News

US senators recently proposed the "Safe and Feasible Chip Export Act", requiring the US Department of Commerce to completely suspend the issuance of advanced artificial intelligence chip export licenses to countries such as China and Russia within the next 30 months. Recently, Republican Senator Pete Ricketts and Democratic Senator Chris Coons jointly proposed the "Safe and Feasible Chip Export Act", which requires the US Department of Commerce to completely suspend the issuance of advanced artificial intelligence chip export licenses to countries such as China and Russia in the next 30 months, and explicitly include Nvidia H200 chips and new chips based on the Blackwell architecture in strict control. As a key computing carrier supporting large-scale artificial intelligence model training, the closure of the export channels for the above-mentioned products is widely regarded by international observers as an important policy tool for the US government to suppress the upgrading of China's artificial intelligence industry from the source of technology supply.   The "Congressional Supervision Clause" set up in the bill has more institutional features, stipulating that after the 30 month policy implementation period expires, any rule adjustments made by the Ministry of Commerce must be reported to Congress 30 days in advance. This will strengthen the institutional rigidity of the export control of chips to China. The joint proposal by senators from both parties essentially forms legislative constraints on the executive branch's policy easing tendency, continuing the evolving characteristics of the US chip control policy towards China, which alternates between looseness and tightness. This policy dynamic can be traced back to early 2025, when Nvidia's customized H20 chip for the Chinese market was included in the regulatory list. After a brief policy adjustment, the regulatory system is now further strengthened through legislative procedures. Facing the chip export control policy, AMD CEO Su Zifeng announced on December 4th local time that some of AMD's MI 308 chips have obtained permission to be exported to China and are ready to pay a 15% tax to the US government when shipped. Previously, AMD had estimated that the US government's export restrictions on MI 308 chips would result in a loss of approximately $800 million. NVIDIA continues to be at the forefront of opposition. Huang Renxun has repeatedly emphasized publicly that "the Chinese market cannot accept technology products with limited performance", and through quantitative analysis, pointed out that the continuous export restrictions have caused the company to lose $15 billion in revenue in China. If the new law takes effect, it is expected that the revenue in the Chinese market will return to zero in the next two quarters. And just one day before the proposal of the Safe and Feasible Chip Export Act, Nvidia successfully lobbied Congress to suspend the GAIN AI Act. This core argument, which emphasizes that "artificial intelligence technology is different from nuclear weapons and should promote global technology sharing," emphasizes the important supporting role of technology circulation in the US job market and industrial ecology. However, this business oriented policy proposition has not gained widespread recognition.   Republican Congressman John F. Kennedy of the United States Senate Banking Committee publicly questioned the objectivity of his proposed policies, arguing that his views overly focused on corporate commercial interests; Hardliners, represented by Steve Bannon, advocate for strict measures similar to the Cold War era nuclear technology control. From a global perspective, this chip regulation game is essentially a strategic competition for technological dominance, but it may result in a "lose lose" situation. The restricted flow of technology will slow down the pace of global artificial intelligence technology innovation, while supply chain fragmentation will drive up industry operating costs. At a critical stage of the development of artificial intelligence technology, seeking a balance between maintaining national security and promoting global technological cooperation tests the strategic decision-making wisdom of China and the United States. The ultimate direction of this game will have a profound impact on the future global technological development pattern and international economic order.

Global semiconductor giant Samsung Electronics recently announced a major organizational restructuring, disbanding its high bandwidth memory (HBM) special development team that had only been established for one year, and integrating the team's personnel and overall business into the design team of the DRAM (Dynamic Random Access Memory) development department. Adjustment details: Smooth transition of personnel and business. According to multiple media reports, this adjustment marks Samsung's HBM business returning from an independent "special attack team" model to the mainstream research and development system of storage chips. Samsung Electronics stated that this move aims to optimize resource allocation, enhance synergies between HBM and core DRAM products, and improve overall operational efficiency and competitiveness.   Samsung Electronics urgently established an independent HBM development team in July 2024. At that time, Samsung was clearly lagging behind its main competitor SK Hynix in the rapidly growing HBM market, and the establishment of this team was seen as a key measure to concentrate resources and accelerate technological catch-up. HBM is a core component in the field of artificial intelligence (AI) servers and high-performance computing, and market demand has exploded with the AI wave. According to the adjustment plan, all members of the original HBM development team will be transferred to the design team under the DRAM development department. Son Young soo, the Vice President (EVP) who previously led the HBM team, has been appointed as the head of this design team and will continue to lead the original team in advancing the research and development of the next generation of HBM products.   Despite the change in team organization, the core research and development tasks, including the development of cutting-edge technologies such as HBM4 and HBM4E, will remain continuous. Industry analysis indicates that this restructuring reflects a change in Samsung's assessment of the development stage of its HBM technology. After more than a year of concentrated research and development, Samsung believes that its HBM technology has gained core competitiveness and no longer needs to "make up for shortcomings" in the form of an independent team. Instead, it is shifting towards deeper integration with basic DRAM technology to strengthen long-term technological advantages. This move is also interpreted as Samsung having full confidence in the upcoming next-generation HBM products, such as HBM4. Market impact: Consolidating position and catching up with market share. Market data shows that in the second quarter of this year, Samsung's market share in the global HBM market fell to 17%, ranking third, far behind SK Hynix, which holds 62% market share. However, with Samsung's HBM3E successfully entering Nvidia's supply chain and all HBM4 product orders sold out, the company is optimistic about its market performance next year. Samsung's Executive Vice President of Storage Business, Kim Jae joon, revealed during a recent earnings conference: 'Compared to this year, we have significantly increased our HBM production capacity for next year, but customer demand still exceeds supply.'. The company expects that as the supply scale of HBM4 gradually expands, its market share will steadily rebound from next year. Market research firm TrendForce predicts that Samsung's market share in the global HBM market is expected to exceed 30% by 2026. It is worth noting that this organizational adjustment coincides with the global storage market entering a super cycle. According to supply chain sources, the price of DDR5 16Gb chips increased by as much as 102% in October, while Counterpoint Research predicts that by the second quarter of 2026, memory module prices will rise by 50% compared to the current level. While consolidating the traditional DRAM market, Samsung continues to lay out its high-end HBM product line, demonstrating its dual track strategy of "bread and the future".   At present, Samsung has a strong foundation in DRAM production capacity and order fulfillment. In the HBM field, its HBM3E product has achieved mass production and been delivered to customers such as Nvidia and AMD, and orders for the next generation HBM4 have also been sold out. Although Samsung ranked third in global HBM market share in the second quarter of this year, behind SK Hynix and Micron, market research firm TrendForce predicts that with the expansion of HBM4 supply, Samsung's market share is expected to exceed 30% by 2026. This organizational adjustment is Samsung's strategic layout to more effectively integrate internal resources, seize market opportunities, and achieve market share growth. The Korean Economic Daily reported that Samsung's HBM4 product is rapidly expanding its market share and benefiting from a 56% annual increase in standard DRAM prices, resulting in a significant increase in overall storage revenue; In addition to the rebound in storage prices, the price of NAND Flash has also rebounded, which will help improve the overall gross profit structure. It is reported that Samsung Electronics plans to complete the organizational restructuring this week and is expected to hold a global strategic meeting in early December to review next year's business plans.

On October 28, 2025, Skyworks Solutions and Qorvo jointly announced the final merger agreement. This transaction, completed in the form of "cash+stock," will create a new giant in the RF industry with a total value of up to $22 billion (including corporate debt) and annual sales of approximately $7.7 billion. After the merger, the new company is expected to save over $500 million in operating costs annually and will continue to be listed on NASDAQ under the name Skyworks Solutions, Inc., with the stock code SWKS remaining unchanged. Image: Skyworks and Qorvo merger announcement The announcement shook the industry. These two companies, which have been deeply rooted in the RF front-end field for more than 20 years and have long occupied the top two global market shares, have finally merged. According to Yole's statistics, although Qualcomm and Broadcom have larger sales volumes from an overall revenue perspective, their business models have particularities: Qualcomm's advantages largely rely on the bundled sales of its mobile platform, while Broadcom's business is highly concentrated in Apple, and a large part of its revenue comes from the Wi Fi/Bluetooth module systems it provides for Apple phones. In contrast, Skyworks and Qorvo have always maintained an absolute leading advantage in the open market, especially in the Chinese market, and are regarded as the true Top 2 in the RF field. Figure: Distribution of RF front-end market In fact, the integration of the RF industry is not a new phenomenon. Looking back at the development history of the industry, multiple major mergers and acquisitions have already reshaped the market landscape: In 2015, RFMD merged with Triquint to form Qorvo; In 2016, II-VI Company (later renamed Coherent), a manufacturer of optoelectronic and semiconductor components, completed its acquisition of Anadigics, and Anadigics withdrew from the historical stage. In addition to the integration between these top manufacturers, the RF industry has also undergone multiple cross-border mergers and acquisitions: In 2001, TriQuint strengthened its core technology in the field of RF filters by acquiring Sawtek; In 2009, passive component supplier TDK integrated with Epcos, and in 2017, it was incorporated into RF360, a company established by Qualcomm. In 2015, Skyworks acquired Panasonic's filter solutions business; Although mergers and acquisitions are commonplace in the RF industry, the merger of Skyworks and Qorvo still brings unprecedented impact, mainly reflected in three fundamental changes: Strategic motivation shift: Unlike the previous "strong alliance" aimed at strengthening competitiveness, this merger is a strategic contraction between the two leading companies, marking the industry's shift from expansion thinking to efficiency priority; The supply chain has a profound impact: as a major supplier of RF solutions for Chinese mobile phone manufacturers, especially flagship models, Skyworks and Qorvo originally formed a balanced pattern of mutual backup. The merger of the two will profoundly affect the stability and bargaining power of China's mobile phone supply chain; Technological development path reconstruction: In the past two decades, the RF industry has basically followed the development path of "Skyworks and Qorvo technology leading", "adoption by Chinese mobile phone manufacturers", and "follow-up by Chinese RF enterprises". This merger will break this established order and force the industrial chain to seek new development logic. Why did Skyworks and Qorvo, two RF giants, choose to turn at this moment? How will the merger affect the future direction of China's RF industry? This article will attempt to conduct in-depth analysis around these issues. The formation of order in the RF front-end industry The RF front-end industry in which Skyworks and Qorvo operate is an essential core component of wireless communication devices. The RF front-end is located before the transceiver and mainly consists of four modules: power amplifier (PA), low noise amplifier (LNA), filter, and switch. Every wireless terminal cannot do without these key components.   Figure: Functional composition of RF front-end According to market research institutions such as Yole, the global RF front-end market size is expected to reach approximately 15.4 billion US dollars by 2025, equivalent to over 100 billion yuan in Chinese yuan. This is one of the few sub sectors in the semiconductor industry with a scale exceeding 100 billion yuan, and the market capacity is huge. Despite the large market size, the industry has a strong sense of order. Both the technical roadmap and product solutions exhibit highly unified characteristics. This feature is closely related to the application scenarios of RF front-end. Its core customers are mobile phone manufacturers, and the global mobile phone market has an annual shipment volume of over one billion units, with a large and highly stable scale. RF performance directly affects the communication experience of end users. Currently, mobile phone brands are highly concentrated, including Apple, Samsung, as well as a few giants in China such as Vivo, Huawei, Honor, Xiaomi, OPPO, etc., each with an annual shipment volume of tens of millions or even hundreds of millions of units. In this context, the reliability and performance of the RF front-end, as the core chip of the terminal, cannot be compromised. Another notable feature of the mobile phone market is rapid iteration. As a consumer electronics product, mobile phones are frequently updated, and the continuous evolution of communication technology from 2G to 5G, as well as the advancement of 6G in the future, require continuous upgrades of RF front-end to adapt to new protocol standards. In such an industrial environment, the evolution model of "platform driven definition and certification" has emerged and gradually become the mainstream solution in the market. To coordinate the technological evolution between mobile phone manufacturers and RF solutions, SoC platform manufacturer MediaTek (MTK) has taken the lead in introducing the "third-party definition" mechanism. MediaTek collaborates with RF front-end suppliers to define standard solutions based on customer needs and its own platform planning. After determining the plan, the RF manufacturer and platform provider will synchronize their development work to provide complete chip and RF solutions for end customers who choose the MediaTek platform. Figure: Ecological conduction path diagram of RF front-end solution This path seems to achieve a win-win situation for RF manufacturers, platform manufacturers, and terminal manufacturers, but the platform's selection of RF suppliers becomes particularly crucial in the initial stage. Leading manufacturers can lay out designs in advance, launch products first, solve problems in a timely manner, and can be included in official reference designs for collaborative promotion. However, once mobile phone manufacturers adopt a reference design for development, it is often difficult to invest additional resources to replace the solution in the future. For a long time, the first batch of cooperative RF manufacturers selected by platforms such as MediaTek have been limited to American and Japanese manufacturers such as Skyworks and Qorvo, and domestic manufacturers have been excluded. This means that domestic manufacturers can only seek limited opportunities for adaptation and replacement after their solutions have matured and become commercially available, posing significant challenges for the development and introduction of domestic products. Simultaneous homogeneous development has become a key bottleneck that has long constrained the development of domestic RF enterprises. The above is the basic pattern of the current RF industry order. However, the formation of any order is not a one-day effort. Looking back at the development history of the global RF front-end over the past forty years, its order evolution can be roughly divided into three stages: Initial stage: Maintaining order through technological barriers (1985-2012) Mid stage: Maintaining order through integration of design capabilities and solutions (2013-2022) Recent stage: Maintaining order through market business strategies (since 2023) The three stages and the changes in revenue of the two companies are shown in the following figure. Figure: The correspondence between the three stages of RF order formation and Skyworks and Qorvo revenue 1. Initial stage: Maintaining order through technological barriers (1985-2012) In the era of rapid development of wireless communication technology, many emerging companies have emerged with innovative technologies and market opportunities. However, enterprises that can survive and achieve long-term development in fierce competition not only need to master advanced technology, but also need to launch products that meet market demand at critical moments. Anadigics, Triquint, RFMD, Skyworks, and Avago were among the top performers, ranking among the top 5 in the GaAs PA market at the time. Figure: 2008 Global GaAs PA Supplier Ranking Anadigics was founded in 1985, initially focusing on power amplifiers in cable television (CATV) systems. With the rise of the mobile communication market, the company quickly turned to this field and launched a series of high-performance PA products based on its existing technological accumulation. Its highlight lies in the high-efficiency and high-performance amplifiers, especially the HELP ™ Energy saving technology. Anadigics adopts indium gallium phosphide (InGaP) material and HBT structure, and further introduces BiFET process, which is described in the technical paper "InGaP Plus ™:  Advanced GaAs BiFET Technology systematically describes the progressiveness of this composite process. Image: Anadigic's BiFEM process and HELPTM technology for improving PA performance Triquint was also founded in 1985, spun off from Tektronix, inheriting its profound accumulation in compound semiconductor processes such as indium phosphide (InP) and gallium arsenide (GaAs). Since its establishment, the company has established a development path centered on advanced technology, and has established significant advantages in fields such as GaAs and GaN, becoming an important supplier of military RF microwave devices in the United States. Triquint has also proactively deployed gallium nitride (GaN) technology and participated in relevant projects of the US Defense Advanced Research Projects Agency (DARPA), building a solid technological barrier for defense and base station applications. Unlike Anadigics, Triquint not only focuses on active circuits, but also actively integrates passive devices, expanding its business to the entire RF front-end by mastering surface acoustic wave (SAW) and bulk acoustic wave (BAW) filter technologies. The world's first PAMiD (power amplifier module and integrated duplexer) module, launched in 2003, became an important milestone in the development of RF modularity. Figure: (a) Triquint's module product TQM71312 launched in 2003 (b) Triquint's explanation of its module products RFMD was founded in 1991, coinciding with the global promotion of the GSM standard. The company focuses on RF microdevices and decisively chooses the GaAs HBT process route. The advantages of this technology in terms of linearity, efficiency, and power density have quickly made it the mainstream choice for mobile PA in the 2G to 4G era. RFMD achieved a virtuous cycle of technological leadership and scale expansion by binding with mainstream mobile phone brands such as Motorola and Nokia, and relying on its own IDM model. In contrast, AnaDigics, which used GaAs MESFET technology in the early days, gradually lost competitiveness in the consumer electronics market in terms of cost and ease of use due to the negative voltage supply required for the devices. Despite the later shift towards HBT technology, RFMD has taken the lead with its first mover advantage and large-scale production capacity, and MESFET technology has gradually withdrawn from the mainstream civilian PA market. Skyworks was established in 2002 through the merger of Alpha Industries and Conexant Systems' wireless division. The merger brings significant technological and business synergies: Alpha's accumulation in semiconductor processes, combined with Conexant's expertise in wireless systems, gives Skyworks the ability to define system level RF solutions. The company clearly focused on the RF field and decisively withdrew from the baseband business in 2006, concentrating resources on deepening GaAs HBT technology and modular integration. By continuously optimizing processes and driving architectural innovation, Skyworks not only pushes the performance of GaAs HBT devices to new heights, but also combines them with system level packaging technology to launch a series of highly integrated front-end modules, laying the foundation for the evolution of 4G/5G terminal RF architecture. The development path of Avago is also representative. Its predecessor was Agilent Semiconductor Division, which operated independently after being acquired by private equity funds in 2005. The company focuses on III-V compound semiconductor technologies such as GaAs and has developed differentiated advantages in devices such as FBAR filters. Through a series of precise mergers and acquisitions and technological integration, Broadcom eventually became a global semiconductor giant, and its development process reflects the dual drive of technological depth and strategic mergers and acquisitions. The above five companies, with their respective focus and breakthroughs in key technologies such as processes, modularization, system integration, or filters, have jointly occupied about 66% of the global GaAs market share by 2008, laying the foundation for a relatively stable market pattern in the following two decades. During this period, a number of RF startups emerged in both China and the United States. On the Chinese side, RDA、 A large number of RF startups such as Weijie Chuangxin, Huizhi Microelectronics, Zhongpu Microelectronics, Hantianxia, Haoxin, Guomin Feixiang, and Zhengyuan Microelectronics have been established one after another, attempting to replicate the successful path of American enterprises. However, due to its relatively late start compared to American companies and immature industrial chain, it has not yet had a substantial impact on leading enterprises. Start up companies such as Amalfi, RFaxis, SiGe, Cavendish have also emerged in the United States, most of which were acquired by existing giants between 2010 and 2020. The innovative technologies of these startups are absorbed and integrated, further strengthening the technological barriers of large enterprises. Image: A US startup RF company acquired by a giant Overall, during this stage, the top five companies in the industry have led and maintained the early formation of the global RF front-end order by building and continuously strengthening their technological advantages. 2. Mid term stage: Maintain order by integrating design capabilities and solutions (2013-2022) With the increasing maturity of compound semiconductor (GaAs HBT) foundries (such as WIN Semiconductors) and the accelerated diffusion of RF technology, the solutions launched by Chinese manufacturers are gradually reaching a "usable" level. In the face of this challenge, the response of international manufacturers such as Skyworks and Qorvo is to build a deeper competitive system: with continuous design technology innovation as the cornerstone, platform reference design as the core carrier, high integration modules (such as PAMid/L-PAMid) as barriers, and strategic mergers and acquisitions to supplement key capabilities. 1）. The cornerstone of order: the continuous leadership of RF design technology At this stage, the core competitive strategy of international manufacturers has shifted from the gap in technology to achieving sustained leadership through design and technological innovation. Faced with the gradual diffusion of process technology, enterprises such as Skyworks and Qorvo have not only consolidated their performance advantages but also laid a solid foundation for the subsequent construction of platform ecology and module barriers through systematic innovation in core design aspects such as circuit architecture, packaging forms, and power supply systems. Specifically, the application of a series of key technologies in the RF front-end of mobile terminals collectively supports this leading position. Specifically, it includes: Flipchip packaging achieves a dual improvement in RF performance and integration potential by replacing traditional Wirebond; The introduction of advanced circuit architectures such as Push pull and Doherty optimizes the linearity and efficiency of power amplifiers; Envelope tracking (ET) technology improves power efficiency at the system level;

"Omdia reported today that the market size of cloud infrastructure services in Chinese Mainland will reach US $12.4 billion in the second quarter of 2025, the first quarterly year-on-year growth rate of more than 20% since the beginning of 2024 (specific proportion is 21%). ” Omdia reported today that the market size of cloud infrastructure services in Chinese Mainland reached US $12.4 billion in the second quarter of 2025, the first quarterly year-on-year growth rate of more than 20% since the beginning of 2024 (specific proportion is 21%).   In terms of market share, Alibaba Cloud, Huawei Cloud, and Tencent Cloud have market shares of 34%, 17%, and 10% respectively in Q2 2025, with the remaining 39% shared by other service providers.   Overall, the exceptionally strong demand for AI played a key role in the rebound of Q2 growth. The demand for enterprise AI is shifting from basic model calling to application scenarios that better meet business needs, including early exploration of industry-specific models and AI agent/agent applications.

The revenue and upcoming sales of chip giant Nvidia have exceeded Wall Street's expectations, easing investors' concerns about market turbulence caused by massive spending in the field of artificial intelligence (AI). The company's quarterly financial report released on Wednesday showed that revenue surged 62% to $57 billion in the three months ending in October, mainly due to the increasing demand for its chips in artificial intelligence data centers. The sales of this department increased by 66%, exceeding $51 billion. Reaching $51.2 billion, easily exceeding analysts' previous expectations of $49.09 billion, a 25% increase from the previous quarter and a 66% increase from the same period last year. The remaining $6.8 billion in revenue comes from Nvidia's gaming business ($4.2 billion), followed by sales in the professional visualization and automotive sectors. According to the Generally Accepted Accounting Principles (GAAP) in the United States, the company's net profit was $32 billion, a year-on-year increase of 65%. Both revenue and profit exceeded Wall Street's expectations.   Nvidia stated that most of the growth was driven by initial sales of its GB300 chips. The network business, even components that can work together like a computer with dozens of GPUs, contributed $8.2 billion in sales to data centers. Nvidia Chief Financial Officer Colette Kress stated in a statement that the company's best-selling chip series is now the Blackwell Ultra, which is the second generation version of the company's Blackwell chips. In a statement to shareholders, Nvidia Chief Financial Officer Colette Kress pointed out that the accelerated development of computing power, powerful AI models, and the rise of intelligent agent applications have driven the growth of the company's data center business. Kress stated during the company's third quarter earnings conference call that last quarter, the company announced an AI factory and infrastructure project with a total of 5 million GPUs. Kress said, "This demand covers all markets, including cloud service providers, sovereign states, modern construction enterprises, and supercomputing centers, and includes multiple iconic construction projects Blackwell Ultra GPU was released in March, offering a variety of configuration options and performing particularly strongly. It has now become the leading product within the company. According to the company, previous versions of the Blackwell architecture have also maintained strong demand. Nvidia stated that it expects sales for this quarter to be around $65 billion, compared to analysts' previous expectations of $61.66 billion. The company stated that its net profit for this quarter increased by 65% to $31.91 billion, or $1.30 per share, compared to $19.31 billion, or 78 cents per share, in the same period last year. Nvidia is the world's most valuable company and is seen as a barometer of the prosperity of artificial intelligence. The performance of this chip manufacturer may affect market sentiment. CEO Huang Renxun stated in a statement that the sales of his artificial intelligence Blackwell system "far exceeded expectations" and that "cloud GPUs (graphics processing units) are sold out". Huang Renxun stated in the company's third quarter financial report that "Blackwell's sales far exceeded expectations, and the cloud GPU has also been sold out. The demand for computing continues to accelerate exponentially in terms of training and reasoning. We have entered a virtuous cycle of artificial intelligence. The artificial intelligence ecosystem is rapidly expanding - with more new foundational model builders emerging, more AI startups covering more industries and countries. Artificial intelligence is ubiquitous and capable of anything. ” "There are many opinions about the artificial intelligence foam. But from our perspective, what we see is quite different," he said in a conference call with analysts. We have performed excellently in all stages of artificial intelligence He has stated that people's willingness to pay for artificial intelligence tools indicates that the technology is "profitable," even though most tech companies are now reinvesting the money they make in new infrastructure. Due to growing concerns about overvaluation of artificial intelligence stocks, the quarterly report of this chip manufacturer has attracted more attention on Wall Street than ever before. Due to concerns about the return on investment in artificial intelligence, the S&P 500 index fell for four consecutive days before Wednesday. Before Nvidia released its performance, people had high expectations for its performance. LPL Financial's Chief Technology Strategist Adam Turnquist stated that the question is not whether the company will exceed expectations, but rather how much it will exceed. Despite artificial intelligence valuation dominating the headlines, Nvidia continues to conduct its business in an elegant manner, "said Matt Blitzman, a senior stock analyst at Hargreaves Lance. He said that the valuation of certain areas in the field of artificial intelligence "needs to take a breath, but Nvidia does not belong to this category". Huang Renxun previously stated that he expects artificial intelligence chip orders to reach $500 billion by next year. Investors are paying attention to when the company expects to achieve these revenues and how to fulfill these orders. Nvidia Chief Financial Officer Colette Kress told analysts that the company "may" take on more orders than the $500 billion already announced. During the earnings conference call, she also creatively listed the highlights of recent financial reports from chip manufacturer partners, emphasizing that artificial intelligence is bringing returns to the company. For example, she mentioned that Meta's artificial intelligence recommendation system allows users to "spend more time on apps like Facebook and Threads"; Anthropic recently announced that it expects its annual revenue to reach $7 billion this year; Salesforce's engineering team has improved efficiency by 30% since using artificial intelligence for coding. She also cited a series of examples of corporate clients. But she also expressed disappointment with regulatory restrictions that hinder the company from exporting chips to China, saying that the United States "must win the support of every developer, including Chinese developers". Kress stated on the earnings conference call, "Due to geopolitical issues and increasingly fierce competition in the Chinese market, we were unable to secure large purchase orders this quarter. Although we are disappointed with the current situation that is hindering our ability to export more competitive data center computing products to China, we are committed to continuing to communicate with the governments of the United States and China, and will continue to advocate for enhancing the United States' competitiveness globally Tech giants are increasing their investment in artificial intelligence, competing to profit from the boom that has driven stock prices to historic highs. The financial reports released by Meta, Alphabet, and Microsoft last month once again confirmed that these companies have invested huge amounts of money in various aspects from data centers to chips. Sundar Pichai, CEO of Alphabet, the parent company of Google, also stated that although the growth of AI investment is an "extraordinary moment," there are also some "irrational" factors in the current AI boom. His remarks were made at a time when other industry leaders were issuing warnings. The chips produced by NVIDIA are crucial for artificial intelligence data centers, as it has entered into a series of deals with key players in the field of artificial intelligence such as OpenAI, Anthropic, and xAI, with NVIDIA being at the core of these deals. These transactions have attracted much attention due to their cyclical nature, as mutual investments between artificial intelligence companies are increasing. These agreements include Nvidia investing $100 billion in OpenAI (the company behind ChatGPT).

The expansion of global investment in artificial intelligence (AI) has exacerbated the shortage of semiconductor DRAM, and price negotiations, previously conducted monthly or quarterly, are now shifting towards long-term supply contracts of six months or longer. Demand-side companies are also actively responding to six-month contracts as a severe DRAM supply shortage is expected to continue driving up prices next year. The market is even beginning to discuss supply contracts for 2027, as securing supply next year is becoming increasingly difficult.   On November 17th, a semiconductor industry insider explained, "The DRAM market has shifted to a long-term contract-driven market," adding, "The purchasing demand generated by this situation is stronger than the supercycle market of 2017." DRAM serves as temporary data storage, enabling central processing units (CPUs) and graphics processing units (GPUs) to process information quickly.   With the emergence of large language model (LLM) AI such as ChatGPT, the role of GPUs is becoming increasingly prominent. To support these models, memory semiconductors, including high-bandwidth memory (HBM) that employs multi-layered stacked DRAM, are facing supply shortages. A prime example is that major US tech companies, including NVIDIA, the world's largest AI semiconductor company, have secured HBM chips from SK Hynix and Samsung Electronics through long-term annual contracts.   However, recently, not only is there a shortage of HBM memory, but general-purpose DRAM is also facing supply shortages. This is because with OpenAI and Meta announcing AI infrastructure investment plans worth hundreds of billions of won, and with major companies and governments worldwide building data centers for their AI research, demand for general-purpose DRAM, including Double Data Rate (DDR), Graphics (G) DDR, and Low Power (LP) DDR, has surged. While general-purpose DRAM has lower performance than HBM, it is crucial for AI inference and computing.   According to industry insiders, DRAM demand is growing significantly, primarily concentrated in US and Chinese companies. In particular, Samsung Electronics and SK Hynix are reportedly signing six-month contracts with major demanders for supplies next year.   Typically, semiconductor DRAM supply contracts are signed monthly, with a fixed price each month, and then the product price is adjusted according to market prices. However, starting in the second half of this year, with the surge in semiconductor DRAM demand, contract cycles are shifting from quarterly contracts to six-month or longer supply contracts. This is because demanders are not only willing to offer prices higher than market rates, but also want to sign supply guarantee contracts for at least six months.   In fact, DRAM market sales are declining rapidly. Samsung Electronics' semiconductor division (DS) had finished goods inventory assets of 3.404 trillion won at the end of the third quarter, a 14.6% decrease (580.4 billion won) from the previous quarter. SK Hynix's inventory is also declining. The company's finished goods inventory assets in the third quarter of this year were 2.152 trillion won, a decrease of 368.9 billion won from the end of last year.   Due to the DRAM shortage, market discussions have even extended to supply contracts for 2027. SK Hynix has already secured all DRAM supply contracts for next year and is currently negotiating supply for 2027. With SK Hynix's DRAM sold out, demanders, including major global technology companies, are turning to Samsung Electronics. However, Samsung Electronics has already signed supply contracts for most of next year's production. With surging demand, Samsung Electronics is even discussing plans to raise DRAM supply prices by more than 40%.   Industry insiders predict that the super-cyclical performance of Samsung Electronics and SK Hynix will continue until at least 2027 as the DRAM market shifts towards long-term supply contracts. When manufacturers have more long-term supply contracts, production planning, including production costs and distribution, will become easier, and profits will increase accordingly. One industry insider explained, "Prices have risen again because semiconductor contracts that were originally signed monthly or quarterly are now renewed semi-annually." He added, "By 2027, long-term supply contract prices will be higher than current levels, further enhancing profitability." Distributors Unprecedentedly Mandate Bundling of Memory and Motherboards   According to Taiwan's *Economic Daily News*, a severe global shortage of DRAM memory has led some Taiwanese distributors to impose unprecedented bundling requirements on buyers. The report states that some channels now require customers to purchase motherboards and DRAM memory modules in a 1:1 ratio, or risk being unable to purchase memory at all.   This allocation control method is reportedly unprecedented in the DRAM industry. Distributors appear to be using the high demand for memory modules to boost motherboard sales, a strategy more common in the tight-supply consumer electronics market than in the hardware sector. ASUS, Gigabyte, MSI, and Chinese motherboard manufacturer Chaintech are reportedly direct beneficiaries of this practice.   Taipei-based financial analyst Dan Nystedt, known for translating and tracking Taiwanese tech industry media, relayed this news in a post on X, noting that the bundled sales policy "triggered a surge in motherboard sales."   Ultimately, this all reflects the rapid changes in the memory market since the beginning of the year. DRAM contract prices are currently up approximately 170% year-over-year, primarily driven by demand from AI server manufacturers. TrendForce recently raised its Q4 DRAM market growth forecast to 18%-23% quarter-over-quarter.   On the client side, mini-PC maker Minisforum recently increased prices for pre-installed configurations that include both DRAM and SSDs, while keeping prices for basic SKUs unchanged. The company explicitly stated that this move was due to a "significant increase" in its overall costs.   While the bundled sales strategy described in Taiwanese media currently appears to be limited to the Taiwanese market, it illustrates that the distribution of DRAM throughout the supply chain is tightening.   The report also notes that downstream buyers may face new hurdles as hyperscale data center and smartphone manufacturers continue to capture the majority of available capacity. Morgan Stanley Downgrades Ratings for Major OEMs   The continued DRAM memory shortage and price doubling or even more in recent weeks pose a significant challenge to potential PC assemblers and could lead to sustained price increases for computers and electronic devices for at least the next few years. According to X posts by @juklanosreeve (Jukan), Morgan Stanley market analysts believe even large manufacturers and integrators will be impacted, and have even downgraded their stock investment recommendations for some companies.   For reference, Morgan Stanley uses three ratings for stock performance forecasts: OW (Overweight, or Good), EW (Neutral, or Neutral), and UW (Underweight). Dell's rating was reportedly downgraded significantly from OW to UW, while HP, ASUS, and Pegatron were downgraded from EW to OW.   Other OEMs like Acer and Compal, already manufacturers of ultrawide devices, have also seen their target prices lowered by approximately 20% by Morgan Stanley. Dell faces harsher forecasts than other companies because it sells a large volume of servers, which typically consume significant amounts of memory.   In another chart, Microsoft estimates that memory costs account for 40% of the bill of materials (BOM) for high-end servers. General-purpose servers far fare similarly at 30%. Standard PCs and “AI” PCs (whatever the definition of “AI” today) account for 20% and 15% of the BOM, respectively.   If you're wondering why Apple is still rated OW, Jukan suggests that the Cupertino giant made large purchases before the full-blown DRAM shortage and also had a long-term agreement with Kioxia, presumably for which Kioxia produced some of its DRAM.   Given Apple's past experience in handling such crises and its consistent proactive approach, Jukan's speculation may not be unfounded. It's not hard to predict that even if prices for Macs, iPads, and iPhones rise, the increase will likely be small, and ample supply will remain in the short term.   Morgan Stanley seems to believe that original equipment manufacturers (OEMs/ODMs) are likely to absorb some of the DRAM costs themselves, thus reducing profit margins, rather than passing all the costs on to customers. This perhaps best illustrates just how serious the crisis truly is.

On November 11, semiconductor giant AMD announced the completion of its acquisition of MK1, an American AI inference startup. This deal marks AMD's further expansion in the AI field, aiming to enhance its capabilities in high-speed inference and enterprise-grade AI software stacks.     According to reports, MK1 is led by Paul Merolla, co-founder of Neuralink, whose team possesses deep technical expertise in AI inference technology. Merolla previously assisted in leading chip design efforts at Neuralink and developed algorithms capable of decoding brain activity. Other members of MK1 also come from renowned companies such as Neuralink, Meta, Tesla, and Apple. MK1's Flywheel technology is optimized for AMD hardware and currently processes over 1 trillion tokens daily.   Merolla stated that joining AMD is the right next step for MK1's technology and mission, as AMD's resources and platform will enable MK1 to achieve larger-scale deployment and development. MK1's technology is closely integrated with the memory architecture of AMD Instinct GPUs, delivering precise, cost-effective, and fully traceable inference capabilities.   This acquisition is one of the key initiatives for AMD to advance its broader AI strategy. Recently, AMD has been active in the AI field, continuously enhancing its software capabilities through multiple acquisitions. In August of last year, AMD spent $4.9 billion to acquire server manufacturer ZT Systems, accelerating the development of rack-level systems based on Instinct GPUs. In October of this year, AMD sold ZT Systems' manufacturing division to Sanmina for $3 billion, while retaining its design and support teams.   In addition, AMD has also invested $36 million in several other acquisitions this year, including the purchase of silicon photonics chip startup Enosemi, the employee team of Canadian AI inference chip startup Untether AI, and compiler startup Brium.   According to the latest financial report, AMD's revenue in the third quarter reached $9.25 billion, up 36% year-on-year, with a net profit of $1.2 billion, up 61% year-on-year. The company forecasts a mid-point revenue of $9.6 billion for the fourth quarter, exceeding market expectations. AMD also mentioned its collaboration with OpenAI for 600 gigawatts of computing power, with the deployment of the first 100 gigawatts of AMD Instinct MI450 GPUs set to begin in the second half of 2026. CEO Lisa Su revealed that AMD expects sufficient supply capacity in 2027 and 2028, with AI business revenue projected to reach tens of billions of dollars in 2027.   So far this year, AMD's stock price has doubled, reaching $243.98 per share, with a total market capitalization of $397.2 billion. AMD's sustained investment and strategic positioning in the AI sector are delivering significant market returns and growth potential.

Since its introduction to the memory market in the late 1980s, NAND flash memory has fundamentally changed the way large amounts of data are stored and retrieved. This non-volatile memory designed specifically for high-density data storage is applied in almost every field of the electronic market, from smartphones to data centers, covering everything. It is used in most removable and portable storage devices, such as SD cards and USB drives. In recent years, 3D NAND has also played an important role in the booming development of artificial intelligence, providing an efficient storage solution for the large amount of data required for training AI models. With the explosive growth of data storage demand, chip companies are competing to increase the storage cell density of NAND flash memory (in gigabits per square millimeter (Gb/mm ²)) while reducing the cost per bit. More than a decade ago, the semiconductor industry transitioned from 2D NAND to 3D NAND to overcome the limitations of traditional memory size reduction. In recent years, companies have increased storage density by increasing the number of storage cell layers per chip and the number of storage bits per cell (commercial NAND flash memory can reach up to four bits). One of the most important advances is the transition from floating gate transistors to charge trap units. Floating gate technology stores charges in conductors, while charge trap units store charges in insulators. This reduces the electrostatic coupling between storage units, thereby improving read and write performance. In addition, due to the smaller manufacturing size of charge trap units compared to floating gate transistors, it also paves the way for higher storage densities. But as 3D NAND technology continues to break through physical limits, the semiconductor industry is turning to various new technologies to arrange storage units more tightly - not only horizontally, but also vertically. Several innovative technologies developed by IMEC have achieved vertical expansion without sacrificing the performance and reliability of the memory: air gap integration and charge trap layer separation. Inside the Charge Trap Unit: The Basic Building Blocks of 3D NAND The semiconductor industry plans to apply full ring gate (GAA) or nanosheet transistors to logic chips in the coming years. But the GAA architecture has been widely applied in the field of 3D NAND flash memory and is the main force in high-density data storage. In this 3D architecture, storage units are stacked in vertical chains and addressed through horizontal word lines. In most cases, charge trap cells act as storage devices in 3D NAND. This storage unit is similar to a MOSFET, but it embeds a thin layer of silicon nitride (SiN) within the gate oxide layer of the transistor. This turns the gate oxide layer into a semiconductor material layer called an oxide nitride oxide (ONO) stack, where each layer serves as a barrier oxide layer, a trap nitride layer, and a tunnel oxide layer (Figure 1). The figure shows a 3D NAND GAA architecture with a series of vertical charge trap cells, which have oxide nitride oxide (ONO) gate dielectrics and a limited number of word lines (WL). When a positive bias voltage is applied to the gate, electrons in the channel region tunnel through the silicon oxide layer and are captured in the silicon nitride layer. This will increase the threshold voltage of the transistor. The state of a storage cell can be measured by applying a voltage between the source and drain electrodes. If current flows, it indicates that no electrons are captured and the storage unit is in the "1" state. If no current is measured, the storage unit is in a so-called "electron captured" state, corresponding to "0". The charge trap unit is implemented in a 3D NAND structure using the GAA vertical channel method. Imagine rotating a planar transistor 90 degrees, with the vertical conductive channel surrounded by a gate stack structure. The manufacturing process of GAA channel first involves alternately stacking conductors (silicon, used as word lines) and insulation layers (silicon oxide, used to separate word lines). Next, use advanced dry etching tools to drill down and form cylindrical holes. Finally, alternate deposition of silicon oxide and silicon nitride layers on the sidewalls of the holes, with the channel of the polycrystalline silicon transistor located at the center of all layers. This structure is commonly referred to as the 'macaroni channel'. Next Generation 3D NAND: Cell Stacking and Cell Scaling In the coming years, the memory industry will push the GAA based 3D NAND flash roadmap to its ultimate limit. Nowadays, mainstream manufacturers are launching 3D NAND flash memory chips composed of over 300 layers of oxide/word line stacks (Figure 2). It is expected that by 2030, this number will further increase to 1000 layers, equivalent to approximately 100 Gbit/mm ² of storage capacity. The challenge is how to maintain a consistent word line diameter in a 30 micron thick stacked layer. However, maintaining uniformity of all components in such a small space will continuously increase the complexity and cost of the process, placing higher demands on high stack deposition and high aspect ratio etching processes. This 3D NAND flash image highlights the z-spacing between adjacent word lines. In order to accommodate stacking more layers, semiconductor companies are investing in the development of various supporting tools to improve the storage density of 3D NAND. These 'expansion accelerators' include increasing the number of bits per unit and reducing the xy spacing of GAA units (lateral expansion). In addition to improving bit density and cell density, companies are also taking measures to increase the area efficiency of storage arrays. Another method to increase storage capacity is stacking technology, which involves stacking flash memory devices on top of each other to increase the total number of layers. In 3D NAND flash memory, storage cells are connected in series to form a chain, which is achieved by alternately stacking insulation layers and conductor layers and drilling holes on them. The unit stacking process can be repeated two to three times - possibly even four times in the future - to create longer chains on each chip. Each unit stack is sometimes referred to as a 'layer'. By stacking a large number of storage units and stacking each layer to create higher 3D NAND chips, enterprises can increase the total number of layers without having to manufacture all layers at once. For example, a company can assemble 250 layers of storage units and stack four of them into a 3D NAND chip with 1000 layers. The main challenge is how to etch deep enough holes on these multi-layer storage chips and evenly fill these holes. In addition, some companies are separating the underlying logic from NAND arrays and re integrating it onto NAND arrays in a configuration called CMOS bonded array (CbA). In this configuration, CMOS chips are manufactured on separate silicon wafers and then connected to NAND arrays using advanced packaging techniques, particularly hybrid bonding technology. CbA is the next stage of development for CMOS Down Array (CuA), in which NAND chips are directly manufactured on top of CMOS chips in the same single-chip process. Looking ahead, companies are considering bonding multiple storage arrays onto a single CMOS wafer as an alternative to layered stacking - even bonding multiple array wafers onto multiple CMOS wafers. In order to control the continuously rising manufacturing costs, IMEC and other semiconductor companies are actively exploring vertical or "z-spacing" scaling technologies to reduce the thickness of oxide layers and word line layers. In this way, more storage layers can be stacked at a controllable cost. Advantages and disadvantages of Z-spacing scaling in 3D NAND flash memory Reducing the spacing between storage layers is crucial for continuously lowering the cost of next-generation 3D NAND. The spacing between adjacent word lines is about 40 nanometers, and the purpose of scaling the z-axis spacing is to further reduce the thickness of the word line layer and the silicon oxide layer in the stacked structure. In this way, for every micrometer increase in stacking height, the number of storage layers can be increased, thereby increasing the number of storage units and ultimately reducing costs. However, without optimization, scaling the z-axis spacing will have a negative impact on the electrical performance of the storage unit. This may lead to a decrease in threshold voltage, an increase in subthreshold swing, and a decrease in data retention capability. In addition, it will also increase the voltage required to program and erase the data stored in the storage unit, which will inevitably increase power consumption, reduce the speed of the storage unit (RC delay), and may lead to breakdown of the gate dielectric between adjacent units. These effects can be traced back to two physical phenomena that become more pronounced when memory cells are squeezed closer together: intercellular interference and lateral charge transfer. When the thickness of the word line layer decreases, the gate length of the charge trap transistor also shortens accordingly. As a result, the control ability of the gate over the channel gradually weakens, thereby promoting electrostatic coupling between different cells. In addition to mutual interference between cells, the reduction of storage cells in the vertical direction can also lead to lateral charge transfer (or vertical charge loss): the charges captured inside the storage cells often migrate out of the vertical SiN layer, thereby affecting data retention. The charge trap unit has two geometric directions: z and xy (due to the cylindrical symmetry of the unit, x and y have the same size). Charge can leak from the storage unit in these two directions. The charge will pass through the tunnel in the gate along the xy direction and/or block the oxide from escaping the unit, while also escaping along the z direction, ultimately entering the interior of adjacent units or being too close to them. This is due to lateral charge transfer, which becomes more significant as the vertical size of the cells decreases and the distance between them decreases. Next, we will discuss the technological driving factors that can address these drawbacks, enabling researchers to unlock z-spacing scaling for future generations of 3D NAND flash memory. Between word lines: using air gaps to reduce cell interference Integrating air gaps between adjacent word lines is a potential solution to address inter cell interference issues. The dielectric constant of these air gaps is lower than that of the gate to gate dielectric, thereby reducing the electrostatic coupling between storage cells. This technology has been widely applied in planar two-dimensional NAND flash memory architectures. However, integrating air gaps into high silicon oxide/word line stack structures is more challenging. To overcome these complexities, IMEC proposed a unique integration scheme at the IEEE International Memory Workshop (IMW) in 2025, which enables precise control of the air gap position between word lines. In 3D NAND memory, a thin layer of silicon oxide is placed inside the gate of the storage unit - as a "gate dielectric", separating the word line from the transistor channel - and between the word lines of different storage units - as a "gate to gate dielectric", separating adjacent units from each other (Figure 3). The gate dielectric constitutes the tunnel layer and barrier layer of the ONO stack structure, and surrounds the charge trap SiN layer. 3. The 3D integrated process flow of the air gap (ad) shown in the figure, as well as the transmission electron microscopy (TEM) and energy dispersive X-ray spectroscopy (EDS) images of the air gap (ef). Therefore, silicon oxide not only exists inside each storage cell, but also between cells. Due to the manufacturing process of 3D NAND storage cells, the gate dielectric extends continuously from one cell to another and intersects with the inter gate dielectric in the space between adjacent storage cells. IMEC believes that this is the ideal location for placing the air gap. However, with current process technology, removing (or cutting) the charge trap SiN layer between cells remains a huge challenge. At IMEC, we have found a new method to integrate air gaps without cutting SiN from the storage unit. This innovation introduces an air gap from within the storage hole region by concaving the intergate silicon oxide before depositing the ONO stack layer. The air gap and word line self align to achieve very precise placement. This method also has potential scalability, which is the main issue with other proposed solutions. The results indicate that devices with air gaps are less sensitive to interference from adjacent cells than devices without air gaps. This conclusion is drawn by applying a so-called "on voltage" on the unselected gate, which results in a smaller threshold voltage shift for bandgap devices (Figure 4). This result was obtained on a test device with limited word line layers, a spacing of 30 nm (gate length of 15 nm, thickness of the silicon oxide dielectric layer between gates of 15 nm), and a storage hole diameter of 80 nm. 4. Threshold voltage changes of charge trap devices with and without air gaps (left) at different passing voltages. IMEC researchers also investigated the impact of air gaps on memory performance and reliability. The results indicate that the air gap does not affect the operation of the memory, and its durability can reach 1000 programming/erasing cycles, which is comparable to devices without air gaps. Based on these results, hole side air gap integration is considered a key step in achieving future z-axis spacing scaling. Charge trap cutting: its position in the future development of flash memory IMEC has proven that introducing an air gap in the gate dielectric layer is feasible. However, currently these cavities in storage units only exist before blocking the oxide layer. What if we could drill deeper into the storage unit and introduce air gaps into the regions of the barrier oxide layer and charge trap layer? We tested this method in simulation and the results showed that this charge trap layer separation (or charge trap cutting) can increase the storage window of the storage unit (Figure 5). In addition, charge trap cutting can prevent the captured charges in the storage unit from laterally migrating along the SiN line from top to bottom along the height direction of the oxide layer/word line stack. 5. The difference between a continuous gate stack (left) and a gate stack with charge trap layer cutting and air gap integration (right). The data is stored in flash memory units by programming the threshold voltage to different levels. To store one bit of data, a cell requires two levels: for example, 0V and 1V. To store two bits of data, a cell requires four levels: for example, 0V, 0.5V, 1V, and 1.5V. As the number of bits increases, the number of required voltage levels also increases. It is necessary to increase the total range of threshold voltage (storage window) or reduce the interval between adjacent levels (using a 1-bit interval of 1 V and a 2-bit interval of 0.5 V). However, when these voltage levels are too close, distinguishing them becomes even more difficult. By increasing the storage window, charge trap reduction technology can help each storage unit achieve more levels, thereby storing more bits. However, integrating charge trap cutting in 3D NAND flash memory is not an easy task, as it requires directional etching and deposition of extremely deep and narrow hole walls. For this structure, the technical toolbox used for 2D NAND flash memory is no longer applicable. Currently, IMEC is collaborating with its suppliers to develop new technologies to achieve controllable charge trap cutting. Once the charge trap layer can be interrupted, IMEC plans to combine it with an air gap integration scheme to provide a complete and scalable solution for the z-spacing scaling challenge.  

Last week, Nvidia made history by becoming the first company to surpass a market value of $5 trillion. But this is just one of the many ways it affects the global economy. This chip manufacturer leading the artificial intelligence revolution is not only the world's most valuable company to date, but may also be the most influential stock in Wall Street history. Since the beginning of 2023, Nvidia has been the main driving force behind market growth, bringing huge returns to shareholders and earning billions of dollars for CEO Huang Renxun. Nowadays, its market value has exceeded 6 out of 11 sectors in the S&P 500 index, and even surpassed the market value of most countries' entire stock markets. From a historical perspective, this is clearly a huge anomaly and a shocking move, "said Matt Miskin, Co Chief Investment Strategist at Manulife John Hancock Investments. Just last week, Nvidia announced cooperation agreements with Nokia, Samsung Electronics, and Hyundai Motor Group. Although the company will not release its financial report until mid November, the recent performance of large technology companies highlights their enormous growth potential. Microsoft, Amazon, and Meta Platforms have all pledged to continue investing heavily in artificial intelligence. According to data compiled by Bloomberg, it is expected that the total capital expenditures of these four companies will increase by 34% in the next 12 months, reaching approximately $440 billion. These expenses are the main reason why Nvidia's projected revenue for the next fiscal year is expected to reach $285 billion, compared to only $11 billion in revenue for the 2020 fiscal year. All these help to explain why the artificial intelligence based stock market foam has become so popular, and Nvidia is the center of this foam. Last week, Huang Renxun played down people's concerns about the runaway market frenzy at the company's annual GTC conference, and Jerome Powell, the chairman of the Federal Reserve, also refuted the idea of comparing the current situation with the Internet foam at the end of the 1990s at a press conference on Wednesday. This trend will reach its peak and reverse, and we expect this situation to eventually happen, "Miskin said. However, currently, companies at the center of the artificial intelligence race are performing the best in terms of profitability, and this situation needs to change in order to achieve leadership succession. Nevertheless, the S&P 500 index seems to have put too many eggs in one basket The following five charts document the process of Nvidia's market value soaring to $5 trillion and demonstrate its importance to the stock market:   As the world's largest company by market capitalization, Nvidia naturally has the highest weight in major stock indices (calculated by market capitalization). Its stock accounts for 8.5% of the S&P 500 index, exceeding the total weight of the 240 companies with the lowest market capitalization. Howard Silverblatt, a senior index analyst at Standard&Poor's, said that this is likely the highest weighted record among any constituent stock, but he also pointed out that it is difficult to find daily data for a century. In mid-2023, Apple's weight reached a peak of 7.7%, while Microsoft also reached 7.4% later that year. At present, the total weight of the seven major technology stocks in the S&P 500 index exceeds 36%, with Apple Inc. ranking second with a weight of 6.9%.   Nvidia is not only the world's most valuable company, with a market value about $1 trillion higher than the second ranked Apple, but according to data compiled by Bloomberg, Nvidia's market value even exceeds the total market value of five countries' stock markets: the Netherlands, Spain, the United Arab Emirates, and Italy. This Santa Clara, California based company's market value now exceeds that of all stock markets except for the United States, China, Japan, Hong Kong, and India.   Almost all Wall Street analysts are bullish on this stock, with about 91% of analysts giving it a "buy" or "buy" rating. The market generally believes that this upward trend will continue, and HSBC analyst Frank Lee recently raised the target price of the stock to the highest on Wall Street at $230- meaning its market value will approach $8 trillion. However, there is also an analyst who holds the opposite opinion on the stock: Jay Goldberg, an analyst at Seaport Global Securities, has maintained a "sell" rating since April and has set a Wall Street low target price of $100. During this period, the stock price has more than doubled.   As the company expands, its sales growth rate often slows down due to a larger base. The average annual revenue growth rate for S&P 500 index companies with expected sales of $100 billion or more is 6%. As a result, Nvidia has become an exception, with its revenue expected to increase by nearly 60% this fiscal year. Although this growth rate has slowed down from 126% and 114% in the previous two years, it still far exceeds other giant companies of Nvidia. The expected annual revenue growth rates for the second and third ranked companies - Microsoft and Apple - are 15% and 6.2%, respectively.   With the soaring stock price of Nvidia, Huang Renxun's net worth has also skyrocketed. According to the Bloomberg Billionaires Index, his net worth has reached $176 billion. This year alone, his wealth has increased by over $60 billion, enough to make him one of the top ten billionaires in the world. According to documents submitted to the US Securities and Exchange Commission in October, Huang Renxun holds approximately 3.5% of the company's shares in his personal name and family trust.

Just now, Skyworks, a leading global high-performance analog and mixed signal semiconductor company, and Qorvo, a leading global provider of connectivity and power solutions, announced that they have reached a final agreement to merge the two companies in cash and stock transactions, with a combined valuation of approximately $22 billion, to create a globally leading high-performance RF, analog, and mixed signal semiconductor company headquartered in the United States. Skyworks CEO and President Phil Brace said, "This merger is an important milestone for both our industry and Skyworks. The combination of Skyworks and Qorvo's complementary product portfolio, along with a world-class engineering team, will enhance our ability to meet the growing customer demands in the mobile and diversified markets. With stronger scale, a more diverse customer base, and operational synergies, we can bring superior innovation to our customers and create sustainable value for shareholders. ” Qorvo CEO and President Bob Bruggeworth said, "Qorvo and Skyworks share a common culture of innovation and are committed to solving our customers' most complex challenges. By partnering with Skyworks, we can accelerate innovation and provide broader and more comprehensive solutions in numerous growth areas. We are pleased to leverage the combined strengths of our teams, products, and technology combinations to strengthen our capabilities in the mobile field and significantly expand our influence in industries such as defense and aerospace, edge IoT, artificial intelligence data centers, automotive, and others driven by long-term growth trends. ” Strategic basis and transaction highlights It is expected that this transaction will bring significant long-term value to customers, employees, and shareholders. Enhanced scale and financial condition: The combined company is expected to have a total revenue of approximately $7.7 billion and adjusted EBITDA of $2.1 billion. The combined company will be more capable of competing with larger companies - thanks to a stronger and more balanced revenue base, resulting in more predictable performance, more efficient cost structure, and flexible cash generation over the cycle. Stronger innovation capability: This merger will create an innovative global RF, analog, and power technology company, providing customers with more integrated complete solutions and a wide range of products and technologies. The merged company will bring together world-class engineering talent, including approximately 8000 engineers and technical experts, as well as over 12000 authorized and pending patents, enabling accelerated development of advanced system level solutions and unlocking new Design Win opportunities to meet growing customer demands. Create $5.1 billion mobile business: This merger will integrate complementary RF technologies and top-notch products, expand opportunities for mobile business, and enhance revenue stability. A broader product portfolio will enhance our cross platform competitiveness, deepen customer integration, and enrich our technological foundation, while consolidating our advantages in dealing with increasingly complex RF businesses. Establishing a diversified broad market platform worth $2.6 billion: This transaction will create a broad market platform worth $2.6 billion, with a continuously growing potential market size (TAM) and strong profitability, covering the defense and aerospace, edge IoT, artificial intelligence data centers, and automotive markets. The characteristics of these markets are good long-term growth trends, long product lifecycles, and good gross profit margins. Enhancing the domestic manufacturing status and utilization rate in the United States: The merged company will strengthen its domestic production capacity and improve its capital efficiency, and provide support through a strong supply chain partner network to meet the needs of large quantities and highly specialized customers. Immediate and Significant Value Added: It is expected that this transaction will immediately and significantly increase non GAAP earnings per share upon completion of the transaction, and generate $500 million or more in annual cost synergies within 24-36 months after the full integration of the two companies. Transaction Details According to the terms of the agreement, at the end of the transaction, Qorvo shareholders will receive $32.50 in cash and 0.960 shares of Skyworks common stock for each share of Qorvo stock they hold, which means the combined enterprise value is approximately $22 billion. After the transaction is completed, Skyworks shareholders will hold approximately 63% of the merged company's shares, while Qorvo shareholders will hold approximately 37% of the merged company's shares (calculated on a fully diluted basis). Phil Brace will serve as the CEO of the merged company; Bob Bruggeworth will join the board of directors of the merged company. The board of directors of the merged company will consist of 11 directors, including 8 from Skyworks and 3 from Qorvo. Skyworks plans to pay for the cash portion of the transaction through existing cash and additional financing. Skyworks has received a debt financing commitment from Goldman Sachs Bank of America. This transaction does not come with any financing conditions. It is expected that the net leverage ratio of the merged company at the end of the transaction will be approximately 1.0 times its adjusted earnings before interest, tax, depreciation, and amortization (EBITDA) for the past 12 months. This favorable capital structure will help the company continue to invest in its business, thereby enhancing shareholder value. Time and Approval The boards of directors of both companies have unanimously approved the transaction, which is expected to be completed in early 2027, subject to obtaining necessary regulatory approvals, approvals from Skyworks shareholders and Qorvo shareholders, and meeting other customary closing conditions. Starboard Value LP, which holds approximately 8% of Qorvo's shares, has signed a voting agreement to support the transaction.

By the early 1970s, engineers already had oscillators, timers, and monostable circuits, though they weren't integrated into a single integrated circuit. Analog circuits using discrete components were cumbersome, and existing timer chips lacked flexibility. This changed in 1971, when Hans Camenzind, a contractor at Signetics, designed a general-purpose timer chip using only about 25 transistors, a few diodes, and a clever resistor divider network. The result was the NE555: an eight-pin analog timer that could generate monostables, square waves, and triggers, depending on how it was connected. It operated from a single supply, had a robust output stage, and could withstand noisy environments. Signetics's introduction quickly became a hit among hobbyists and industry.   In the decades that followed, the 555 chip found widespread use in LED flashlights, motor drivers, servo testers, voltage-controlled oscillators (VCOs), debounce circuits, and more. Engineers used it to control relays, play melodies, detect missing pulses, and ensure the proper functioning of watchdog timers. Tens of billions of units have been shipped, and today's CMOS versions continue this tradition, retaining the mindset that millions of engineers grew up with while significantly reducing power consumption.   Depending on how the pins are connected, the 555 can operate in one of three core modes: Monostable (one-shot): A negative pulse at the trigger pin pulls the input below 1/3 VCC, setting the latch and driving the output high. The capacitor then charges through the resistor, and when the voltage reaches 2/3 VCC, the latch resets and the output falls. The pulse width is t ≈ 1.1 RC. Astable (free-running): Placing a capacitor between ground and pin 6 (THRES), with the charge/discharge time controlled by two resistors (RA and RB), causes the output to oscillate continuously. The frequency is approximately f ≈ 1.44 / ((RA + 2RB) x C). Bistable (Flip-Flop): TRIG and THRES can be used as set/reset inputs, enabling the 555 to be used as a basic storage element or a debounced switch. The IC also includes a control voltage pin for external modulation or PWM control, providing a 2/3 VCC threshold. If unused, it is typically bypassed to ground with a 10 nF capacitor to prevent instability. One of the 555's most useful design features is its output stage. The bipolar output current is approximately 200 mA, sufficient to directly drive small loads, making it a popular choice for low-component-count circuits long before microcontrollers became inexpensive and commonplace. Despite its analog simplicity, the 555 continues to appear in new designs where its inclusion still makes sense. While a microcontroller might seem overkill, the 555 chip is a fast, deterministic, and robust solution for implementing pulse, delay, or blinking behavior with just a few passive components. With the introduction of CMOS versions like the TLC555 and LMC555, the chip's appeal expanded to low-power, battery-powered, and rail-to-rail swinging applications.   Of course, there were trade-offs. The bipolar NE555, due to its large output transients, could introduce noise spikes on VCC. It also had asymmetrical output drive strengths and a minimum duty cycle exceeding 50% in its basic astable mode, though a shunt diode across RB or other topologies could address this.   Nevertheless, for engineers who grew up controlling blinking LEDs with 555 chips on breadboards, this chip was the epitome of analog circuit elegance: no firmware, no initialization, just a few volts and a capacitor. And it still ships in volumes that most digital ICs can only dream of.

For a long time, Chinese contract manufacturers, such as Foxconn, Luxshare Precision, and Goertek, have been dubbed "Apple workers." This term, frankly, isn't a pleasant one, as it often implies that Chinese companies are forced to rely on cheap labor to perform low-paying tasks in the consumer electronics and tech product value chains, arguably the world's most lucrative. But things are changing now. In September of this year, a bombshell announcement rocked the tech world: OpenAI, Silicon Valley's most prominent AI giant, had approached China's Luxshare Precision to collaborate on AI hardware. This wasn't a simple contract manufacturing relationship. According to The Information, OpenAI and Luxshare Precision have reached a strategic partnership agreement to jointly develop a consumer AI device, with mass production expected as early as late 2026 or early 2027. This time, Luxshare Precision is no longer simply a "contract manufacturer," but will work with OpenAI to "jointly define products and participate in software and hardware collaboration." This news is truly significant: it signifies that a large number of Chinese companies, represented by Luxshare Precision, have achieved substantial progress in industrial upgrading. Chinese companies with mastered manufacturing expertise have already gained considerable influence. Luxshare Precision's move is no accident. Since the tariff war began in April of this year, Apple has gradually shifted production to India and Vietnam. Coupled with Apple's increasingly sluggish innovation in recent years, investors have become increasingly aware that the growth of Apple supply chain companies is simply a function of Apple's new product sales multiplied by market share, lacking a rationale for independent growth. If Apple's innovation slows (as seen with the iPhone's slowdown for several generations), the valuations of Apple supply chain companies will immediately come under pressure. This situation raises the question: Can domestic contract manufacturers survive without Apple? Or are there more promising options? The beginning of transformation: Luxshare Precision's collaboration with OpenAI is just a microcosm of the broader transformation of Apple supply chain companies. In recent years, every Apple supply chain company with a modicum of strength has been desperately seeking new paths forward. Luxshare Precision's transformation over the past few years can be considered one of the most successful examples. Overall, Luxshare Precision's revenue growth is fastest from its new energy vehicle business. Luxshare Precision primarily manufactures wiring harnesses, connectors, charging components, and smart cockpit electronic modules for new energy vehicles—critical components that safely and efficiently connect electricity, signals, and data within the vehicle. Currently, new energy vehicle revenue accounts for 39.47% of Luxshare Precision's revenue, reaching 4.998 billion yuan. The company boasts a formidable automotive client list, including Tesla, CATL, BMW, Mercedes-Benz, and Volkswagen. Goertek has taken a different approach, placing its bets on the VR/AR market. In 2020, Goertek signed an exclusive contract with Meta's Oculus for its next-generation products, securing an order. In 2022, Goertek's smart hardware business, including AR/VR products, accounted for 60% of its revenue, surpassing its acoustic device business for two consecutive years, with revenue increasing by over 90% year-on-year. The most thorough and successful transformation has undoubtedly been achieved by Foxconn Industrial Internet. This company, which was born out of Foxconn, has now transformed itself into the absolute leader in AI server manufacturing. In the first half of 2024, Foxconn Industrial Internet's net profit reached 12.113 billion yuan, a year-on-year increase of 38.6%. More importantly, through cloud computing business, Foxconn Industrial Internet's revenue exceeded the communications and mobile network equipment business for the first time, becoming a new revenue pillar. On the technical level, Foxconn Industrial Internet is also quite capable. The superfluid liquid cooling solution jointly developed by the company and NVIDIA can meet the heat dissipation requirements of the 1200W power consumption of the Blackwell chip. From product design, key components to system delivery, Foxconn Industrial Internet has established a complete AI server industry chain capability. Xinwangda, which used to make batteries for Apple, has also vigorously transformed into the new energy track in recent years and has become a battery supplier for car companies such as Weilai, Xiaopeng, and Dongfeng. At present, these three companies account for 72% of its installed capacity in 2023. More crucially, Wei, Xiaoli, and Li Auto directly invested in Xinwangda, establishing a deep "customer + shareholder" relationship. Xinwangda achieved this not through storytelling, but through proven performance—244Wh/kg energy density, 50,000 cycle life, and reliable operation at -30°C—to carve out a niche for itself amidst the competition from CATL and BYD. Looking at the transformation cases of these Apple supply chain companies, it's easy to spot a common thread: they've all moved beyond simple contract manufacturers and have begun penetrating into new sectors like new energy vehicles and AI servers. The logic behind this is simple. While manufacturing is an industry that emphasizes standards, leading companies in each niche market possess their own unique "processes." When these processes reach peak performance, they transcend mere "cheap labor" and expand across a wide range of related industries, becoming essential "solutions" for the industry. The "real gold" forged in contract manufacturing: When it comes to the unique "processes" honed through contract manufacturing, Foxconn Industrial Internet is a prime example. In the AI ​​era, Foxconn Industrial Internet (FII) provides far more than just hardware assembly. It offers a one-stop solution encompassing design, R&D, manufacturing, and even advanced thermal technologies like liquid cooling. As early as 2015, FII partnered with Alibaba to develop immersion liquid cooling (also known as "immersion cooling") products. Their fourth-generation AI servers currently utilize both water and air cooling technologies. The immersion cabinets achieve a Power Use Effectiveness (PUE) of 1.03, with values ​​closer to 1 indicating higher efficiency. Traditional air-cooled data centers typically achieve PUEs above 1.5. By 2023, FII's AI server shipments will account for nearly 40% of global shipments, placing it firmly in the top tier internationally for liquid-cooled models. This integrated capability, from design, manufacturing, to delivery, is unmatched by many pure liquid cooling technology companies, such as GRC in the US and Submer in the UK. At this point, some may wonder: How did these Apple supply chain companies make the leap from making mobile phone parts to manufacturing cars and AI servers? Can technology be easily swapped? The answer is not necessarily true. There's a very interesting logic behind this "technology transfer." Let's first discuss the core: precision manufacturing technology. Take Luxshare Precision, for example. They originally made connectors—those small parts that plug into mobile phones. Don't underestimate these things; they involve a lot of skills: micron-level precision, guaranteed to last tens of thousands of plug-in and unplug cycles, high and low temperature resistance, and electromagnetic interference resistance... These skills all embody the essence of "precision." But think about it another way: don't the charging ports and battery connectors in new energy vehicles face the same demands? In fact, the requirements are even higher—car connectors are expected to last for over a decade, while mobile phones only last three to five years at most. So, Luxshare Precision's entry into the automotive market is essentially upgrading its "mobile phone-grade" precision manufacturing technology to "automotive-grade" levels. The technical path is the same, only the standards are more stringent. The example of Foxconn Industrial Internet is even more interesting. As we all know, iPhones are getting thinner and thinner, but their processors are getting more powerful. This poses a major problem: heat dissipation. How can heat be efficiently dissipated in such a small space? To address this issue, Foxconn Industrial Internet (FII) has invested heavily in researching various heat dissipation technologies: thermal conductive materials, heat dissipation structure design, and heat pipe technology. In the AI ​​era, AI servers have far greater cooling requirements than iPhones. Nvidia's H100 chip consumes 700W of power, and the new Blackwell chip reaches 1200W. This is equivalent to the heat generated by hundreds of iPhones concentrated in a single tiny chip. However, the technical principles are the same: the goal is to quickly transfer heat away from the source. FII's accumulated experience in thermal conductive material formulations, heat dissipation structure design, and even its understanding of air flow can all be utilized. The only difference is the upgrade from milliwatts to kilowatts, and from air cooling to liquid cooling. Furthermore, Apple's notoriously stringent product quality standards—"zero defects" and "full traceability"—have long been fundamental to Apple supply chain companies. The quality management system developed as a result has become their key strength. Therefore, the core reason for these Apple supply chain companies' successful transformation lies not in their mastery of specific product manufacturing technologies, but rather a universal "precision manufacturing methodology." This is the true value earned day after day on the production line. While Apple supply chain companies have partially transitioned from manufacturing to intelligent manufacturing, leveraging the skills and knowledge accumulated through their previous contract manufacturing experience, simply "breaking away from Apple" is only the first step to survival. The real path forward lies in finding a new role. If these companies simply focus on switching major clients (e.g., from Apple to Tesla), they're still stuck in their old ways and will never be able to achieve independence. Furthermore, given the current Sino-US tensions, this structure, where all major clients are from mainland China, is inherently fraught with uncertainty. Therefore, the true potential for these former Apple supply chain companies lies in transitioning from "OEM parts" to "system definition." In the Apple era, Apple supply chain companies were "hanging on a single phone." But in today's increasingly important era of AI, manufacturing needs are fragmented: there are startups developing AI glasses, laboratories developing humanoid robots, and local governments building edge data centers... They all require reliable, flexible manufacturing partners capable of small batches and rapid iteration. If Apple supply chain companies can package the entire suite of capabilities they used to serve Apple—supply chain, quality control, and automation—into "Manufacturing as a Service" (MaaS), they could become the "water, electricity, and gas" for AI hardware innovation. Just as TSMC doesn't make phones, but its chip manufacturing underpins the entire semiconductor ecosystem, future Apple supply chain leaders may not make cars or robots, but all AI hardware will be inseparable from it. This "Manufacturing as a Service" model bears a strong resemblance to the traditional SaaS model. SaaS, standing for "Software as a Service," literally translates to "software as a service." Think of the apps on your phone. Take Didi Chuxing: In the past, hailing a ride meant either manually hailing one on the street or calling a taxi company. Now, just open your phone and tap it. In the past, if you wanted takeout, you had to call the restaurant for delivery. Now, if you want Sichuan cuisine or hot pot, you can just open Meituan and order it with a few taps. "Manufacturing as a Service" (MaaS) simply means transforming complex "factory management" into a simple, easy-to-use program like a mobile app. This transforms Apple supply chain companies from "OEMs" into "intelligent manufacturing service providers." They're no longer simply selling physical labor; they're creating an "industrial Meituan Takeout"—allowing any business to access world-class manufacturing services anytime, anywhere, just like ordering takeout. Moreover, this concept is no longer just wishful thinking; it's already being demonstrated in real-world cases. A prime example is Foxconn's "MIH Electric Vehicle Open Platform." MIH (Mobility in Harmony) is Foxconn's open ecosystem for electric vehicles, encompassing "hardware, software, and supply chain," launched in 2020. Its core logic is simple: "You have a brand, algorithms, and users, but can't build a car? No problem. We'll provide you with the chassis, three-electric system, electronic architecture, supply chain, and even help you find a contract manufacturer—all you have to do is define the product." This essentially packages Foxconn's decades of experience serving Apple, Dell, and Cisco, encompassing vehicle-level integration capabilities, global supply chains, automated production lines, and quality control systems, into a single "electric vehicle manufacturing operating system." Currently, over 1,900 companies have joined the MIH Alliance, including Qualcomm, Nvidia, Arm, and CATL. Even Fisker in the United States and Thailand's national electric vehicle program are using this platform to develop models. In the AI ​​era, if these Apple supply chain companies leverage the data and knowledge they've amassed over the years into various "industrial apps," a single process solution could simultaneously serve thousands of factories; new process optimizations could be distributed to all customers via "cloud updates"; and every factory's production data could be used to train AI models, optimizing process standards across the industry. This allows contract manufacturers to leap from the bottom of the value chain to the top, transforming from "contract manufacturers" into "intelligent manufacturing platforms." They then hold onto these software-based patents and reverse-license them to brands, collecting platform and patent fees. Some might argue that with the widespread adoption of robots and automated factories, China's demographic dividend and cheap labor advantages will eventually disappear. At that point, AI and automation will gradually close the global manufacturing gap. However, this idea of ​​relying on "automation" to conquer the world clearly underestimates the complexity of manufacturing. The reason is that while manual labor can be replaced by machines, valuable knowledge and experience are the industry's treasures, which even AI cannot replace. Manufacturing is indeed an industry that values ​​standards, but these are "customer-given standards," not "standards of how they are achieved." Apple's requirements for its Chinese contract manufacturers are indeed clear: "Total earphone thickness 5.3mm, ±0.05mm tolerance, IPX4 waterproof rating, Bluetooth latency <80ms..." But the question is, how can you actually manufacture products that meet these requirements? The answer is: it all depends on the contract manufacturers' own "trial and error," assembling and understanding. This is called industry know-how—without decades of deep industry experience, you simply can't master it. This knowledge is ingrained in the minds of veterans, hidden in process documentation, and etched into equipment specifications. And this is precisely the greatest asset that Chinese companies have in becoming "intelligent manufacturing platforms" in the AI ​​era. The laws of nature are clear, and change is inevitable. In this era of great change, only those companies that can turn "crisis" into "opportunity" will be able to buck the trend. Luxshare Precision has gone from being an Apple component supplier to a partner of OpenAI; Foxconn Industrial Internet has gone from being an iPhone OEM to a leading AI server manufacturer; Goertek has gone from being a traditional acoustic component manufacturer to a major player in the VR/AR field... Ten years ago, without Apple, China's OEMs would have been in mourning. But now, even without Apple, China's OEMs are on their own growth trajectory.