebuffed.

Ji-Pu's point of view.

If Sun's plan succeeds, it would be the largest investment in AI since Microsoft Corps invested more than $10 billion in OpenAI's development of ChatGPT, a generative AI chatbot. The AI chip company that Sun is considering creating to complement Arm, the cell phone chip design company in which SoftBank holds a majority stake, plans to work with Arm's CEO Rene Haas on the development of "Izanagi". The head of Armor also said earlier this month that the company sees strong demand for Armor's central processors, comparable to the demand for Fidelity's data center artificial intelligence chips. This move is supposed to transform ARM into an IC design company, using Arm's strengths to seek greater profits.

Feb W2｜Photonix Establishes New Division for Customized AI Chips

Phaidon currently dominates the design and supply of AI chips worldwide, with a market share of about $80% in high-end AI chips, which has seen its market capitalization soar by $40% so far this year to $1,730 billion after more than tripling last year. Reuters quoted nine sources as saying that Pfizer is building a new business unit specializing in designing custom chips for cloud computing companies and others, including advanced AI processors.

According to Reuters, Pfizer hopes the new unit will take a bite out of the rapidly growing global demand for customized AI chips, which are booming at a rapid pace now that the company is facing a growing number of companies developing alternatives to capture its market. Pfizer's customers, which include chatbot ChatGPT developer OpenAI, Microsoft, Google's parent company Alphabet, and Facebook's parent company Meta, are now competing to buy smaller and smaller supplies of chips to compete in the fast-emerging AI space.

For these technology giant customers, the H100 and A100 wafers supplied by Phaidon were intended to be general-purpose, multi-purpose AI processors. However, these companies have begun to develop their own chips to meet specific needs, reducing energy consumption and potentially saving cost and time in chip design. With that in mind, the sources said, Fidelity is trying to play a role in helping these companies develop customized AI chips, a type of business that would have gone to Fidelity's rivals such as Broadcom and Mitsubishi.

Pfizer has not announced a price for the H100, but it is bound to be higher than that of the previous generation A100. Depending on sales volume and other factors, the price per wafer should be in the range of $16,000 to $100,000. Meta plans to have a total inventory of 350,000 H100 wafers in stock this year.

Fidelity has met with representatives from Amazon, Meta, Microsoft, Google and OpenAI to discuss making customized chips for them, two sources familiar with the meetings said. In addition to data center chips, Pfizer is targeting telecom, automotive and gaming customers.

Ji-Pu's point of view.

According to analysts at an independent investment bank in New York, the customized chip market will reach about US$30 billion in 2023, accounting for about 5% of annual global chip sales. Vickers, a researcher at U.S.-based market research firm 650 Group, estimated that the custom data center chip market will grow to as much as US$10 billion this year and double to US$20 billion next year. The market is currently dominated by Broadcom and Mitsumi, and if Fidelity joins the market, it may erode the sales of both companies.

Feb W1｜Canon leverages ASML for chip manufacturing equipment

Japan's camera and printer maker Canon (Canon) is ready to attack the market for advanced semiconductor manufacturing equipment, will launch a new machine that does not use photo-etching but instead uses imprinting, with a price that is claimed to be less than that of the Dutch industry's ASML equipment by a fraction of a point and a reduction in power consumption by up to 90%, which will be shipped as early as this year.

The Financial Times (FT) reported that Canon released its nanoimprint lithography technology in mid-October last year, which does not use light to etch, but instead stamps the wafer design onto the silicon wafer. Canon said that this technology has been developed for more than 15 years, but not commercially viable until now, compared to ASML's market-dominating extreme ultraviolet (EUV) lithography, the new process is cheaper by "one digit", and power consumption can be reduced by up to 90%.

We hope to start shipping this year or next year, and want to launch when the market is hot," said Yoaki Takeshi, who is in charge of developing the new microfilm device at Canon, "It's a very unique technology that allows for simple and low-cost manufacturing of cutting-edge wafers."

According to Canon, the nanoimprint exposure equipment is not intended to replace existing EUV and DUV exposure equipment, but rather to coexist in the production line. In addition to logic wafers, the equipment will also be used to make 3D NAND Flash memory.

Currently, due to the high demand for exposure equipment in the semiconductor market, Canon may find further entry into the semiconductor equipment market by utilizing cheaper equipment and production costs as a selling point. However, nano-embossing technology is not compatible with EUV and DUV exposure equipment, and it would be complicated to add to the existing manufacturing processes of semiconductor manufacturers. However, market sources also point out that Chinese manufacturers are currently finding it difficult to purchase advanced EUV and DUV exposure equipment due to stringent U.S. export restrictions on semiconductor equipment. It will be interesting to see if Canon's inexpensive and efficient nanoimprint exposure equipment can help China's semiconductor industry.

Ji-Pu's point of view.

Extremely precise EUV equipment is a necessity for TSMC, Samsung, Intel, and others to produce the latest generation of wafers, and ASML is currently the only company capable of producing such equipment. ASML is currently the only company capable of producing this equipment. However, ASML equipment is also the most expensive part of the semiconductor manufacturing process, costing more than $150 million per unit, with the latest High-NA EUVs costing as much as $400 million per unit, with a long lead time for delivery, and with major companies around the globe scrambling for them. Canon's biggest challenge is twofold. First, it has to improve the success rate of its further microfabrication process. The company started at 5 nanometers and is aiming for 2 nanometers. Yomei Takeshi, who is responsible for the development of the new microfilm equipment, was reluctant to disclose the potential yield of the nanoimprinting equipment, but analysts said that the yield must be close to 90% in order to compete with the EUV; and the second is the introduction of major customers. For the semiconductor process, the original use of the process sequence film -> yellow light -> etching -> diffusion -> CMP, etc. will be imported into the embossing to be all changed, which means that the product needs to be re-tested to test all the validation, which will be a huge cost, so it should be very difficult to see in the short term in the mainstream process.

Jan W4｜Intel joins hands with UMC for 12 nanometers

Intel and UMC have announced a partnership to develop a 12nm process platform, which is expected to enter production in 2027. With UMC providing a wide range of technical services and Intel providing off-the-shelf facilities, the two companies are operating together in what should be a win-win situation.

Taiwan has been a key member of the semiconductor and broader technology ecosystem in Asia and globally for decades, and Intel's commitment to partnering with Taiwanese innovators such as UMC will better serve our customers around the world," said Stuart Pann, senior vice president and general manager, Foundry Services (IFS), Intel. Intel's strategic partnership with UMC further demonstrates its commitment to providing technology and manufacturing innovation to the global semiconductor supply chain.

UMC's collaboration with Intel on U.S.-made 12nm FinFETs is an important part of the Company's strategy to pursue cost-effective capacity expansion and technology node upgrades," said Steve Wang, co-president of UMC. This collaboration will help customers to smoothly upgrade to this critical technology node while benefiting from the supply chain resiliency that comes with expanding capacity in the North American market. We look forward to working strategically with Intel to capitalize on their complementary strengths to expand potential markets while significantly accelerating the technology development timeline.

This 12nm process will combine Intel's U.S.-based large-scale manufacturing capabilities and FinFET transistor design experience to provide a powerful combination of maturity, performance and power efficiency. UMC's process leadership and decades of experience in providing PDK and design to customers allows for more efficient foundry services. The new processes will be developed and manufactured at Intel's Fabs 12, 22 and 32 at Ocotillo Technology Fabrication in Ocotillo, Arizona, U.S.A., where the fab's existing equipment will significantly reduce upfront investment and optimize utilization.

Both companies are committed to meeting customer needs by collaborating to support design enablement for 12nm processes through electronic design automation (EDA) and intellectual property (IP) solutions provided by ecosystem partners. Production is expected to begin in 2027.

As a matter of fact, not long ago, it was reported in the market that UMC had transferred its 12nm advanced process technology to Intel and received a $10 billion license. Moreover, UMC does have advanced process technologies, including the 14nm Finfet process, the performance of which has reached the competitive level in the industry, and is also actively promoting the 12nm process. In this regard, market participants said that it would not be cost-effective for Intel to license UMC for tens of billions of dollars and then manufacture in other regions. Instead, it will be more cost-competitive to order directly from UMC.

Ji-Pu's point of view.

In the past, UMC has been rumored to be transferring 12nm advanced process technology to Intel's dominant Arm architecture, which is complementary to Intel's focus on the x86 architecture. Intel has also had experience with the Arm architecture in the past, such as infrastructure processors (IPUs) and the development of low-power system-on-chips (SoCs) based on Intel's 18A process, and has strengthened its presence in end-use applications through its partnership with Arm. As a result, the market speculated that Intel may be laying the groundwork for a complete breakup of the Fab by taking the lead in seeking IP licenses for the 12nm process Arm architecture that was previously rumored to be in the works at UMC. This partnership not only helps Intel to bridge the gap from IDM to foundry business model, increase flexibility in process tuning, and gain foundry operation experience, but also allows UMC to flexibly utilize FinFET capacity without incurring huge capital expenditures, thus creating an alternative way to survive in the mature process competition, and indirectly expanding the international distribution of factories and dispersing geopolitical risks by co-operating Intel's U.S. sites. This should be a win-win situation for both companies.

Related News : Intel joins hands with UMC to attack 12nm

Jan W3｜Intel says German 1.5 nm still not as good as customer's perspective

Intel Corp. CEO Pat Gelsinger revealed that once the German plant is up and running, it will be the most advanced foundry in the world, building 1.5nm wafers at that time. It is worth noting that TSMC (2330) revealed that the 18A process performance is similar to the company's N3P, but from the customer's point of view there is a big difference.

According to Tom`s Hardware, Gelsinger said at the World Forum in Davos, Switzerland, on the 17th that Intel's plant near Magdeburg, Germany, will not only be the most advanced in Europe, but also the world's most sophisticated foundry. The facility will utilize the 18A Post-Advanced process and will build wafers for both Intel and Intel Foundry Services customers.

However, Gelsinger did not specify which post-18A advanced process would be used at the Madsenburg plant, but only vaguely indicated that it would be similar to the 1.5 nm process. According to the report, Intel will unveil a roadmap for the post-18A process at the end of February, presumably indicating which facility will be the first to implement the process. Previous rumors have suggested that Intel's post-18A advanced processes may be 16A and 14A.

So far, Intel's most advanced processes are Intel 4 and Intel 3, behind TSMC's 3nm process "N3". Intel expects 18A and later processes to lead the industry in performance, power consumption and area (Performance, Power, Area, PPA).

Gelsinger has previously said that the Intel 18A process and TSMC N2 process transistors seem to be about the same, with no significant advantages, but everyone says Intel's backside power delivery is superior. This, he says, gives silicon chips better area efficiency, which means lower cost and higher performance.

Ji-Pu's point of view.

A few days ago, TSMC held a press conference on the 18th. When asked about Intel's claim that its process PPA is superior to TSMC's 2nm, CEO Chieh-Chia Wei reiterated that Intel's latest technology is very similar or equivalent to TSMC's N3P. TSMC Chairman Dennis Liu followed up by saying that although Chieh-Wei was very modest in saying that the N3P is similar to the 18A PPA, we can look at it from another angle - it's just for Intel's own products. An IDM usually optimizes its technology for its own products, while a foundry, TSMC, optimizes its technology for its customers' products. There's a big difference.

Jan W2｜TSMC's revenue dominated last year, surpassing Samsung's.

Well-known research organization TechInsights Canada's semiconductor market trends research department McClean Report (formerly IC Insights) announced 2023 semiconductor vendors ranked in dollar sales forecast of the top 25, the organization's data show that Samsung Electronics due to memory market headwinds, resulting in a significant decline in revenue, TSMC dollar revenue is expected to overtake Samsung! TSMC is expected to overtake Samsung in terms of dollar revenue, with Intel coming in second.

According to the organization's projection, TSMC's dollar revenue in 2023 is estimated to reach $68.852 billion, a yearly decrease of about 9%, but still tops the list, followed by Intel's $51.401 billion, a yearly decrease of 16%. Samsung Electronics had been the top semiconductor sales company in 2022, but fell to fourth place in 2023, with an estimated $48.304 billion in dollar sales in 2023, a yearly decrease of 37%. The company's dollar revenue in 2023 is estimated to be $48.304 billion, a decrease of 371 TP3T annually.

In addition, it is estimated that among the top 25 companies in the world (including IC design companies), there are only seven companies that have realized annual growth in revenue, and these companies are Phaidon in the third place, Broadcom in the sixth place, Infineon in the ninth place, STMicroelectronics in the tenth place, NXP Semiconductors in the fifteenth place, Sony in the seventeenth place, and Microchip in the nineteenth place.

Intel, which was ranked second in 2023, improved its ranking by one place, although revenue declined 161 TP3T from the previous year.

In third place is Pfizer with an estimated annualized revenue growth rate of approximately 102%, which jumps Pfizer's ranking from eighth in 2022 to third in 2023.

Ji-Pu's point of view.

TSMC announced its December 2023 revenue report on the 10th, with consolidated revenue of approximately $176.3 billion, a decrease of 8.4% from the same period last year. although it is projected that TSMC's full year 2023 revenue is estimated to reach $68.852 billion, a decrease of approximately 9% annually, it is still in the top spot. mcclean pointed out that last year's semiconductor company's revenue downturn was mainly due to the "declining demand" and "smart phones, personal computers, data center servers and other end-use products inventory increase".

Jan W1｜Intel Launches New Company to Focus on Generative AI

According to a new report from Reuters, Intel, backed by digital-focused asset manager DigitalBridge Group and other investors, has announced the formation of a new, independent company that will operate around its artificial intelligence software business. Intel executives did not say whether they would retain a majority stake in the new company, but they did confirm that the new company, called Articul8 (pronounced "Articulate AI"), will have an independent board of directors. But they did confirm that the new company, called Articul8 (pronounced "Articulate AI"), will have an independent board of directors, with Intel remaining a shareholder and Arun Subramaniyan, former vice president and general manager of Intel's data center, serving as CEO.

This achievement resulted from a collaboration between Intel and Boston Consulting Group (BCG) in the area of enterprise AI technology, where Intel used its supercomputers to develop a generative AI system that analyzes text and images through a combination of open source and proprietary technologies. The system was then used to run in BCG's data center, addressing privacy and security concerns.

Intel said in a press release that Articul8 was created on the basis of intellectual property (IP) and technology developed by Intel, and that the two companies will maintain a strategic alliance on go-to-market opportunities and collaborate to drive GenAI adoption in the enterprise.

Subramaniyan told Reuters that many potential customers are concerned that costs could spiral out of control if many of their employees start using the system. He also said that Intel and partners have been working hard to address the biggest gap that exists in generative AI today, which is that "it's easy to set up a proof-of-concept, but it's hard to put it into production in a safe and cost-sustainable manner.

Subramaniyan said in the interview, "We've been trying to address the biggest gap in biomorphic AI today, which is that it's easy to build a proof-of-concept, but getting things into production and doing it in a safe and cost-sustainable way is really what's missing right now."

In addition to DigitalBridge, Articul8's other investors currently include Fin Capital, Mindset Ventures, Communitas Capital, GiantLeap Capital, GS Futures and Zain Group.

Ji-Pu's point of view.

In an interview, Subramaniyan noted that Articul8's main goal is to fill the generative AI gap, demonstrating safe and cost-effective concepts. Intel's move is in line with its broader strategy of seeking external capital to support its business units, following the split of Mobileye Global and the planned initial public offering (IPO) of its programmable chips division. programmable chips division is also planning an IPO. Articul8 has built a scalable and easy-to-deploy GenAI software platform that has also enabled organizations to unlock value from their proprietary data.

Dec W3｜Taiwan TSMC Kumamoto Plant JASM will be completed soon.

TSMC's foundry services subsidiary JASM (Japan Advanced Semiconductor Manufacturing), established in Kumamoto Prefecture, Japan, said Yuichi Horita, president of TSMC's Kumamoto plant construction project is smooth and will soon be completed, with the goal of increasing the proportion of local purchases in Japan to 60% in 2030.

Regarding the progress of construction of TSMC's semiconductor plant in Kikuyo Town, Kumamoto Prefecture, Nikkei News 15 reported that Yuichi Horita, president of JASM, a subsidiary responsible for the operation of the Kumamoto plant, said in a speech at the Japan International Semiconductor Expo "SEMICON Japan 2023," that "the construction project is almost on schedule and will be completed soon. Mr. Horita also pointed out that as many as 6,000 workers were involved in the construction of the plant at its peak, and that they have been moving in manufacturing equipment and connecting equipment since October.

In addition to TSMC's existing suppliers, there are 120 new Japanese companies conducting transactions, and the proportion of local procurement in Japan is currently around 25%, and Yuichi Horita pointed out that "it is planned to increase to 50% in 2026 and further increase to 60% in 2030".

It has been rumored that TSMC's Kumamoto plant will have an opening ceremony in February 2024, but Yuichi Horita did not mention whether the new plant will have an opening ceremony or anything related to TSMC's second plant under evaluation. However, Yuichi Horita did not mention whether an opening ceremony will be held for the new plant, nor did he mention anything about the second Kumamoto plant that TSMC is evaluating.

TSMC Kumamoto plant is expected to start mass production by the end of 2024, producing 22/28nm and 12/16nm process wafers, with a monthly capacity of 55,000 wafers, while Sony's "Sony Semiconductor Solutions Company", as well as Japan's Toyota group of companies, Japan's automotive component maker Denso are all investing in JASM.

Ji-Pu's point of view.

TSMC previously indicated that it was considering the construction of a second plant near the first plant in Japan. The Japanese media pointed out in October this year that TSMC was evaluating the construction of a second plant in Kumamoto Prefecture, Japan. The total investment in the new plant is estimated to be as high as about 2 trillion yen, and the Ministry of Economy, Trade and Industry (METI) of Japan is considering a maximum subsidy of 900 billion yen, and it plans to utilize the second plant in Kumamoto for the production of 6nm wafers. In addition, Bloomberg reported on November 21 that TSMC is considering building a third plant in Kumamoto Prefecture to produce the most advanced 3nm wafers.

Dec W2｜Intel launches new AI chip to capture market next year

Chip maker Intel announced its new chips for generative artificial intelligence (AI), personal computers (PCs) and servers on April 14, U.S. time, focusing on the performance of AI PCs and AI servers, in the hope of capturing the AI hardware market. Intel CEO Pat Gelsinger said that he has seen the generative AI boom as the star product of 2023, and AI PCs will be the protagonist in the coming year, and Intel's new Core Ultra processors will also play a big role.

The Intel Core Ultra laptop processor, code-named Meteor Lake, is the first chip to utilize the Intel 4 process, and is the biggest architectural change in Intel's 40-year history. For the first time, the Intel Core Ultra processor is equipped with a Neural Processing Unit (NPU), which complements the AI processing of both the CPU and the GPU to achieve high-performance AI acceleration, and improves energy efficiency by up to 2.5 times compared to its predecessor, demonstrating the company's intention to aggressively capture the AI PC market. Intel expects that by 2024, the AI PC market will be dominated by the NPU. Intel expects more than 230 computers powered by the new Intel Core Ultra processor to launch in 2024, and more than 100 million Intel AI-enabled PCs in 2025.

In addition to the Core Ultra chipset, Intel also announced its fifth-generation Xeon server processor, touting it as the only mainstream data center processor in the market with built-in AI acceleration. The processor, which includes an NPU, will be available in the first half of next year for deployment in large enterprises such as cloud companies. Intel said the processor performs particularly well on AI inference tasks, with a performance improvement of 36% as measured per watt of power.

Intel also showed off the first working version of its Gaudi 3 chip, which it hopes will challenge Nvidia in the data center AI market. Not much detail was given about the chip, though, other than to say that Gaudi 3 will compete with Prada's H100 and Supermicro's upcoming MI300X. The former is the main choice for companies that have built large chip farms to power AI applications, and the latter's new MI300X will ship to customers in 2024.

Gaudi3 performance has taken a big leap forward, with 1.5 times the bandwidth of its predecessor Gaudi2 7nm, a four-fold increase in BF16 (Brain Floating Point) power, and a two-fold advancement in network arithmetic. Gaudi3 is also expected to be equipped with up to 128GB of HBM3e memory, which will dramatically improve AI learning and training performance, surpassing the rest of the Phaidon H100 and challenging the H200 gas pedal.

Ji-Pu's point of view.

Intel's new AI products Intel Core Ultra and 5th generation Xeon processors, emphasizing that AI solutions are everywhere, from the data center, cloud, network, edge, to the PC, expanding the entire product line. 2024 AI gas pedal market is highly competitive, such as Fidelity's H100 and AMD's upcoming MI300X, and now with the addition of the Gaudi3, Intel stressed that Gaudi3 is expected to capture more of the market in 2024, due to its continued growth and proven performance advantages, as well as its highly competitive TCO (Total Cost of Ownership) and pricing advantages. Intel emphasizes that Gaudi3 is expected to capture more market share in 2024 due to its growing and proven performance advantages, as well as competitive TCO and pricing.

Dec W1｜Eckel to set up $2 billion advanced packaging and testing plant in the U.S.

Apple and Amkor, a major packaging and testing company, announced the expansion of their partnership on 30th of March, U.S. time. Amkor announced that it plans to invest US$2 billion (approximately NT$62.86 billion) in a new advanced packaging and testing plant in Arizona, U.S., to package and test the Apple wafers produced by TSMC's (2330) new plant in Arizona. In addition to Apple's strong support, TSMC's President Chieh-Chia Wei also praised the project, emphasizing the partnership between the two companies over the years.

Eckel has acquired approximately 55 acres of land in the area and will create an advanced manufacturing campus with over 500,000 square feet of space for clean rooms. The first phase of the plant is targeted for production in the next two to three years. Upon completion, the new facility is expected to be the largest advanced packaging and testing facility in the U.S. The facility is planned to provide U.S. volume production leadership for advanced packaging and testing of semiconductors to support key markets such as HPC, automotive and communications.

As the largest advanced packaging and testing company in the U.S., Eckel is pleased to lead the U.S. in the advancement of advanced packaging capabilities as the U.S. semiconductor supply chain expands," said Giel Rutten, president and chief executive officer of Eckel.

TSMC President Chieh-Chia Wei said that Eckel, as an OSAT facility, has been working with TSMC for many years. TSMC congratulates Eckel on building an advanced packaging and testing facility in Arizona, U.S. Eckel's new facility will bring value to TSMC, its customers, and the ecosystem.

Ji-Pu's point of view.

The company is the only U.S.-based semiconductor backend outsourcing assembly and testing (OSAT) facility, and the new facility, when completed, will be the largest semiconductor outsourcing advanced packaging facility in the U.S.," said Eckel. Eckel continues to layout the global, including the new plant in Vietnam has been formally operated in October, and is located in Bac Ninh Province (Bac Ninh), the main layout of advanced system-in-package (SiP) and memory packaging production lines, as well as part of the test program, the advanced packaging accounted for the total revenue of 73~76%, the main customers are Apple, Qualcomm, Phillip, Broadcom, Kaiser and so on. Senator Mark Kelly of Arizona also said that this is a major step in the microchip supply chain to reduce dependence on foreign countries. The $2 billion investment program will create good-paying jobs, strengthen the local economy and help improve national security.

Source OMDIA

Nov W5｜Next-generation Intel processor to be mass-produced in 3nm by TSMC

Intel's next-generation Lunar Lake MX processors feature a new microarchitecture redesigned to deliver breakthrough performance-per-watt efficiencies targeted at mobile devices. the Lunar Lake MX processors feature up to 8 general purpose cores (4 high-performance Lion Cove cores + 4 Skymont performance cores), a 12MB cache, up to 8 Xe2 GPU clusters, and up to 8 general purpose cores. 8 general purpose cores. The operating module is based on TSMC's 3nm N3B process. Intel's Lunar Lake CPUs are based on Intel's 18A process.

The Lunar Lake MX processor retains Intel's multi-chip Foveros 3D interconnect design. To reduce the physical footprint, Intel plans to consolidate the CPU, GPU, and memory controller on the same module and place the other modules in the SoC block. Lunar Lake MX is designed for laptops and is available with either 16GB or 32GB of LPDDR5X-8533 on-package memory to reduce platform footprint and improve performance. In addition, Lunar Lake MX saves 100~250mm² of space compared to typical designs with CPU off-package memory.

TSMC has previously pointed out that 3nm process technology is the most advanced technology in the industry in terms of PPA (performance, power consumption and area) and transistor technology. Benefiting from the strong demand for high-performance computing (HPC) and smartphone-related applications, 3nm will account for a mid-single digit percentage of the company's revenue this year, and the proportion will be even higher next year.

Ji-Pu's point of view.

This is the first time that Intel's flagship CPU uses a third-party process, and it is planned that the graphics, high-speed I/O, and other related chips other than Core will be outsourced to TSMC. Currently, it is rumored that part of the Arrow lake will be made by TSMC's N3E process, while the lunar lake will be made by TSMC's own 18A process. It is expected that TSMC will start mass production in the first half of next year to prepare for the PC recovery in the second half of next year. If TSMC wins this order, Intel will become one of TSMC's top three customers and the second largest 3nm foundry customer in 2025.

Nov W4｜South Korea's memory chip exports grow for the first time in more than a year

According to the latest data released by South Korea's Ministry of Industry, Trade and Resources (MITR) on November 14th, memory chip exports increased by 1% in October, a big improvement over September's annual decline of 18%. Among them, Multi-Chip Package (MCP) led the export rebound with a growth of 12.2%, and the annual decline rate of DRAM wafer exports also shrunk to a single digit, making this export rebound the first growth in 16 months.

South Korea's central bank forecasts that if trade continues to improve, the economy is expected to grow by 2.2% next year, better than this year's estimate of 1.4%. This export growth can also be seen in South Korea's semiconductor industry will lead the export growth, supporting economic momentum.

In addition to memory chips, the annual growth rate of display exports in October increased to 13.1% from 1% in September, and although cell phone exports still decreased by 3.3%, the rate of decrease was smaller than that of 5.2% in September.

Ji-Pu's point of view.

Semiconductors are the backbone of South Korea's economy, and the overall growth in exports in October has become a positive sign of growth prospects and recovery in global economic demand. Currently, the market is observing that the average price of DRAM and NAND Flash has started to rise since the fourth quarter. In terms of DRAM, it is estimated that the contract price will increase by about 3~8% in the fourth quarter, and whether or not this wave of increase can continue depends on whether or not suppliers continue to adhere to the production reduction strategy and the degree of actual demand warming up, of which the most critical is the general-purpose server field. Among them, PC DRAM and Sever DRAM contract price is estimated to increase by 3~8% quarterly; Mobile DRAM may have more room to increase compared to other applications to return to a healthy level earlier.

Nov W3｜Intel 18A will enter trial production in Q1 next year.

Intel CEO Jishengger came to Taiwan on the 7th for nearly 40 minutes of speeches, advanced process progress as the focus, Jishengger a breath to show Intel 7 since the five technology nodes technology products for Meteor Lake chip Intel 4, this year into mass production, Intel 20A, scheduled for 2024 mass production on the line, in addition to the same as the TSMC 2nm Intel 18A, which is the equivalent of TSMC's 2nm Intel 18A, has already finalized its design.

Intel's CEO, Mr. Jischinger, also said that the 18A has now been designed and is moving toward production. And will complete the production phase of five process nodes in four years, the development team is on track. Next-generation server, PC processors with Intel 3 process is in the "debugging" stage, next year can be put into production.

In addition to the four-year goal of advancing 5G process technology, Kissinger also showed the first glass substrate. Kissinger also showed the first glass substrate, which integrates optical components into packaging and has the advantages of low power consumption and low latency, and will be the product after the 18A process, and the layout of advanced packaging, Intel is firing on all cylinders. Intel hopes to regain its leadership position in semiconductor production by 2025. 18A process will not only be used by the company, but also for Ericsson and other customers to OEM chips.

Ji-Pu's point of view.

In addition to advanced processes, Kissinger also emphasized Intel's advanced packaging footprint, which, coupled with Intel's future development of small-chip technology, will continue to meet future market demands. The last of the five processes, Intel 18A, is a very difficult goal to achieve, but Intel is getting closer. 18A is rolling out test wafers and is on track quickly with finalized, reliable designs. Kissinger wants to innovate and build ecosystems, and change everyone's life through technology. But whether Intel can return to glory depends on the acceptance of the end market.

Nov W2｜AMD's financial report is good, next year AI is the key point.

AMD announced its Q3 2023 financial results on October 31st. Driven by the recovery of terminal demand and server processor demand, AMD's Q3 revenue and earnings both showed significant growth. In addition, the previously announced MI300 will be delivered to customers in Q4 as scheduled, and is expected to provide $400 million in revenue in Q4 and contribute about $2 billion next year.

Q3 Key Figures.

Revenue: $5.8B, up 4%, better than expected $5.7B;

Net Income: $299 million, up 3,53%;

Adjusted Earnings Per Share (EPS) : $0.70, also better than the expected $0.68.

However, AMD announced its Q4 revenue estimate of between $5.8-$6.4 billion, which, based on the median of $6.1 billion, implies that AMD's Q4 revenue will increase by about 9% annually and 5% quarterly; however, the $6.1 billion revenue estimate was not as good as the market's expectation of $6.37 billion, resulting in a drop of AMD's stock price by about 4% after the trading session.

AMD's data center business reported revenue of approximately $1.6 billion for the quarter, roughly unchanged from the same period last year but up from Q2. AMD said the data center business grew from Q2 due to accelerated customer adoption of fourth-generation AMD EPYC CPUs in the quarter, and the MI300A and MI300X GPUs are expected to be in volume production in the fourth quarter to support the deployment of multiple HPC, cloud, and AI customers. The MI300A and MI300X GPUs are expected to reach mass production in Q4 to support the deployment of many HPC, cloud and AI customers.

AMD Client revenue increased 421 TP3T to $1.5 billion, primarily due to growth in Ryzen mobile processor sales. Gaming revenue decreased 81 TP3T to $1.5 billion, primarily due to a decline in semi-customization revenue, partially offset by growth in AMD Radeon GPU sales. Finally, Embedded segment revenue decreased 51 TP3T to $1.2 billion, primarily due to lower communications revenue.

Ji-Pu's point of view.

AMD's upcoming AI chips, MI300A and MI300X, are on track for volume production this quarter, and the MI300 series is expected to be the fastest in AMD's history to reach the $1 billion in sales milestone. The data center business is poised for strong growth in the fourth quarter, and the client business will maintain its growth momentum.

Nov W1｜Qualcomm unveils AI chips for PCs and mobile phones, setting the stage for new competition

Chip giant Qualcomm recently hosted the Snapdragon Summit 2023 to unveil two new chipsets, including Snapdragon X Elite, Qualcomm's most powerful PC chip for AI applications, and its latest 5G flagship handset chip.

The Snapdragon X Elite is a new in-house CPU architecture, code-named Oryon. The X Elite contains 12 Oryon cores, rather than the usual patchwork of small and large cores. The 12 cores can reach as fast as 3.8GHz when fully powered up, though the 1st and 2nd cores can go as high as 4.3GHz. Qualcomm claims twice the performance of its rivals, the Intel i7-1355U and i7-1360P, while consuming 68% less power, outperforming the Apple M2 Max in terms of single-threadedness, and processing more than 13 billion parameters. generative AI models.

Ziad Asghar, Qualcomm's vice president of product management, said on 26th of July, US time, that the AI phone replacement wave is happening now; Kedar Kondap, Qualcomm's senior vice president and general manager of the Computing and Gaming Division, also mentioned that AI PCs are expected to take off in 2024, and will be further popularized in 2025. AI PCs do not need to be on the cloud to perform AI computing, which has the benefit of reducing communication latency and improving data security. The benefits of AI PCs are reduced communication latency and increased data security. Qualcomm expects the first Snapdragon X Elite PCs to be available in mid-2024. Microsoft CEO Satya Nadella supported Qualcomm in a video, praising the Snapdragon Elite X for driving a new era of "AI PCs" for businesses and consumers.

Qualcomm also announced a new generation of Hexagon Neural Processing Units (NPUs) for AI and machine learning tasks that are more than three times faster than the Snapdragon 8cx Gen 3. Currently, Windows NPUs are primarily used to accelerate webcam background removal, but Windows will have more and more native AI capabilities in the future, increasing the need for local processing.

TIRIAS Research analyst Francis Sideco pointed out that with Adobe and other software companies launching solutions based on generative AI, users are becoming more and more accustomed to the use of AI to create and enhance work efficiency, the demand for PCs that natively support AI is bound to increase.

In addition, Qualcomm announced the new generation of flagship AI cell phone chip Snapdragon Series 8 Gen 3, which is Qualcomm's first mobile platform to integrate AI performance. Qualcomm said that early next year, ASUS, Sony, OnePlus and other brands will launch their flagship AI smartphones using Snapdragon Series 8 Gen 3.

Ji-Pu's point of view.

Qualcomm proposes the concept of hybrid AI, which can decide whether to process data by terminal or cloud AI, depending on the situation. Personal data can be processed by terminal AI to take care of information security issues, while for companies, uploading confidential documents to the cloud carries a certain degree of risk, which can now be avoided by processing in the terminal device. As for the advantages and disadvantages of AI development, the industry, government and academia continue to discuss, there are also concerns, Ziad Asghar said that before observing the AI is only to organize the data generated as the main, but also unable to actively think. The benefits of AI seem to outweigh the drawbacks, as it can save people's time or be used for educational purposes.

Oct W3｜Canon develops NIL device to challenge EUV microfilm technology

Canon, a major Japanese semiconductor equipment maker, announced on October 13 that it has launched the FPA-1200NZ2C, a microfilming machine that adopts Nanoimprint Lithography (NIL) technology, which can be used to produce 5 nm wafers, and after modification, it can be used to produce 2 nm wafers in an attempt to challenge the position of ASML-US, the leading semiconductor equipment maker. The FPA-1200NZ2C can be used to produce 5 nm wafers and has been modified to produce 2 nm wafers, challenging the position of leading semiconductor equipment maker ASML-US.

Canon is one of the world's top three manufacturers of exposure micrographic equipment, and is a well-known, century-old camera manufacturer. ASML, a joint venture founded in 1984, is the only manufacturer in the world that can produce EUV (Extreme Ultraviolet) micrographics, a critical piece of equipment for advanced semiconductor processes up to 5 nanometers in size and priced at more than US$150 million.

When circuits are drawn on wafers, a device called an exposure device is generally used for phototechnology. Nanoimprinting is different from the previous exposure process, but similar to the principle of wax sealing, in which the surface is first coated with a photoresist (resin) and then pressed down with a well-made photomask mold to create the desired pattern. Because the molds for nanoimprinting can be repetitively imprinted, they are suitable for mass production of microstructures in optical disks, wafers and solar cells.

In addition, Canon has developed the best photoresist materials for nano-imprinted micrographics, as well as nanoscale alignment technology for imprinting, and realized the particle control technology for clean semiconductors. Compared with EUV technology, which has been commercialized, NIL technology consumes only one-tenth of the power consumption of EUV, and the amount of money is only 40% of that of EUV, which significantly reduces energy consumption and lowers the cost of equipment.

EUV technology is still the only choice for advanced processes, but Nanoimprinted Lithography (NIL) has the advantages of low cost and reduced carbon emissions, and is expected to become an alternative to EUV technology in the future if it can be successfully developed and mass-produced at a certain yield rate.

Ji-Pu's point of view.

Although Nanoimprint Imaging (NIL) can shorten the process time and achieve more power saving, NIL is not able to completely replace EUV microimaging because it requires the use of soft photoresist material, which is suitable for regular structures such as DRAM and NAND FLASH, while only a small portion of the logic chip structure can be used.

Oct W2｜Intel Spins Off PSG Division, Intends to Go Public in 3 Years

Intel announced on the 3rd that its Programmable Solutions Group (PSG) will operate independently from January 1 next year, which will allow the division to accelerate its business growth and increase its competitiveness in the field-programmable logic gate array (FPGA) industry.

Intel even said in a statement that it "intends to initiate an initial public offering (IPO) of PSG over the next two to three years and may explore private investor opportunities to accelerate the growth of the business, with Intel retaining a majority stake."

In a press release, the company noted that Sandra Rivera, head of data centers and chips, will be PSG's CEO, while Shannon Poulin, who was formerly PSG's general manager, will serve as chief operating officer. Last year, Intel also spun off its self-driving technology division Mobileye and sold its NAND memory and storage business to South Korea's SK Hynix.

Intel's decision to spin off its PSG division, which was created to develop Agilex, Stratix and other FPGA products immediately after its $16.7 billion acquisition of FPGA maker Altera back in 2015, is helping to unlock some of that value. Investors applauded the move, which sent Intel shares soaring nearly 3% to $36.57 per share after the bell.

FPGAs are programmable chips that can be adjusted in function or updated even if the chip is already installed in an electronic device. They are used in the design and development of communication hardware, data center equipment, and other chips. Although they are extremely flexible and powerful, they have always been difficult to program incrementally, making it difficult to adopt them in large numbers.

In the most recent quarter, Intel said the PSG division had record revenue, but did not provide a precise figure.Altera posted annual revenue of about $1.99 billion in 2019. Citing third-party estimates, Intel said the overall market for such chips will grow to $11.5 billion in 2027, up from $8 billion in 2023.

Ji-Pu's point of view.

The spin-off of the PSG division is the latest step in Intel's efforts to reorganize the company's business to focus on its core competencies of wafer design and wafer microprocessing. It is worth noting that Intel is not spinning off PSG because of poor performance or low-margin markets, but rather because it believes that the PSG division will perform better on its own and that its business agility will be enhanced. Intel CEO Kissinger is working hard to restore Intel's leadership in the chip industry, and many investors are now increasingly optimistic about the revitalization plan.

Oct W1｜Intel adopts EUV equipment for mass production of wafers in Ireland

Intel said on 29 September that it has begun high-volume production using extreme ultraviolet (EUV) lithography at its plant in Ireland, an important step in its efforts to catch up with its competitors. The move will have a direct impact on the competitive landscape of the semiconductor industry and could change the market position of Intel and Taiwan's integrated circuit makers. It also marks the first time EUV has been used in volume production in Europe, which will help the group reduce its dependence on US and Asian supplies.

Intel, once the world's leading chipmaker, has lost its lead but says it expects to regain it with manufacturing technology that it says will rival TSMC's best. And he aims to deliver five generations of technology within four years, with extreme ultraviolet (EUV) lithography playing a key role in achieving this.

Intel's Ireland facility, located in the Dublin suburb of Leixlip, is the group's first high-volume production site for the Intel 4 process, and is part of Intel's effort to advance to five process nodes in four years. It is part of Intel's effort to advance to five process nodes in four years. It is also a tangible demonstration of Intel's commitment to catching up with wafer technology leaders TSMC and Samsung. The process will use EUV technology to produce the future Meteor Lake laptop chips, paving the way for AI PCs.

Ji-Pu's point of view.

Intel claims that its Intel 3 will "enter production" in the second half of 2023. Its roadmap states that it aims to achieve "process leadership" at the 2nm node by around 2025. Ann Kelleher, Intel's general manager of technology development, said the company is working hard to achieve that goal, with two processes now complete, a third "coming on fast," and the last two well advanced. This marks the first time EUV has been used in volume production in Europe, a new milestone for Intel and the semiconductor industry as a whole.

Sep W4｜Intel Launches First Glass Substrate, Revolutionizing Packaging

Intel announced on Taiwan time on the 18th to launch the industry's first glass substrate for advanced packaging, breaking through the limitations of the existing traditional substrate, so that the semiconductor package transistor number limit maximization, and planning for mass production between 2026 and 2030, will be the first to lead into the need for a larger package, higher speed applications and workload of the field of data centers, AI, graphic processing and other areas.

Pooya Tadayon, Intel's director of assembly and testing, said glass is harder than organic materials and has a low coefficient of thermal expansion, unlike organic substrates, which expand and bend. These characteristics give glass substrates a great advantage in reducing the pitch of connecting wires for large-size packages. He also mentioned that the use of glass materials can achieve some interesting properties, can improve the efficiency of the wafer power supply, and enable the connection bandwidth from 224 G to 448 G. As manufacturing processes develop and demand changes, glass substrates will gradually emerge and coexist with organic substrates, rather than replace them.

According to industry analysis, Intel's next-generation glass substrate advanced packaging solutions will provide larger and more efficient packaging services. Intel's move will revolutionize the world of semiconductor packaging, competing with professional packaging and testing vendors such as Sun Micron and Eckel. Intel emphasizes that its latest glass substrate advanced packaging technology will continue to push Moore's Law forward as it strives to meet its goal of accommodating 1 trillion transistors in a single package by 2030.

Although it did not disclose the names of the supply chains it has worked with, Intel said it has invested in research and development related to glass substrates and has worked closely with materials and equipment manufacturers in the hope of building a relevant ecosystem.

Ji-Pu's point of view.

With the increased demand for more powerful computing and the arrival of the semiconductor industry into the heterostructure concept of using multiple chiplets in a single package, glass substrates are a practical and indispensable revolution for next-generation semiconductors. It breaks through the limitations of traditional substrates and allows more transistors to be connected in a single package, assembling larger chiplet composites, or system level packaging, which allows for greater flexibility and lower overall cost and power consumption, realizing performance and increasing transistor density. Although Intel is a technology leader, the next step in successful commercialization, including customer variety and order intake, will be critical in the future.

Sep W3｜Silicon photonic chips will be the next wave of AI.

Silicon photonics technology has been a hot topic lately, and according to media reports, TSMC is aiming at the upcoming silicon photonics ultra-high-speed chip business opportunity next year. Silicon photonics technology is based on the combination of photons and electrons in silicon material, which can realize high-speed, high-bandwidth optical communication and optoelectronics functions on a chip. The advantages of this technology include high speed, high integration, low power consumption and low cost. Therefore, silicon photonics is regarded as one of the key technologies to solve the bottlenecks faced by data centers and high-performance computing systems.

In terms of silicon photonics, TSMC (2330) and Sun Micron (3711) are best positioned to benefit from this trend, according to major domestic investment advisors. TSMC offers wafer fabrication and advanced packaging integration solutions for PICs and EICs, while Sun Micron already has integrated optical solutions on its VIPack platform. TSMC (2330) is optimistic about the development of silicon photonics technology. Recently, at the SEMICON show in Taiwan, TSMC introduced its silicon photonics platform solution "EPIC". Given the increase in energy efficiency and the need for higher transmission speeds, there is a growing need for integrated solutions for electronic and photonic sources in large-scale data centers.

However, silicon photonics is not an easy technology, and it is facing several key bottlenecks. Yen-Cheung Fang, director of ITRI's Institute of Electronic and Optoelectronic Systems, said, first, the high manufacturing cost of silicon photonics chips limits the promotion of silicon photonics in large-scale applications, so cost reduction is an urgent need to solve the problem; in addition, the performance and stability of silicon photonics components still need to be further improved, and need to be focused on the design, manufacturing and testing in-depth research to ensure long-term stable operation and high performance.

Silicon optical chips have clearly become an area of intense investment for the semiconductor industry, not only for Intel, Cisco and IBM, but also for Pfizer, which has acquired Mellanox Technologies, a multinational supplier of computer networking products, for $6.9 billion in cash in 2021. Many tech companies are eyeing potential uses for silicon optical chips in a variety of areas such as data centers, supercomputers and networking equipment, driverless cars and national defense radar systems, among others. Many technology companies are focusing on its potential use in areas as diverse as data centers, supercomputers and networking equipment, driverless vehicles and national defense lightning systems.

Ji-Pu's point of view.

Silicon photonics industry is a major trend with huge market potential for its application scenarios. It is a challenge to realize efficient light source integration on silicon photonic wafers. The light source, as a key component, needs to be efficiently emitted and detected on silicon materials. Better integration of emission and reception is the key to improving the performance of silicon photonic systems, and effective ways need to be found to integrate the light source and detector on a single silicon wafer. These challenges still require technological breakthroughs and continuous research. Substantial progress may be possible within a few years and will drive the realization of wider applications of silicon photonics technology.

Sep W2｜Huawei launches new handset with Kirin 9000S chip

Huawei's latest smartphone, the Mate 60 Pro, is powered by the Kirin 9000S processor chip from Chinese foundry SMIC. After disassembling the chip, TechInsights, a reverse engineering and disassembly company, confirmed that it was built on China's SMIC N+2 process, and that its performance is close to that of TSMC's 7nm processor. However, the semiconductor industry has pointed out that SMIC's lack of extreme ultraviolet (EUV) equipment means that its performance and wafer density may only be slightly better than that of TSMC's 10-nanometer process, and that yields may not be as good.

Although Huawei's return to the market has been low-profile, the Kirin 9000S mobile application processor (AP) has been the subject of a lot of discussion in the market. According to various media reports, the Kirin 9000S is manufactured by SMIC, and although the manufacturing process has not been announced, the N+1 process may be equivalent to 7nm LPE process, and the N+2 process may be equivalent to 7nm LPP process, which is the first chip to utilize SMIC's most advanced 7nm manufacturing process technology.

Semiconductor industry sources pointed out that SMIC's Kirin 9000S for Huawei only uses a deep ultraviolet (DUV) exposure machine, which is not incapable of creating a 7nm process wafer, requiring multiple rounds of exposure, and I'm afraid that production yields will not be able to reach commercial standards. It is estimated that SMIC's Kirin 9000S chip may only reach a level slightly better than TSMC's 10nm process standard, and the production yield still needs to be improved.

TSMC (2330) built its first-generation 7nm process products with DUV exposures back in 2018, but due to technical limitations of the machine, TSMC quickly pushed forward with EUV exposures to produce the second-generation 7nm, and then the EUV exposures upgraded the software tooling and hardware with ASML's technical support, allowing TSMC to push all the way to the current 3nm process with the EUV exposures. The EUV Exposure machine was then upgraded with software tools and hardware through ASML technical support.

Reuters reports that China will launch a new state-backed investment fund aimed at raising about $40 billion for its chip industry as it ramps up efforts to catch up with the U.S. and other competitors. However, with SMIC supplying chips to Huawei, the key to mass production is actually yield, so there are ongoing concerns about Huawei's ability to maintain production scale and profitability.

Ji-Pu's point of view.

Compared to the previous generation of Kirin 9000, which used TSMC's 5nm process node with one large core (clocked at 3.13GHz), three medium cores (clocked at 2.54GHz), and four small cores (clocked at 2.05GHz), totaling a chip area of about 106mm^2, the current generation of Kirin 9000 utilizes SMIC's advanced process node, which includes one large core (clocked at 2.62GHz), three medium cores (clocked at 2.15GHz), and four small cores (clocked at 1.53GHz), totaling about 140mm^2 of wafer area. The Kirin 9000 uses SMIC's advanced process nodes, including one large core (clocked at 2.62GHz), three medium cores (clocked at 2.15GHz), and four small cores (clocked at 1.53GHz), with a total wafer area of about 140mm^2. The clock is lowered, and the wafer is enlarged.

Theoretically, the physical limit of DUV will be at 5nm, which requires the use of technologies such as 3P/3E and auto-alignment, but I am afraid that the yield and performance will be quite low. Industry insiders have pointed out that it will be difficult for SMIC to mass-produce 7nm process before it has access to an EUV machine. However, this also means that China's semiconductor industry has been able to make technological advances without an extreme ultraviolet (EUV) lithography machine, and the difficulty of this achievement underscores the resilience of the country's wafer technology capabilities. But we have to admire SMIC's ability to forge a path through the layers of U.S. blockade.

Sep W1｜Intel Turns to TSMC, Outsourcing Up to $19 Billion

Intel wanted to utilize TSMC's capabilities in order to solve its own manufacturing shortcomings.

In its latest report on TSMC (2330), Goldman Sachs Securities pointed out that as Intel continues to expand its share of outsourcing in the face of process upgrades since the start of its 10-nanometer process, the total demand for outsourcing is expected to rise to $19.4 billion in 2025, making TSMC the largest beneficiary.

With Intel's manufacturing processes being surpassed by TSMC, Intel has continued to expand its outsourcing orders to TSMC in recent years in order to achieve the optimal balance between cost, performance and capacity, such as adopting TSMC's 5nm process for the 2021 Ponte Vecchio, and announcing last August that it would outsource this year's Meteor Lake GPUs to TSMC, among others.

The market is highly concerned about the development trend of Intel outsourcing orders to TSMC, Goldman Sachs semiconductor industry analyst Lu Kunlin pointed out that although Intel has always been a customer of TSMC's small-scale projects, and the core product CPU is still mostly manufactured by its own production, but after it faces the challenge of the 10-nanometer upgrades, the situation is beginning to change.

Quincy Lu conducted a sensitivity analysis on the trend of Intel outsourcing orders to TSMC. According to the analysis results, it is estimated that the potential TAM of Intel outsourcing market will reach $18.6 billion in 2024 and $19.4 billion in 2025.

Because of the difficulty of technology upgrades, and because wafer manufacturing and cost management have become challenges for Intel, on June 21, 2023, Intel decided to create a "foundry-like" relationship between manufacturing and business units to give business units more flexibility in utilizing third-party foundries to optimize operating expenses and investment decisions.

Ji-Pu's point of view.

Starting out as a process manufacturer, Intel's technology is not to be underestimated, and its performance and power consumption have always been very competitive. However, from 14nm to 10nm, Intel has deviated from the goal of self-sufficiency, but the cooperation with TSMC can realize the best balance between cost, performance and capacity. At present, Intel's SoC, such as Meteor Lake, Raptor Lake, and many other products in the IC, are delivered to TSMC production manufacturing, which to TSMC production IC not only IO, WiFi and even GPU, CORE and other core key chip. In the future, Intel is likely to outsource its key IC production to TSMC in order to better control costs and improve efficiency. This is mainly due to the sales situation of the end products and the speed of ramp up of the yield rate of both parties. As a result, orders from Intel's core CPU outsourcing are expected to contribute significantly to TSMC's revenue growth from 2025 onwards, and we cannot rule out the possibility that Intel may become the second-largest customer, so we continue to be optimistic about TSMC's future development.

Aug W4｜ARM rides on the AI train with an IPO

ARM, the British chip design company owned by Softbank, filed an application with the U.S. Securities and Exchange Commission (SEC) on March 21, planning to list its shares on the NASDAQ under the ticker symbol "ARM," with a preliminary stock release of between $30 and $70 billion. The move is expected to be the world's largest initial public offering (IPO) this year (2023) and the third-largest in tech history after Alibaba and Facebook.

ARM focuses on providing professional chip architecture design and related IP licenses. Headquartered in the U.K., the company's business model is to license its designs to partners in a variety of fields, with approximately 70% products worldwide adopting instruction-set architectures or design ICs developed by ARM, including smartphones, tablets, automotive assisted driving systems and data centers.

ARM was acquired by Japan's Software Bank in 2016 for $32 billion, however, in 2020, ARM was originally planned to be acquired by Phaidon (NVIDIA) for $40 billion, but the deal faced opposition from the UK, the US, and the EU, and these opposing voices were concerned that once ARM lost its neutrality, it could have a serious impact on competition in the marketplace, and so the deal was The deal was shelved, and ARM later decided to move in the direction of an initial public offering.

This year, due to the AI trend set off a wave of global technology industry, technology stock prices climbed all the way up, also created Phaidon as the first U.S. chip maker with a market value exceeding one trillion U.S. dollars. As a semiconductor patent factory, ARM hopes to take advantage of the AI trend to achieve higher valuation in order to satisfy the expectations of its parent company, SoftBank Group.

ARM filed a registration statement on Form F-1, but did not disclose the company's valuation, IPO pricing or fundraising size. The filing disclosed that ARM hired 28 institutions to serve as underwriters for its first IPO this year, with Barclays, Goldman Sachs, JPMorgan Chase and Mizuho Financial Group leading the way, and Morgan Stanley conspicuously absent.

The company is a major player in the global semiconductor industry, with chips designed for tablets, smartphones, sensors, disk storage and servers; almost every smartphone and tablet uses ARM technology. In May of this year, ARM launched a new machine learning model chipset architecture Cortex-X4 core and a GPU called G720, in addition, the company also expanded the open source partnership, and to promote the development of the GPU ecosystem based on it.

Ji-Pu's point of view.

ARM's IPO plan is not only to raise funds, but also to improve the company's public image and market recognition. It is expected that ARM will raise public funds in the form of IC Design, and it is expected to aim at the high computing and automotive fields as the main target markets. For ARM, the maturity of the architecture will increase as more and more people participate in the construction. And existing customers are joining the RISC-V camp, although the short-term impact is not significant, but in the long term, it will still have a considerable impact on ARM's layout, and it is worth paying attention to how ARM is going to overcome these challenges.

Aug W3｜Qualcomm rumored to lay off 200 people, and those who stay will have to accept bonus reduction and no salary adjustment.

Cell phone chip maker Qualcomm has been rumored to start a large-scale manpower streamlining plan for its Taiwan branch, adopting voluntary departure and dismissal in parallel, and is expected to lay off 200 people in October, which is about more than 10% of the total number of Qualcomm employees in Taiwan, and at the same time, it will also implement policies such as reducing bonuses, not adjusting salaries, and discounting dividends by 70% for the employees who stay behind, which shows that demand for the cell phone market is sluggish, and that large manufacturers are actively opening up new sources of income and cutting costs.

Qualcomm has previously announced that in response to the poor market conditions, the U.S. headquarters in San Diego was downsized in June, with a total reduction of approximately 415 employees at the headquarters. It has also been rumored that Qualcomm has submitted a list of new layoffs to the local government, but the number of layoffs has not been disclosed. Qualcomm CEO Amon (Amon) recently pointed out in the financial meeting, Qualcomm has been this year through a variety of cost-saving measures to save 5% of the cost.

According to Qualcomm's announcement of last quarter's financial results, last quarter's consolidated revenue was approximately US$8.540 billion, a 23% annual decrease, of which handset chip revenue decreased by 25% to US$5.26 billion. The only area that still showed year-on-year growth was automotive chips, which grew 13% year-on-year to US$434 million.

The San Diego Union-Tribune (San Diego Union-Tribune) pointed out that most of Qualcomm's business comes from Android (Android) devices, including smartphones and 5G applications, China is an important market for Android devices, the region's weak sales of smartphones, but also on Qualcomm caused considerable pressure. This quarter smartphone market shipments will be at least a single-digit percentage decrease (7 to 9%), the estimated quarterly revenue will be between $8.1 billion to $8.9 billion, the median is much lower than the market analysts' expectations, the estimated EPS of $ 1.8 to $ 2, the same is lower than the market's average valuation of 1.94 dollars.

Taiwan Economic Research Institute Industrial Economics Database Director Liu Pei-jin said, "I think on the one hand, in addition to reflecting the United States to China, in the semiconductor-related some of the controls, the most important is still in the current global semiconductor industry, a boom, then still persistent, and did not see a more pronounced rebound.

Ji-Pu's point of view.

The generally weak demand for consumer electronics such as automobiles, home appliances, mobile phones and tablet PCs, coupled with the fact that inventories have not been fully digested, has made brand owners pessimistic about the market outlook and conservative in placing orders. After the layoffs, the oncoming challenges are price cuts by suppliers and a reduction in purchases. The global market downturn has led to low demand for notebook computers and cell phones, etc., and both the upstream and downstream supply chains may be impacted.

Aug W2｜Tokyo Mighty: China buys mature process wafer equipment

Bloomberg (11th) reported that Asia's largest semiconductor equipment maker, Tokyo Elektronics Limited (TEL), pointed out that as the U.S. and its allies tighten export controls on cutting-edge technologies, Chinese chipmakers are firing on all cylinders, stepping up their purchases of mature process equipment.

Toshiki Kawai, president of Tokyo-based VeriTech, said on an earnings call on Oct. 10 that he had seen "very strong investment" in China and that "this is not just a short-term trend for this year," but that "we expect the demand to continue.

The company noted that the surge in demand for mature process equipment will help make up for delayed investments by high-end logic chipmakers and foundries. China's contribution to the company's second-quarter revenue amounted to $39%.

Tokyo-based VeriTech is a key link in the wafer manufacturing supply chain, providing machinery and equipment for the production of advanced semiconductors by wafer makers such as TSMC, Samsung Electronics and Intel. Strong investment momentum is expected in automotive and industrial applications, a trend seen in the last fiscal year.

Tokyo VeriTech's Q2 results fell by 17% due to the decline in demand in the global electronics market, but full-year revenues are still expected to reach 1.7 trillion yen, with an operating profit of 82.4 billion yen, slightly better than analysts' expectations.

Hiroshi Kawamoto, head of Tokyo-based VeriTech's finance department, said customers from China are well aware of the restrictions and have already redrawn their strategies. The company said Japan's new measures on wafer equipment exports, which went into effect last month, have had no impact on its operations and sales.

Regarding the demand for generative AI-related equipment, Hiroshi Kawamoto said, "We have received a lot of inquiries about AI-related investments, and although the scale is still small, AI should contribute to the revenue in the next fiscal year".

Ji-Pu's point of view.

After the U.S. tightened restrictions on China's advanced chips, China accelerated the expansion of mature processes above 28nm. It is speculated that China wants to take advantage of its dominant market position to gain the upper hand. This could raise concerns in the U.S. that Western companies are relying on China's traditional chip supply, which in turn poses a national security risk.

Aug W1｜Taiwan TSMC to expand into Europe and set up a plant in Germany.

Foundry leader TSMC finalized its investment in Germany yesterday! TSMC, together with Bosch, Infineon, and NXP, announced that they will establish a joint venture in Dresden, Germany to build a European Semiconductor Manufacturing Company (ESMC), with a total investment estimated to be more than 10 billion euros (about 353 billion Taiwan dollars).

And the German government will provide up to 5 billion euros in subsidies. TSMC is expected to take a 70% stake in the joint venture, with each of the three major European plants taking a 10% stake, and TSMC will be responsible for its operation.

TSMC's Board of Directors resolved to invest up to €3,499.93 million in its German subsidiary (ESMC) for a major shareholding.

To provide professional integrated circuit manufacturing services.

The German plant is planned in accordance with the framework of the European Chips Act. The future fab will utilize TSMC's 28/22nm planar complementary metal oxide semiconductor (CMOS) and 16/12nm finned field-effect transistor (FinFET) process technologies for production, with a monthly capacity of approximately 40,000 wafers. The plant is expected to be built in the second half of 2024 and begin production by the end of 2027, creating approximately 2,000 direct high-tech professional jobs.

TSMC's investment in Dresden demonstrates TSMC's commitment to meeting our customers' strategic capabilities, technology needs, and deepens our long-term partnerships with Bosch, Infineon, and NXP Semiconductors," said TSMC President Chieh-Chia Wei.

Ji-Pu's point of view.

Dresden, the capital of the German state of Saxony, is an important cultural, political and economic center in eastern Germany and the largest semiconductor cluster in Europe. Europe is a promising place for semiconductor innovation, especially in the automotive and industrial sectors. This will help Germany become the European capital for semiconductor production, reducing its dependence on the Asian supply chain, while at the same time mitigating the damage caused by wafer shortages to local automakers and providing the innovation and capacity needed. For TSMC, the likely number one competitor (Intel) has already signed a contract with the German government to receive a large grant. Under the assumption that TSMC wants to gain access to the European automotive market, TSMC has to go there. It is expected that the process nodes in the past will be 28~1X nm, and there are even rumors that more advanced processes will be set up in Germany in the future. In terms of product analysis, the major IDM factories in Europe do not contain 28/22nm (eMemory), so basically it is very reasonable to set up 28/22nm factories in the past, but the rumor of providing 1x or more advanced processes is relatively unexpected. If true, it is estimated that it may be aimed at a possible competitor (GlobalFoundries) in the European automotive market, which is not available.

Jul W4｜Heterogeneous Integration Breakthrough Stressor Support for IC Packaging

Applied Materials introduces new materials and systems that provide wafer makers with a major breakthrough in hybrid bonding and TSV technologies to help integrate small wafers into advanced 2.5D and 3D packages. TSV is currently the process bottleneck for TSMC's advanced packaging CoWoS. The ASM solution extends the industry lead in heterogeneous integration, enabling 2.5/3D 3D packaging to continue Moore's Law interconnect technology, which allows multiple logic wafers to evolve into three-dimensional packages through TSV silicon perforation. Currently, the bottleneck in TSMC's advanced packaging CoWoS process lies in the TSV technology. Instead of wire bonding, TSV perforated wafers are stacked by drilling holes in each logic wafer and filling them with metal from the bottom so that they can pass through each layer of the wafer. The holes are then filled with conductive materials such as copper, polycrystalline silicon, tungsten, etc. to form a connection, and finally, the wafers or dies are thinned and stacked and bonded as a three-dimensional stacking technology for inter-chip telecommunication signals. As IC designers continue to integrate more logic, memory, and specialty feature chips into advanced 2.5D and 3D packages, the number of TSV interconnect leads per package is expanding into the thousands. In order to integrate more interconnects and accommodate higher wafer stacks, silicon perforations need to be made narrower and higher, resulting in changes in deposition uniformity, which reduces performance and increases resistance and power consumption. As the largest supplier of heterogeneous integration technology, Inductotherm provides wafer fabrication systems including etching, physical/chemical vapor deposition, annealing and surface treatment equipment. We help semiconductor companies combine small wafers of various functions, technology nodes and sizes into advanced packages so that the combined whole can operate as a single product. Heterogeneous integration brings new solutions and opportunities to the industry as the demand for transistors for applications such as high-performance computing and artificial intelligence grows rapidly, making traditional 2D miniaturization expensive and slow.

Ji-Pu Viewpoint:Through these innovations, wafer manufacturers such as TSMC are able to realize higher performance wafer designs in advanced packaging. Coupled with the fact that Inductotherm emphasizes that the new system launched this time has a significant competitive advantage in the heterogeneous integration field, it will widen the gap between Inductotherm and its peers in the field of heterogeneous integration and satisfy the growing market demand.

Jul W3｜Intel to Attack China's AI Market: Sunrise, Unimicron, Sentec and Other Taiwan Chains to Complement

Intel has launched a downscale version of its AI gas pedal "Habana Gaudi2" specifically designed for the mainland market. The downscale version of Habana Gaudi2, which meets U.S. specifications, is competing with rival Pheonix for the mainland AI chip market, and has already been adopted by mainland server giants such as Longchamp Group, Xinhualan, and Hyperfocus, which has once again heated up the topic of AI. "The downscale version of the Habana Gaudi2 chip is produced in 7nm. The industry is optimistic that the adoption of Intel's AI chips in China by local server giants will help Sun Microsystems to expand its AI market in China.(3711)In addition, the company has a strong supply chain, such as the Wave, Unimicron, and an aftermarket, such as Sentec, Wave's main server foundry.

Ji-Pu Viewpoint:Due to the U.S. ban, chipmakers are unable to sell high-performance processors above 600 gigabits per second (GB/s) to mainland China. To cater to the Chinese market, NVIDIA launched a downgraded version of the A100 processor called "A800", followed by a downgraded version of the H100 processor called "H800". Intel is now following this trend by launching a downgraded version of its AI chipset specifically designed for the Chinese market. This demonstrates the importance that both companies place on the China market, and their desire to lower the barriers to entry and provide more hardware choices to enhance their customers' ability to deploy AI in the cloud and smart edge technologies, while also benefiting Taiwan's related supply chain.

Jul W2｜Inventory Adjustment: IC Design Firms Seek Quarterly Increase in Results

After last year's4The season is the same as the first quarter of the year.1After adjusting for seasonal inventory, someICDesign Factory No.2The revenue performance of these companies is expected to continue to increase this quarter due to the stabilization of the new products and the pulling force of goods.

Table 1: Second Quarter and Year-ago Revenues by Company

Source: Financial reports of various companies; collated by Ji-Pu Industrial Trend Research Institute.

Ji-Pu Viewpoint

Mauda's Q2 revenue was NT$1.8 billion, up nearly 20% quarter-over-quarter, and has been on the rise for two consecutive quarters. The demand for DDR5 is expected to rise gradually from the end of this year to next year.
Sensor IC makers' Q2 revenue was NT$9.4 billion, up nearly 30% quarter-over-quarter, and rebounded for the second consecutive quarter. Mouse and e-sports applications drove the performance, and market demand is gradually stabilizing.
Sitronix's Q2 revenue of NT$9.9 billion increased by over 20% QoQ, ending the sixth consecutive quarterly decline. Industrial control and automotive applications performed steadily, and customer shipments are expected to start in August.
Power management IC maker Hong Guan Dian first-half revenue of 700 million yuan, an annual increase of more than 30%. Main focus on AC-DC high-voltage power management solutions, gaming and game machine applications to drive growth.
Revenue of Guangzhou Silitech increased slightly to $5.1 billion in the second quarter. Continuing to launch new products, the result is expected to be flat to slightly increase.

Jul W1｜U.S.-Dutch Joint Effort Against Mainland Chips

The U.S., Japan, and the Netherlands have made a heavy blow to China's semiconductor industry, the Dutch government yesterday offered to limit the export of some advanced semiconductor manufacturing equipment new rules, requiring the future of ASML (ASML) to export the DUV deep ultraviolet micrographics equipment to mainland China, to apply for prior authorization, the new rules effective since September 1st. Domestic semiconductor manufacturers said that due to the DUV machine in the past two years facing a shortage of chips, the foundries to expand production capacity to frantically scramble for goods, the shortage has not been relieved, and it is expected that this control order, the fear of China's foundries to expand the wave of acquisitions. However, in the future, China's wafer upgrading is bound to encounter resistance, wafer design based on risk diversification, will accelerate the transfer of orders to Taiwan factories, TSMC, UMC, MPEG, the world, and even foreign companies Glorfonds, will benefit.

Ji-Pu Viewpoint:Since October 2022, the U.S. has implemented an export ban restricting the sale of U.S. wafer fabrication tools from U.S. equipment manufacturers such as Collin Research and Development and Applied Materials to Mainland China due to national security concerns. To avoid improper use of sensitive technologies, the US has also actively lobbied other countries with important suppliers to adopt similar measures. Subsequently, Japan joined in the effort by restricting the export of 23 types of semiconductor manufacturing equipment. The Dutch government's recent decision to impose controls, presumably in agreement with Eskimo, does not cut off all business to China. DUV deep ultraviolet micrographing equipment is a modification of the 193 wavelength exposure machine, which is not as good as EUV's ability to expose a fine line, but is now widely used in critical sites at the smallest 7 nm process nodes. This move is said to have tightened China's throat even more, and will undoubtedly add insult to injury for the development of advanced processes (below 16nm) in the next few years.

Jun W4｜Taiwan TSMC's operation will be stronger in the second half of the year.

TSMC's (2330) operation is expected to turn stronger in the second half of the year. The industry's assessment is still mainly based on the injection of new Apple-related applications, while non-Apple mobile phone applications are still weak, with limited contribution from rush orders and short-term orders. As for the popular high-end AI-related applications, it is difficult to assess TSMC's contribution in the short term, and it is expected to be more quantitatively calculated next year. Industry insiders point out that AI-related applications are still at an early stage. In fact, the largest contribution to TSMC's revenue still comes from high-speed computing (HPC) and smartphone-related fields, while AI is still part of high-speed computing in TSMC's categorized projects and cannot be independently calculated for the time being. According to TSMC's statistics, in the first quarter, revenue from HPC accounted for 44%, followed by smartphone-related revenue of 34%. After the inventory adjustment in the smartphone market in the second half of last year, the industry believes that smartphone inventory depletion is nearing the end, and that the market share of high-end flagship brands will continue to grow, presenting a trend of the strongest being the strongest.

Ji-Pu Viewpoint:Apple accounted for about 19.2% of the global cell phone market in 2022, and achieved a record high profit share of 85%. It is expected that Apple's shipments will remain good this year, with an annual estimate of about 230 million units. In addition, Sony (SONY) with high-end cell phones with CMOS sensor (CIS) chip, for many years for the solid Apple supply chain; TSMC Kumamoto plant is mainly for the production of CMOS sensing elements for Sony, Sony believes that the plant's production capacity is not enough to meet the demand for more production capacity, which means that the release of the latter part of the outsourcing of the sealing and testing orders synchronized with the increase in the CIS sealing and testing TSMC subsidiaries, Precision Materials, Caiyu will be synchronized to benefit. On the other hand, TSMC is still making inventory adjustments in the server field, and needs to continue to observe the impact of AI applications on the server industry to determine whether it will have a crowding-out effect on the budgets of large manufacturers.

Jun W3｜Technology News Highlight：Three Advantages...Hon Hai Group

Hon Hai (2317) Group is rich in topics, not only the parent company Hon Hai is the largest assembly plant of Apple, almost all Apple products, including the recent popular AI servers, electric vehicles, semiconductors, and low-orbit satellites, can be said to have everything, so that the Hon Hai Group has the largest economic scale, the broadest global layout, coupled with the most in-depth vertical integration of the three major advantages, its companies are therefore the same as the rain and dew. The companies under the Hung Hai Group have all the advantages. Hon Hai itself is supported by four major themes: Apple stocks, AI server concepts, low-orbit satellites, and electric vehicles. In terms of Apple's supply chain, Hung Chun and GIS-KY, which are part of the "Hung Family Army", are familiar manufacturers in the market, supplying chassis and lenses for Apple's first MR helmet respectively. Hung Hai's Chairman, Mr. Liu Yang-wei, has already clearly revealed the three major directions of Hung Hai's operation this year, namely, electric vehicles, semiconductors and low-orbit satellites. As for electric vehicles, HONGHAI will start from its parent company, cooperate with its connector company Guangyu, and integrate semiconductor-related investment energies to create a powerful one-stop service model and expand the order volume.

Ji-Pu Viewpoint:The Hon Hai Group's aggressive expansion into the semiconductor field demonstrates a keen eye for emerging technologies, investing heavily in semiconductor sourcing and the electric vehicle field, and securing the Group's competitiveness in the electronics manufacturing field and key IC supply chain through acquisitions, re-investments, and equity holdings. In addition, Hon Hai has also capitalized on the rising trend of satellite communication technology by actively participating in the low-orbit satellite field to ensure that its diversified deployment can continue to maintain its leading edge and lay a solid foundation.

Related News:Triple Advantage...Hung Hai Group

Jun W2｜Sun Micron Receives Advanced Packaging Order from TSMC

Fidelity's AI chips have become so popular that customers are frequently pursuing orders. While TSMC is actively satisfying Fidelity's foundry capacity demand, it is rumored that CoWoS, the advanced package that matches with the shipment, is experiencing a tight production capacity, with a shortfall of as high as 10% to 20%, and is urgently looking for Sun Micron.(3711)The rescue of Sun Micron Semiconductor, a subsidiary of Investment Control, has pushed up the utilization rate of Sun Micron's high-end packaging capacity, and has caught up with Fidelity's AI boom, resulting in a surge in gross profit margin. TSMC, Sun Micron has never commented on a single customer and order information. According to the industry, it has recently been rumored that TSMC's advanced packaging CoWoS production capacity has been severely strained due to the clustering of orders from large high-speed computing customers (such as Phaidon), and that the customers have requested TSMC to expand its CoWoS production capacity. However, TSMC has chosen to outsource the orders to the world's largest packaging and testing vendor, Sunrise, after taking into account the contribution to its own gross profit margins. Currently, CoWoS-L/R and InFO can share the same machine, but at this stage, InFo capacity is already full to meet the demand of Apple's new machine, and there is no other space for the time being; CoWoS-S has been rumored to be partially debottlenecked for expansion, but there is no large-scale upgrading of production capacity.

Ji-Pu Viewpoint:TSMC's CoWoS capacity is unable to meet customer demand, and Sun Microsystems has taken an outsourcing tonic. Taking the Fan-Out-Chip-on-Substrate-Bridge technology, which was announced in mid-May this year, as an example, the latest breakthrough has been realized, which enables the integration of two ASICs and eight High-Bandwidth Memory (HBM) components through eight bridge connections in a large, high-efficiency package with a size of 70mm x 78mm. FOCoS-Bridge technology not only enables more efficient use of computing resources and accelerates the processing of large amounts of data, but also drives the development of compute-intensive tools such as AI algorithms, deep learning, and scientific simulation. the combination of AI and HPC has tremendous impact and potential across industries and is driving the need for innovative packaging solutions to meet the demand for high-performance, high-performance, 70mm x 78mm packages. solutions to meet the growing demand for high-performance computing.

Sun Microsystems Receives Advanced Packaging Order from TSMC

Jun W1｜Tech News Highlight：AI demand soars! NVIDIA to build supercomputers in Israel

Nvidia Corporation, a major U.S. chip maker, has been successful in developing the AI market and has become a hot topic in the industry. Co-founder and CEO Jen-Hsun Huang's recent visit to Taiwan has also set off a wave of "NVIDIA fever", with foreign media reporting that in response to the rising demand for AI, Nvidia is constructing an AI "supercomputer" in Israel. As reported by Reuters and The Israel Times, Gilad Shainer, senior vice president of Fidelity, said that Fidelity is cooperating with about 800 startups and tens of thousands of software engineers in Israel, and is spending hundreds of millions of dollars to develop a computer system called "Israel One", which is said to have a computing power of up to eight exaflops per second, and will become one of the world's fastest AI supercomputers. It will be one of the fastest AI supercomputers in the world, and part of it is expected to be in operation by the end of the year. According to the report, the operation of "Israel One" will be based on Pfizer's newly launched Spectrum-X network platform, which is a high-performance Ethernet architecture specifically built for generative AI. Shiner claimed that AI is "the most important technology of our lifetime" and that the process requires GPUs (graphics processing units). Now that generative AI is ubiquitous and requires large databases for training, Israel One is a major investment that will help drive innovation in Israel and around the world.

Ji-Pu Viewpoint:Nvidia's sales for the quarter are expected to reach $11 billion, far exceeding market expectations due to strong demand for AI. In addition to supply chain benefits, Nvidia will become one of the fastest AI supercomputers in the world by building the Israel One AI supercomputer in Israel to meet growing demand and strengthening its collaboration with Israel. The emergence of these transformative technologies provides unprecedented opportunities for companies, and we can foresee that they will continue to intensify their exploration of their applications and stay ahead of the curve in the ever-changing business environment.

May W4｜Technology News Highlight：Semiconductor shocked by Mainland China's ban on Micron

China's State Internet Information Office (SIIO) announced yesterday that their review has found serious cybersecurity issues with Micron, the largest U.S. memory chip maker. In order to address this issue, the mainland has asked critical information infrastructure operators in China to stop purchasing Micron's products. This news has caused a huge shock in the global memory market, and the market landscape is expected to change. The industry interpreted this as China's first countermeasure to the U.S. tech wars. Although the White House and Beijing have said in public that they want the relationship to develop in a positive direction, the fact that Micron, another major U.S. semiconductor producer, has been severely sanctioned by China underscores the lingering concern that a U.S.-China technology war is still on the horizon. Micron is the world's third largest DRAM maker and one of the top five NAND chip makers. The incident has created difficulties for Micron's development in the China market, while competitors such as Taiwan's Nan Ya Technology Corp. and South Korea's Samsung and SK Hynix may benefit. However, the global memory market is currently oversupplied, so Micron may face greater inventory pressure and may need to promote its products through price cuts, which could slow down the market's recovery. In March, China's State Internet Information Office announced that it was conducting a network security audit of Micron products sold in China. Micron responded by saying that it is aware of the situation and has been actively communicating and cooperating with the relevant parties.

Zipo's point of view: China's State Internet Information Office last night posted on its WeChat public number "Net Information China" news about a network security review of Micron's products, saying that the move is aimed at protecting the security of the country's critical information infrastructure. From an industry chain perspective, Micron's products in China should be transferred to Samsung, SK Hynix, CK Storage, Fujian Jinhua, etc. It is also expected that Micron will compete for market share in North America and other non-Chinese markets, which will be another shift in the industry landscape. The size and impact of the shift will continue to be observed. In terms of price, DRAM / NAND Q3 stabilized but when it will rise to continue to observe, and Micron was blocked by China will have an impact on the price, so far there is no clear sign.

May W3｜Technology News Highlight: MediaTek's 6G NTN Global Seamless Overlay

MediaTek (2454) released the latest 6G NTN (Non-Terrestrial Networks) technology white paper on the integration of satellite and terrestrial networks on May 15. Compared with the 5G standard, which will incorporate NTN technology at a later stage, MediaTek believes that 6G NTN technology should be considered at the initial stage of the design of the 6G physical layer and the protocol layer, and that it will create a three-dimensional coverage of land, sea, air, and space by complementing satellite and terrestrial networks in the future. In the future, the satellite network and terrestrial network will complement each other to create a three-dimensional coverage of land, sea and air space. 3GPP member meeting was held in Taiwan on May 2, and MediaTek, TAICS, and the Bureau of Industry joined hands to strive for 6G business opportunities for Taiwan, pursuing the complete integration of satellite and terrestrial communication standards, and, in comparison with the privately owned satellite communication technology, once 6G NTN becomes the 3GPP open standard and definition, it will be able to utilize the existing global communication standards and definitions. Once 6G NTN becomes a 3GPP open standard and definition, it will be able to utilize the economic scale of the existing global communications industry chain and make satellite communications from niche markets to become a technology that can be used by mainstream consumers on a daily basis, and can be applied to mobile phones, automobiles, and the Internet of Things (IoT), etc., to provide uninterrupted global coverage of network communications. MediaTek's white paper points out four key 6G NTN technology R&D directions, including high-efficiency waveform design, predictable mobility management, large satellite antenna arrays and beamforming, and satellite and terrestrial network spectrum sharing technologies. MediaTek said that with the commercialization of 5G, the expectations of the outside world for new-generation communications have been raised, and satellite smartphones with global coverage have become a hot spot in the near future. Satellite communication chips will penetrate into the general end-consumer market through various mobile devices, and the integrated 6G satellite and terrestrial communications will make a greater contribution to promoting global development, increasing productivity, and creating a new business model, which will bring about the transformation of society. This will lead to a multifaceted transformation of society. Through the 6G NTN satellite network, network services can penetrate into previously inaccessible parts of the world, including sparsely populated areas, uninhabited mountains, oceans, etc., which can help reduce the gap between digital cities and rural areas. In addition, through the combination of IoT NTN technology and devices, it can be used for monitoring water sources, fire alarms, risk prevention, and even as a way to detect illegal logging and poaching, which is of great benefit for sustainable development and protection of the environment. MediaTek will continue to develop 6G technology and seize the market opportunity. According to a survey conducted by Precedence Research, the 5G non-terrestrial communications market will reach $3.79 billion in 2022 and is expected to reach $27.69 billion in 2032, with a compound annual growth rate of 22% from 2023 to 2032, demonstrating the degree to which various sectors are actively promoting the development of NTN. This time, MediaTek published a white paper on 6G technology development and initiated the layout of 6G next-generation communication to grasp the development opportunity of next-generation communication.

Wisdom: The NTN (Non-Terrestrial Networks) technology white paper emphasizes the compatibility and complementarity between satellite networks and terrestrial networks, creating a three-dimensional coverage of land, sea, and air across the entire terrain and space, and providing users with seamless and intelligent communication services. MediaTek and Qualcomm are the world's leading smartphone chipset manufacturers (roughly 1/3 on both sides), so it is a reasonable strategy for MediaTek to proactively lay out next-generation 6G technology to stay competitive. The current 6G vision is a flexible, multi-dimensional, integrated wireless communications system that provides immersive mobile connectivity in a truly ubiquitous manner. Unlike 3G~5G services, which still rely on the deployment of base stations to transmit signals, 6G networks including satellite communications can improve the problem of terrestrial network coverage limitations, jumping over the limitations of terrestrial base stations to provide basic satellite text messaging, voice, and Internet services to areas around the world that do not have terrestrial network coverage. In the past, satellite communications were mostly for a few, special applications and relatively expensive, but in the future, with the development of 5G and 6G mobile communications technology, it is likely to become more popular and affordable, and is expected to give rise to more innovative application scenarios, as well as services such as emergency disaster relief, uninhabited area monitoring, wildfire prevention and control, and maritime buoys information collection. What companies will benefit from 6G in China: At present, the popularization of 6G is still in the early stage, and there are still some technical and spectrum sharing issues that need to be dealt with. If this layout is fermented in the future, those who will directly benefit from it are of course MediaTek's supply chain [TSMC, UMC, MPEG, Precision Measurement, Yongzhi, Sunrise, Kyoyuan Electric, and Silicon Power]; and those who can be concerned about it at present should be the LEO satellite-related supply chain.

May W2｜TSMC 3nm to be highly mass-produced

TSMC (2330), a leading foundry, uploaded its annual report to the shareholders' meeting today. The report points out that it will increase its investment in research and development to continue to expand its overall competitiveness and technological leadership, and that it will continue to optimize its manufacturing operations, including the "digitization" of its fabs, in order to improve efficiency and productivity in support of 3-nanometer high-volume production this year and beyond. As we move into 2023, TSMC noted that general economic and geopolitical uncertainties remain, and as the global COVID-19 outbreak subsides, people are entering a smarter and more connected world. As semiconductors become increasingly important and ubiquitous in all facets of our daily lives, and as semiconductor technology becomes the foundation of the modern digital economy, the value of semiconductors continues to rise in the global supply chain, creating greater value opportunities for customers and TSMC," said TSMC. For TSMC, fulfilling our mission to be a trusted technology and capacity provider to the global logic IC industry is more important than ever. The company will pursue and capitalize on strong growth opportunities by maintaining the triad of technology leadership, manufacturing excellence and customer trust. TSMC noted that it is increasing its investment in research and development to continue to expand its overall competitiveness and technology leadership, and to maintain TSMC's technological rhythm with industry-leading advanced manufacturing processes and 3D IC solutions, providing the value of the Company's technology platform to help customers enhance their product competitiveness and support their future growth advantages. TSMC will also continue its efforts to optimize manufacturing operations, including fab "digitization," to improve efficiency and productivity to support high volume production of N3 this year and beyond.

Chi-Pu's view: TSMC will further increase its investment in R&D and strive to optimize its manufacturing operations, including fab digitization, to improve efficiency and productivity in order to support 3nm high volume production. TSMC's major customers in the advanced manufacturing process have released their financial forecasts for this year, and they are currently unable to find enough capacity to support 80,000 to 100,000 wafers per month. Therefore, strengthening the cost structure in other areas or enhancing the conditions necessary for future growth will be the possible strategic direction for the time being.

May W1｜TSMC Launches New Software to Accelerate the Adoption of 3nm Processes in Automotive ICs by 2025

TSMC said it will launch new software this year to help customers invest in building advanced automotive chips and accelerate adoption of TSMC's latest 3nm process technology. The new software, announced at a Silicon Valley conference on June 26, will enable automotive IC designers to start designing their chips about two years earlier. This will enable the industry to utilize this technology as soon as TSMC launches its "automotive version" of the 3-nanometer chip process in 2025. According to Reuters, many large automotive chip suppliers, including NXP and STMicroelectronics, have chosen to entrust TSMC with the production of their wafers. However, automotive wafers must meet higher standards for ruggedness and durability than wafers for consumer electronics. TSMC's special processes for the automotive industry are typically introduced several years later than similar processes for consumer electronics wafers. In the past, automotive chipmakers have needed additional time to design chips for these specialized process lines, resulting in automotive chips that are years behind the processes used in the chips in the latest smartphones. The goal of TSMC's new software is to help automotive IC designers design "automotive-ready" processes before they are launched, so that they can be adopted and produced in parallel when the processes are launched.

Ji-Pu Viewpoint:Despite the downturn, TSMC revealed its latest technology developments at the latest North American Technology Forum, which included three new members of the 3-nanometer technology family, as well as advances in 2-nanometer technology using nano chip transistor architectures. The introduction of these new technologies not only improves power, performance and density performance, but also helps address the issue of automotive chips lagging behind smartphone chips. In addition, TSMC has introduced new software to help automotive IC designers design before the process is launched, so that the process can be adopted and manufactured simultaneously when it is launched, helping to drive technological innovation and advancement.

Apr W4｜ Technology News Highlights: Creative orders strong, optimistic year-round operations

IC design services company Creative (3443) is optimistic that system and brand manufacturers will continue to invest in the development of customized application-specific ICs (ASICs). Benefiting from the technology and capacity support of its parent company, TSMC (2330), Creative has smoothly pushed its advanced processes to 5nm and 3nm, and launched InFO and CoWoS advanced package integration solutions to expand its market share in the artificial intelligence (AI) and high-performance computing (HPC) markets, and has been actively pursued by international customers. The company has also launched InFO and CoWoS advanced packaging integration solutions to expand its market share in the artificial intelligence (AI) and high-performance computing (HPC) markets, and has been actively pursued by international customers. Last year, Creative's revenue and profit hit record highs, and net income per share after tax of NT$27.69 was better than expected. The Board of Directors resolved to distribute a cash dividend of NT$14 per common share. Although the first quarter of this year has entered the traditional low season, with the AI/HPC applications boosting the increase in the number of commissioned design (NRE) cases and ASIC shipments, the first quarter consolidated revenue of NT$6,529 million is still the second highest quarterly revenue in the history of the company, and we are optimistic about the full year of quarterly growth in operations. Last year, Creative benefited from the booming AI/HPC projects, and this year, Creative will gradually shift to ASIC volume industry to drive revenue growth. In order to win more NRE projects from international large-scale manufacturers and to increase the advantage of receiving orders in the trend of chiplet design, Creative has not only obtained TSMC's 5nm and 3nm advanced manufacturing process, InFO and CoWoS advanced packaging technology and capacity support, but also launched a number of integrated solutions, which have been successfully favored and adopted by customers. GLink 2.3, TSMC's third-generation 5nm interconnected silicon IP that combines TSMC's InFO and CoWoS packaging technologies, has been completed and silicon validated. The company has mass-produced ultra-large System-on-Chips (SoCs) equipped with two to eight high-bandwidth memories (HBMs) and CoWoS for six customers, and has successfully advanced into the AI/HPC field. The HBM3-enabled controller and Physical Layer (PHY) have completed 5nm silicon validation and have been adopted by several customers. CREATIVITY has also finalized the design of an AI/HPC computing ASIC that integrates TSMC's 5nm advanced process and 2.5D advanced packaging to create a large-scale cloud data center for a customer, and finalized the design of a test chip that supports the 3nm N3E process and integrates the HBM 3 and GLink IPs in the first quarter of this year. In addition, Creative has also finalized 7nm and 16nm workstation-grade optical communication chips for customers, which can be used in ultra-large data centers and will enter mass production this year.

Ji-Pu Viewpoint:Benefiting from AIGC and HPC, Creative achieved record high revenue and profit last year, and although the first quarter of this year has entered the traditional low season, it still achieved the second highest quarterly revenue in history. In addition to the original high-speed computing field, CREATIVITY has gradually transformed into an ASIC industry to drive revenue growth. Meanwhile, CREATIVITY has also obtained a number of integrated solutions, which have been successfully favored and adopted by customers. Therefore, we are optimistic about Creative's future development in ASIC development and advanced manufacturing processes, packaging integration solutions, etc., and are optimistic about its operational growth.

Apr W3｜ DDIC Major Manufacturers Show Growth Momentum as Demand Recovers

Elon's Q2 results are up:ELAN (2458), a major NB touch IC maker, has benefited from the inventory restocking effect of its customers and has recently received rush orders for a number of product lines. Elon's first quarter consolidated revenue was NT$2.405 billion, up 22.4% quarter-over-quarter and down 41.3% year-over-year. According to the company's estimation, even though the end buyer's sentiment has not yet strengthened significantly and no long term orders have been received, Elan's results for the current quarter will be better than those of the first quarter, and the growth rate will depend on the orders it receives in the near future.

Unitech's Q1 revenue exceeded expectations, and Q2 was even more aggressive:Though still affected by inventory adjustments in the consumer electronics production chain, panel driver IC maker Unitech (3034) has surpassed its high Q1 consolidated revenue of NT$24.046 billion, as panel makers' inventory depletion nears the end of the quarter. Lianyong's recent integration of touch panel driver ICs (TDDI) shipments have become stronger and prices have risen, and large-size panel driver ICs (LDDIC) have also seen manufacturers continue to pull in stronger shipments, coupled with OLED panel driver ICs that have been sampled by Apple, the company is optimistic that operations in the second quarter will enter into a cycle of growth.

Wonderland Automotive panel driver IC shipments are surging:Panel driver IC maker Himax automotive panel product line shipments are expected to surge again this year. According to Himax, the Local Dimming (Local Dimming) Timing Controller (TCON) solution has successfully penetrated into the high-end vehicle models of many car manufacturers, and shipments of this product line are expected to grow strongly. According to Xeikeng, the automotive TCON technology for local dimming, in addition to improving the contrast ratio of automotive screens and other functions, can also significantly reduce the power consumption of displays, which is very critical for electric vehicles and larger displays. Legal person pointed out that Chi-King in the automotive panel driver IC market has successfully entered the Mercedes-Benz, BMW and Fuchs Group and other large European and American car manufacturers, as well as the rise of China's electric car brands in recent years have used Chi-King's automotive panel driver IC products, due to the size of the automotive panels have been expanding in recent years, it is expected to become Chi-King this year's shipments of the main reason for the momentum of the surge in the high.

Ji-Pu Viewpoint:ELAN, Uni-President and Wonderful are all IC design houses in different fields. In the short term, these three companies are benefiting from the effects of inventory replenishment and demand recovery on the customer side, and their results are showing a continuous upward trend, and overall revenue growth is expected to continue. Meanwhile, Vision's automotive panel product line is also expected to reach new heights, especially its automotive zone dimming TCON technology, which has important application value.

Related News:

Elong's Q2 results upward

Unitech's Q1 revenue exceeds expectations, Q2 is even more aggressive

Xeikyo Automotive Panel Driver ICs Shipment Surge

Apr W2｜Intel's New Generation GPUs: TSMC Eats It All!

With the demand for artificial intelligence (AI) and e-sports and other related applications to maintain the growth trend remains unchanged, Intel adjusted and spun off the graphics processing unit (GPU) business organization, but the GPU research and development momentum to maintain the original plan to proceed, and according to the industry news, Intel's second-generation Battlemage and the third-generation Celestial will be launched in 2024 and 2026, TSMC won the 4-nm and 3-nm wafer foundry large orders. TSMC won the 4nm and 3nm foundry orders. At the end of last year, Intel restructured and spun off its GPU business organization. The Accelerated Computing Systems and Graphics Group (AXG), which was originally responsible for the GPU product line, will be split into two, with the consumer-focused business being merged into Intel's Customer Computing Group (CCG), and the data-centric and supercomputing GPU business being merged into the Data Center and Artificial Intelligence Business Group (DCAI). Intel is optimistic that gaming and AI-related applications will drive strong demand for GPUs, and that discrete graphics and accelerated computing will be key growth engines in the future. Therefore, Intel will launch product roadmaps that can be paired with Intel's core CPUs for each target market, in the hope of focusing on competitors such as Supermicro and Fidelity to gain a higher market share. In terms of the development of existing Arc architecture GPUs and consumer discrete graphics cards, Intel has maintained its GPU R&D program despite the fact that its operations are facing greater challenges this year due to the pressure of inventory depletion in the production chain. According to industry sources, Intel will launch its second-generation Battlemage graphics chip on the Xe2 architecture in the second half of 2024 and its third-generation Celestial graphics chip on the Xe3 architecture in the second half of 2026. Although Intel has invested heavily in building 4nm and more advanced process fabs, the GPU product line will continue to operate in partnership with foundries, with TSMC as the sole partner.

Ji-Pu Viewpoint:Intel has restructured and spun off its GPU business organization, but the GPU R&D momentum remains on track, with product roadmaps targeting different market needs in the hope of capturing a higher market share, especially in gaming and artificial intelligence related applications. The chart below shows last year's GPU market share, with Intel having a market share of approximately 6%. Intel's second-generation Battlemage and third-generation Celestial will be launched in 2024 and 2026, respectively, using TSMC's 4-nanometer and 3-nanometer foundries, and Intel continues to operate in partnership with foundries, with TSMC as the sole partner.

Reference: JPR

Intel is also building the Ponte Vecchio graphics chip for AI and HPC, which is expected to be produced using TSMC's 7nm and 5nm. Overall, Intel is committed to developing its GPU product line and OEM partnership with TSMC to address strong demand and increase market share in e-sports and AI-related applications.

Apr W1 | Samsung to cut memory production, industry applauds

Global memory chip leader Samsung Electronics of South Korea is rumored to change its previous attitude of "adamantly not reducing production" and is considering to start reducing production, which has been loudly cheered by the industry, which has said that this is "the most crucial action to save the market of DRAM and NAND Flash", and that good news is expected to come to memory manufacturing and module factories such as Nanya, Macronix, WaPac, Qunlian, and Vigantas. Samsung is scheduled to announce its quarterly financial results on April 7 and hold a presentation, and the market is paying close attention to whether the company will maintain its previous policy of not artificially reducing production. Market forecasts, Samsung's semiconductor business and device solutions division (DS) this year's operating losses may exceed 10 trillion won. Previously, Samsung insisted that there is no artificial production reduction plan, will not directly reduce the input wafers, but to seek a natural reduction in wafer production. If Samsung can reduce memory production, not only favorable to the global memory market quickly warmed up, but also allows its own profits gradually return to the past glory, can be said to be a win-win situation, especially at the time of the overall semiconductor inventory adjustment has not yet ended, Samsung has a huge pressure on profitability, and the need for memory business to help, and therefore the reduction of production is inevitable. Previously, Micron announced its financial report, although there was a sharp decline in revenue, but the stock price can be turned red, so there are rumors in the industry that there are unknown reasons for the DRAM/NAND price to see the bottom of the barrel.

Ji-Pu Viewpoint:Observing the situation in the industry, memory inventories continue to improve but still high, supply and demand are still a huge mismatch, the price trend is still challenging, and profits this quarter are expected to be lower than originally estimated. We are still holding a conservative and pessimistic view, memory prices have been falling since Q4 last year, Samsung insisted on not reducing production, there is no sign of stopping the decline, and at the beginning of the year, it was expected that it would probably fall all the way to Q4 this year. now, Samsung's attitude may be softening, but at present, in the case of uncertainty about the content of the reduction in production and the market inventory is still on the high side, it is expected that the price will continue to be shocked in Q2, and there is a chance of bottoming out only in Q3. The price is expected to continue to oscillate in Q2, with a chance to bottom out in Q3.

Samsung to cut memory production, industry applauds

Mar W4 | Security Transfer Authorization Shocks IC Designs

The Financial Times (FT) reported that silicon intellectual property (IP) maker Anmou (Arm) has recently notified a number of large customers such as MediaTek, will significantly change the mode of authorization to collect money from the previous chip price as the basis for calculating the value of the terminal device products, the earliest next year into line, which will make Qualcomm, MediaTek and other IC design vendors cost a great deal of money, subverting the ecology of the industry. Market assessment, Anmou is scheduled to be listed in New York this year, its parent company, Softbank Group, this planning should be to increase revenue and profitability, in order to attract investors. The industry believes that more than 95% of the smart phones on the market are using Amlogic's IP as the infrastructure, once Amlogic changes its charging model, it will have a big impact on cell phone chips and brand manufacturers. Ammann, MediaTek (2454) and other industry yesterday (23) on the above news did not comment. A number of IC designers that have adopted Amo's architecture believe that they still need to observe the subsequent development of this matter. A number of Amo's industry executives and former employees said that Amo used to base its IP licensing fees on the price of the chip, but now it plans to charge device manufacturers based on the price of the device, which is the biggest change in its business strategy in decades, and is estimated that the IP licensing fees will be several times higher than the original fees.

Ji-Pu Viewpoint:Amor's hopes for an IPO are back on the market, and Amor's proposed new business model suggests that license fees will be based on the "average selling price" (ASP) of the mobile device, rather than the price of the chip. The adjustments are mainly related to Amor's most important "Cortex-A" design, which is essential for the development of smartphone application processors. Cell phone IC, for example, the current ASP of a single IC is about 10 ~ 50 USD royalties of about 1 ~ 2%, the average price of a smart phone (2022) fell about 330 ~ 375 USD, if the change to device manufacturers to charge, will significantly increase the Amor license income, but also increase the attractiveness of the IPO to investors. In the face of such a major change in Anmou's business strategy for the first time, a number of IC design industry customers using Anmou's structure, said that the willingness to accept the new arrangement is not high, and whether next year as expected in the next year into the licensing payment model, but also to be observed again after the development.

Mar W3 | OpenAI Re-introduces Language Model GPT-4 with Improved Academic Performance but Pitfalls Remain

OpenAI, a startup company, today unveiled a stronger language model, GPT-4, which Microsoft says is already being used in its Bing chatbot. OpenAI says GPT-4 performs better on some professional exams, but is still "illusory" and can't be fully trusted. Today, OpenAI officially launched its "more powerful" language modeling system, GPT-4. ChatGPT has been used in a large number of applications since its launch, and has attracted both positive and negative comments. OpenAI said that GPT-4 took six months to test, and even though it is "far from perfect" and does not perform as well as humans in real-world situations, the results of this enhanced version of the language model are better than the previous generation of the system in different professions and academic levels. Comparing GPT-4 with its predecessor, GPT-4 scored in the top 101 TP3T, or 90%, of the test takers in the Bar Exam simulation test, while the latter scored in the bottom 101 TP3T, and GPT-4 is also more evolved than its predecessor in other tests, such as the SAT Reading and Math exams. In everyday conversational situations, the difference between GPT-4 and its predecessor is not obvious, but when the complexity of the task reaches a certain threshold, GPT-4 is more creative and able to execute more subtle commands.

Ji-Pu Viewpoint:OpenAI has released its newest artificial intelligence (AI) model called "GPT-4". In addition to greater accuracy, creativity, and collaboration, this newest AI model also supports image input, allowing GPT-4 to interpret image content and even answer questions in text. Based on current information, GPT-4 is powerful but not entirely reliable, and may provide "hallucinatory" answers or make errors in reasoning. However, it is still difficult to "predict" GPT-4, and Open AI will continue to train models in this area. So just as with autonomous driving (L3+), we're still waiting for GPT4 to be commercially available, but we can't trust it to provide a solution.

A closer look at GPT-4 and 14 more amazing details! Might it help crime, beyond human control?

GPT-4 Continues to Drive AI Concerns Lawmakers Expect Major 3 Companies to Benefit

Mar W2｜IC design houses expected to bottom out in Q1

Influenced by the traditional off-season effect and the supply chain's continuous de-inventorying, most of the IC design houses expect the Q1 results to slip, however, the operation will bottom out in Q1, and Q2 should be able to gradually stop the decline and rebound. IC design industry has recently seen the emergence of rush orders from customers, with the majority of them being for TVs, monitors, and high-end laptop-related products that underwent inventory correction earlier, and the demand for the consumer electronics market continues to be sluggish due to the fact that market demand for a number of products is still weak. Market demand for consumer electronics products continues to be sluggish as demand for most products remains weak and is still at the inventory adjustment stage. Coupled with the Lunar New Year holiday, fewer working days, and the traditional low season for the semiconductor industry in the first quarter, most IC design houses anticipate that their results in the first quarter of this year are expected to continue to decline compared to the fourth quarter of last year. Cell phone chipset maker MediaTek expects Q1 revenue to be around NT$93bn to NT$101.7bn, a decrease of 6% to 14%, while image sensor and micro-optical component maker Caiyue expects Q1 revenue to be reduced by 2-digit percentage points. After several quarters of inventory adjustments, most IC design houses expect supply chain inventories to bottom out in the first half of the year, and supply and demand should gradually stabilize. In the second quarter, with the increase in the number of working days and the unsealing of the epidemic in China, the economy will have a chance to recover, and operations should be able to stop the decline and rebound.

Ji-Pu Viewpoint:According to the current industry observation, the bottoming time of each product is not the same, about the first quarter of 2023 to May, so gradually see the IC design industry continue to chip orders to the foundry. The part of the foundry, from the beginning of the OLED DDIC rush orders, the United States to limit the AI chip sales in mainland China buffer period will expire this year and the current stocking wave, ChatGPT hot, accelerate the NV in TSMC N3B/E NTO schedule, are all good signs of the industry to meet the turnaround.

Mar W1 | Technology News Highlights: U.S. Semiconductor Grants Set Profit Conditions

The U.S. Department of Commerce began accepting applications for Wafer Act grants on February 28th, and will proceed with a five-phase review process. Not only do semiconductor applicants have to comply with numerous restrictions, but those who receive more than $150 million in grants must also share profits with the government. The Wafer Act, signed into law by President Joe Biden last August, provides $39 billion in grants for semiconductor companies to invest in U.S. wafer manufacturing, which, when combined with semiconductor R&D grants, total $52 billion for companies to apply for. In announcing the opening of applications on February 28, Commerce Secretary Gina Raimondo emphasized again that "the Semiconductor Industry Grant Program is a matter of national security, and the government will not arbitrarily write a blank check to any company. The Department of Commerce estimates that each company will receive a grant of about 5 to 15 percent of its expected capital expenditures. Even taking into account government loans and loan guarantees, the total amount of grants to a single industry will not exceed 35 percent of expected capital expenditures. In order to ensure that these grants actually restore the competitiveness of the U.S. semiconductor industry and to prevent future supply chain issues from impacting the U.S. semiconductor industry, the Department of Commerce has set stringent thresholds for the grants.

The Bill seeks to apply for the relevant specification:

Ji-Pu Viewpoint:In early February of this year, the U.S. media reported that during a previous visit to Taiwan by Speaker of the House Carlos Pelosi, TSMC founder Chang Chung-mou said outright that if the U.S. thinks it can enter the world's most complex electronics manufacturing market by throwing money at it, that's "terribly naïve," and that protecting Taiwan is the right solution. From the U.S. political point of view, regardless of whether the cost of semiconductor manufacturing has skyrocketed or not, pulling the semiconductor manufacturing chain back to the U.S. will not cost too much, and it will solve the geographic risk of advanced processes being produced only in Asia, and will boost the U.S. market for equipment, jobs, and other industries around the country.

Feb W3 | Technology News Highlight: Microchip Expands Silicon Carbide 8-inch Wafer Fab and Completes Third-Generation Semiconductors

Microchip, a semiconductor company, today announced plans to invest $880 million in an 8-inch fab to expand its silicon carbide (SiC) and silicon wafer production capacity in Colorado, USA, to strengthen third-generation semiconductors for automotive, aerospace, and defense applications, according to a press release today. Microchip President and CEO Ganesh Moorthy said Microchip has a long history of working with the City of Colorado Springs and the State of Colorado Springs. Microchip and the City of Colorado Springs (City of Colorado Springs) and the state government has a long history of cooperation, he pointed out that the U.S. Chip and Science Act (CHIPS and Science Act) investment tax credits, positive impact on the business of Microchip, Microchip is active in several semiconductor factories to seek capacity expansion grants, including Colorado Springs, Microchip pointed out that the The Colorado Springs facility currently employs more than 850 people and produces 6-inch wafers, and Microchip is building an 8-inch wafer manufacturing facility that will dramatically increase the number of wafers produced, with the new facility expected to create 400 new jobs ranging from production specialists to technical positions in equipment purchasing and management, process control and test engineering.

Ji-Pu Viewpoint:Even with the downward correction of the global economy and growth, the global situation of SiC substrate is still in a state of oversupply. Among the third type of semiconductors, the current situation is still that the substrate wins the world. Therefore, large manufacturers continue to increase the production capacity of substrates and move closer to 8-inch. At present (2022e), the global production capacity is around 700,000~900,000 wafers/year at 6-inch equivalent capacity, and the largest supplier is Wolfspeed, with the ratio as shown in the chart below. Other major substrate manufacturers are summarized in the table below:

Microchip Announces $880 Million Investment Program in Class III Semiconductor Silicon Carbide Manufacturing

Feb W2 | Technology News Highlight: US-China tech war to escalate again, will restrict investment in China's technology

Business is like a war zone, and the US is once again offering ironclad policies! Within two months, the Biden administration is expected to introduce a new law restricting investment in some of China's technology areas, such as quantum computers, advanced semiconductor processes, and AI. Scholars analyze that the U.S. is intervening for the sake of increasing domestic employment, and also for political reasons other than national security issues.

Sacrifice more tech bans

The New York Times reports that the Biden administration has been preparing to restrict U.S. companies from investing in China, especially in high-tech areas. It is estimated that the new law will be introduced within two months. The U.S. Treasury Department is said to have been in discussions with the European Union and other governments to ensure that U.S. funds in similar areas do not flow into China. Experts say that in the near term, the U.S. will continue to ban direct investment in sensitive areas in China, such as quantum computers, advanced semiconductor processes, and AI with military and surveillance capabilities, to control the flow of funds.

75% Wafer in East Asia

The Diplomat Magazine reported that the U.S. passed the Wafer Act in an attempt to pull semiconductor production lines back to the mainland to counter China, and to avoid over-reliance on East Asia to produce global semiconductors and high-end wafers of up to 75%, forcing many large manufacturers to choose sides between the U.S. and China. U.S.-China economic relations have been tense since the 1990s, when the U.S. believes that China manipulates the exchange rate and makes U.S. domestic industries uncompetitive, so it tries to rebalance the economic relationship between the two countries, but it is not completely cut off the economic and trade relations, but rather hope to increase the total amount of U.S. exports to China.

China as a national security issue?

In the 1990s, China-US economic and trade relations were based on the reciprocity of each other's needs, with China hoping to rapidly boost its economy, reduce poverty, and rise to international prominence, while the US hoped to obtain cheap products and loans to maintain the momentum of the US economic locomotive. However, globalization has not allowed all people to benefit equally, the public's sense of relative deprivation is increasing, countries have produced populist political parties and policies, such as the United States since 2017, China and the United States in the issue of trade tariffs and trade surplus and deficit issues do not see eye-to-eye, the ruling Trump administration will be on China's imports of goods extensive tariffs, increased retaliatory measures, and will be China's economic and technological upgrading of the national security issue.

Ji-Pu Viewpoint:Possibly as a result of the recent controversy over the 'balloon incident', some observers are suggesting that Washington will be taking more punitive measures against Beijing in the near future, which could include long-awaited restrictions on foreign investment. While the White House has declined to comment on the executive order at this time, sources suggest that the Biden administration's reason for declining to comment is that it wants to give the industry a chance to weigh in on the proposed rules before the program goes into effect. The impact of the executive order on China, if implemented, will depend on the content of the order. However, as the U.S. is the largest recipient of foreign technology investment from China, it will be interesting to see if there are any other countermeasures that China will take and whether they will hurt the U.S. in the future.

Feb W1 | Technology News Highlight: 2022 Q4 MediaTek Press Conference

Cell phone chip maker MediaTek (2454-TW) held a press conference today (3) and announced its fourth quarter financial results last year, affected by customer inventory adjustments, gross margin, profitability, net income after tax also fell to NT$18.514 billion, a two-year low, net income per share of NT$11.66, while the full year's net income maintained a modest growth of NT$118.141 billion, a year-on-year increase of 6%, net income per share of NT$74.59, a big gain of seven share capital. After-tax net income of 74.59 yuan, a big profit of 7 share capital.

Overall Outlook for the Year

Looking ahead to 2023, MediaTek's CEO, Mr. Li-Hsing Tsai, said that the overall global economic environment remains uncertain, and that he hopes to observe the degree of recovery in demand in the next few months in order to provide a full-year revenue outlook for this year. Cai Lixing said that most customers are approaching their normal inventory levels with a conservative view of market demand and prudent inventory management. With the lifting of the blockade in China and the relative stabilization of the global economy, we believe that demand visibility will gradually improve in the next few months, and MediaTek's business will have a chance to turn around from the second quarter onwards. Demand for TVs and WiFi saw a slight rebound in the first quarter; Li-Hsing Tsai said the current quarter is expected to be a low point for the business, and that it will continue to be price disciplined to protect profitability in such an environment.

Cell Phone Industry Trends Markets Watching Closely

While global smartphone shipments are expected to decline slightly this year, 5G penetration is expected to grow from nearly 50% last year to about 55%, driven by 5G handsets in emerging markets such as India and Southeast Asia, and MediaTek expects to continue to increase its share of flagship handsets this year, as more flagship chips are introduced and penetrated into more markets. The flagship phones of Vivo X90 and X90 Pro, which utilize MediaTek's latest Breguet 9200 chipset, have been well received by consumers and are outselling their predecessors, while the Breguet 9000+ has been introduced to several flagships.Foldable Cell PhoneIn terms of non-handset business, Tsai believes that demand for Wi-Fi and broadband will gradually recover. In terms of non-handset business, Li-Hsing Tsai believes that the demand for Wi-Fi and broadband will gradually recover, and the market share is expected to be maintained or even better. Wi-Fi 7 related revenue is expected to be realized in the first quarter, and it is expected that the volume will be further enlarged in 2024, and the Wi-Fi 7 layout will be just like the 5G, and MediaTek will gradually enter the high-end market.

2022Annual Operating Performance

MediaTek's annual revenue last year was NT$548.796 billion, an annual increase of 11.2%, and net income per share after tax was NT$74.59, a record high. MediaTek said, especially cell phones, smart device platforms and power management chips have been growing for four consecutive years, which is the result of the solid implementation of the technology leadership and global expansion strategy over the past few years, and at the same time, the establishment of industry-leading product portfolio and a more diversified portfolio of global customers to consolidate a strong foundation.

Ji-Pu Viewpoint:Although CEO Li-Hsing Tsai said that he will continue to monitor the degree of demand recovery in the coming months due to the uncertainties in the global economic environment, many foreign investors such as JP Morgan and UBS are optimistic that the first quarter will be the bottom of the barrel. However, many foreign investors, such as JP Morgan and UBS, are optimistic that the first quarter will be the bottom of the barrel. The mainland smartphone market is expected to be stronger than expected in 2023, and some OEMs are already planning to raise their component pull and shipment targets after Q1. It is expected that Q2 revenue will recover moderately compared to Q1.

Jan W3｜Technology News Highlight：Samsung continues to slash orders as storm sweeps through supply chain

The memory market is in bad shape and the storm is spreading to the upstream silicon wafers. South Korean tech giant Samsung is rumored to cut down on silicon wafers purchasing, which will affect the operation of global silicon wafer factories. In addition, Samsung is also bad about this year's performance, rumored to be interested in reducing its investment in foundry, implying that it will first fill the capacity of its own factories as a top priority for this year's low demand for chips.

Ji-Pu Viewpoint:The memory market is in the midst of a storm that is spreading to the upstream silicon wafers. South Korean tech giant Samsung is rumored to be cutting back on its silicon wafer purchases, which will affect the operations of silicon wafer makers around the world. Samsung believes that the weaker-than-expected demand for memory is mainly due to the continued high global interest rates and weak economic outlook, which has triggered a deterioration in consumer confidence and led to further tightening of customer finances and inventory adjustments, while SK Hynix also said that the memory industry is facing an unprecedented deterioration in the market, and will therefore cut its capital expenditures by half this year. Among the Taiwanese factories, Global Crystals directly supplies Samsung wafers, which has the biggest impact. Globalchip disclosed that about half of the customers want to discuss the fine-tuning of the 8-inch and 12-inch product shipment schedule, but Globalchip is currently planning to maintain a high capacity utilization rate this year, so as not to waste capacity. In addition, Samsung is also bad this year's performance, rumor has it that it intends to reduce its investment in foundry, meaning that its demand for chips this year highlights the downturn, it will first fill its own factory capacity as the first priority. According to industry sources, its foundry investment spending is likely to be lower than last year's, and is estimated to be about the same as the 1.2 trillion won ($9.7 billion) it spent in 2020 and 2021. In the past, close cooperation with Samsung UMC, due to the recent foundry mature process market conditions headwinds, if the outsourcing orders from Samsung to shrink, the fear of facing the fear of adding to the misery.

Dec W4｜Technology News Highlight: Infineon to spend billions on mergers and acquisitions to create new business opportunities

Even if the global chip industry downturn, the German semiconductor giant Infineon (Infineon) still intends to spend billions of euros in acquisitions, in order to expand the product line, to create new operational energy, including power semiconductor (power semiconductor), sensor (sensor), software and artificial intelligence, CEO Jochen Hanebeck said he hoped to expand the product portfolio through mergers and acquisitions to strengthen the ability of Infineon in different areas of the industry. Hanebeck, CEO of Infineon, said the company wants to expand its product portfolio through acquisitions to strengthen Infineon's capabilities in different areas of the industry. The company reported significant growth of 63% in operating profit to €3.4 billion (US$3.6 billion) by the end of September 2022, driven by electric vehicles, autonomous driving, renewable energies, data centers, and Internet of Things (IoT)-related applications, and expects sales of €15.5 billion by 2023 (up or down by €500 million). The market is observing destocking through the first half of the year, with product sales yet to be confirmed in the second half of the year.

Corporate News｜Tesla's Q1 2024 law says

Apr W4｜Technology News Highlight : UMC throws a shockwave! Withdrawal from UMC's Board of Directors

Apr W4｜UMC Throws a Shock Bomb! Withdrawal from UMC's Board of Directors

Ji-Pu's point of view.

Apr W3｜SK Hynix and TSMC Collaborate to Develop Next-Generation HBMs

Ji-Pu's point of view.

Apr W2｜Intel's foundry suffers $220 billion loss

Ji-Pu's point of view.

Apr W1｜SK Hynix plans to build a packaging plant in the U.S.

Ji-Pu's point of view.

Mar W2｜Huang Yanxun Showcases New Technology at GTC

Ji-Pu's point of view.

Mar W1｜Leaderboard WiFi 7 takes a big step forward

Ji-Pu's point of view.

Feb W3｜Cheng-Yi Sun creates AI chip company to rival Nvidia.

Ji-Pu's point of view.

Feb W2｜Photonix Establishes New Division for Customized AI Chips

Ji-Pu's point of view.

Feb W1｜Canon leverages ASML for chip manufacturing equipment

Ji-Pu's point of view.

Jan W4｜Intel joins hands with UMC for 12 nanometers

Ji-Pu's point of view.

Jan W3｜Intel says German 1.5 nm still not as good as customer's perspective

Ji-Pu's point of view.

Jan W2｜TSMC's revenue dominated last year, surpassing Samsung's.

Ji-Pu's point of view.

Jan W1｜Intel Launches New Company to Focus on Generative AI

Ji-Pu's point of view.

Dec W3｜Taiwan TSMC Kumamoto Plant JASM will be completed soon.

Dec W2｜Intel launches new AI chip to capture market next year

Dec W1｜Eckel to set up $2 billion advanced packaging and testing plant in the U.S.

Nov W5｜Next-generation Intel processor to be mass-produced in 3nm by TSMC

Nov W4｜South Korea's memory chip exports grow for the first time in more than a year

Nov W3｜Intel 18A will enter trial production in Q1 next year.

Nov W2｜AMD's financial report is good, next year AI is the key point.

Nov W1｜Qualcomm unveils AI chips for PCs and mobile phones, setting the stage for new competition

Oct W3｜Canon develops NIL device to challenge EUV microfilm technology

Oct W2｜Intel Spins Off PSG Division, Intends to Go Public in 3 Years

Oct W1｜Intel adopts EUV equipment for mass production of wafers in Ireland

Sep W4｜Intel Launches First Glass Substrate, Revolutionizing Packaging

Sep W3｜Silicon photonic chips will be the next wave of AI.

Sep W2｜Huawei launches new handset with Kirin 9000S chip

Sep W1｜Intel Turns to TSMC, Outsourcing Up to $19 Billion

Aug W4｜ARM rides on the AI train with an IPO

Aug W3｜Qualcomm rumored to lay off 200 people, and those who stay will have to accept bonus reduction and no salary adjustment.

Aug W2｜Tokyo Mighty: China buys mature process wafer equipment

Aug W1｜Taiwan TSMC to expand into Europe and set up a plant in Germany.

Jul W4｜Heterogeneous Integration Breakthrough Stressor Support for IC Packaging

Jul W3｜Intel to Attack China's AI Market: Sunrise, Unimicron, Sentec and Other Taiwan Chains to Complement

Jul W2｜Inventory Adjustment: IC Design Firms Seek Quarterly Increase in Results

Jul W1｜U.S.-Dutch Joint Effort Against Mainland Chips

Jun W4｜Taiwan TSMC's operation will be stronger in the second half of the year.

Jun W3｜Technology News Highlight：Three Advantages...Hon Hai Group

Jun W2｜Sun Micron Receives Advanced Packaging Order from TSMC

Jun W1｜Tech News Highlight：AI demand soars! NVIDIA to build supercomputers in Israel

May W4｜Technology News Highlight：Semiconductor shocked by Mainland China's ban on Micron

May W3｜Technology News Highlight: MediaTek's 6G NTN Global Seamless Overlay

May W2｜TSMC 3nm to be highly mass-produced

May W1｜TSMC Launches New Software to Accelerate the Adoption of 3nm Processes in Automotive ICs by 2025

Apr W4｜ Technology News Highlights: Creative orders strong, optimistic year-round operations

Apr W3｜ DDIC Major Manufacturers Show Growth Momentum as Demand Recovers

Related News:

Apr W2｜Intel's New Generation GPUs: TSMC Eats It All!

Apr W1 | Samsung to cut memory production, industry applauds

Mar W4 | Security Transfer Authorization Shocks IC Designs

Mar W3 | OpenAI Re-introduces Language Model GPT-4 with Improved Academic Performance but Pitfalls Remain

Mar W2｜IC design houses expected to bottom out in Q1

Mar W1 | Technology News Highlights: U.S. Semiconductor Grants Set Profit Conditions

Feb W3 | Technology News Highlight: Microchip Expands Silicon Carbide 8-inch Wafer Fab and Completes Third-Generation Semiconductors

Related News:

Feb W2 | Technology News Highlight: US-China tech war to escalate again, will restrict investment in China's technology

Feb W1 | Technology News Highlight: 2022 Q4 MediaTek Press Conference

Jan W3｜Technology News Highlight：Samsung continues to slash orders as storm sweeps through supply chain

Dec W4｜Technology News Highlight: Infineon to spend billions on mergers and acquisitions to create new business opportunities

-For more information, please clickContact Us-

Share the article now!

Related Posts

Corporate News｜Tesla's Q1 2024 law says

Corporate News｜FSMC 2024 Q1 Press Conference

Apr W4｜Green Energy News Highlight: Tesla removes old roadblocks in China, but new shadows loom in the U.S.

AprW4｜Big Health News Highlight: TaiGen*-KY's new influenza drug succeeds in Phase III blinding, effective in treating influenza