Core Viewpoint - Google's recent launch of the TurboQuant compression algorithm, which claims to reduce the key cache usage of large models by at least six times, has caused panic in the storage industry. However, at the MemoryS 2026 summit, various storage manufacturers and cloud computing companies expressed that the demand for storage is still expected to rise due to the acceleration of AI applications, leading to a potential continuation of shortages [3][4]. Group 1: AI Impact on Storage Demand - AI is rapidly consuming storage capacity, with AI servers projected to account for over 20% of total server shipments by 2026, significantly increasing storage configurations [4]. - The shift from AI model training to more frequent actual usage has heightened the demand for data reading speed and response capabilities, making high-performance storage a core component for system efficiency [4]. - The automotive sector is emerging as a significant application area for AI, alongside edge AI, which is expected to drive new growth in storage demand [4]. Group 2: Supply-Side Challenges - The expansion cycle for storage capacity is lengthy, lasting 18 to 24 months, making it difficult to alleviate supply shortages in the short term. Structural mismatches in supply and demand have become the norm [5]. - Flash memory is expected to remain in short supply for an extended period, with feedback from system-side experts indicating increasing resource constraints due to AI growth squeezing DRAM capacity [6]. - The storage technology landscape is undergoing a paradigm shift, moving from micro-innovations to architectural revolutions, with concepts like CXL and integrated storage-compute solutions accelerating towards commercialization [6].

Core Viewpoint - The demand for AI is driving the storage industry, which is currently in the mid-stage of a super boom cycle, with supply shortages expected to last at least until 2027 [1] Group 1: Storage Industry Outlook - The storage industry is maintaining a high level of prosperity, supported by better-than-expected performance and guidance from key players like Kioxia, as well as an increase in NAND contract prices [1] - The industry is expected to remain in a supply-demand imbalance until the end of 2027, with a strong recommendation for storage module companies due to their short-term performance potential [1] Group 2: Investment Opportunities - The report highlights four key investment directions in the context of the AI era, focusing on the need for bandwidth and capacity upgrades [2] - Storage solution providers are essential for CUBE, with a focus on companies that have support from storage manufacturers and first-mover advantages [2] - Semiconductor equipment is benefiting from the upgrade in advanced packaging demands, with a focus on etching, bonding, and thinning equipment [2] - Advanced packaging is seen as a critical breakthrough for high-end storage, with Chinese manufacturers leading in capabilities and expanding capacity [2] - Logic chip companies are expected to enhance their competitiveness and accelerate industrialization, particularly in 3D structured logic chips, benefiting from AI-driven demand [2]

Core Viewpoint - The article highlights the increasing demand for storage chips, particularly HBM and NAND flash, driven by AI needs, leading to a supply shortage and rising prices. Nvidia is strengthening its collaboration with strategic partners like Samsung to develop new AI technologies and ferroelectric NAND flash memory, which could address critical industry challenges [1][3]. Group 1: Market Dynamics - The global NAND supply peaked in 2022 at 21.387 million wafers but is expected to decline to 15.408 million wafers in 2023, with a projected recovery to only 17.61 million wafers by 2028, insufficient to meet surging market demand [3]. - NAND prices have surged, with a 90% increase in the first quarter of this year alone due to the worsening supply shortage [3]. - Nvidia plans to introduce a new type of NAND called "Inference Context Memory Storage" (ICMS) in its next-generation AI accelerator "Vera Rubin," which is expected to account for 9.3% of global NAND demand [3]. Group 2: Technological Innovations - Nvidia is investing in new technologies, including a $4 billion investment in silicon photonics startups Lumentum Holdings and Coherent, and plans to establish the Nvidia Accelerated Quantum Research Center (NVAQC) to maximize efficiency by combining GPUs and quantum processing units [4][5]. - The collaboration with Samsung on ferroelectric technology is part of Nvidia's efforts to acquire emerging technologies [5]. - Ferroelectric materials can maintain polarization without external electric fields, potentially reducing voltage requirements and enabling denser stacking of memory chips, which could significantly enhance supply capabilities [7]. Group 3: Competitive Landscape - Samsung has developed a NAND stacking technology with approximately 200-300 layers and is focusing on ferroelectric materials to achieve 1000-layer stacking, addressing the growing AI storage demand [7]. - In the past 12 years, Samsung has filed the most patents in the ferroelectric device field, totaling 255, which represents 27.8% of the total, surpassing competitors like Intel, SK Hynix, and TSMC [8]. - The competition in the industry is intensifying, with companies like Peking University developing a 1nm ferroelectric transistor, indicating a pressing need to advance NAND technology to support AI chip development [8].

Core Insights - The article highlights the significant achievements of Hefei Reliance Memory (合肥睿科微电子) at the ISSCC 2026, showcasing China's advancements in RRAM technology and its potential in AI applications [2][19] - The company has developed a unique ReRAM-based chip that addresses key challenges in AI inference, particularly in large language models (LLMs) and edge AI applications, demonstrating a successful integration of mature process technology with innovative architecture [3][4][17] Group 1: RRAM Technology and Innovations - Hefei Reliance Memory has leveraged its proprietary ReRAM technology to create a chip that excels in performance, energy efficiency, and cost-effectiveness, suitable for AI inference across various scenarios [3][4] - The ReRAM technology offers advantages such as non-volatility, high-speed read/write capabilities, low power consumption, and high-density integration, making it a promising solution for AI, IoT, and edge computing [4][18] - The company’s approach combines mature 55nm CMOS processes with innovative architecture, allowing it to challenge advanced process technologies effectively [4][17] Group 2: Breakthroughs in AI Inference - The company introduced a 55nm LLM accelerator that utilizes a 3D stacked ReRAM-on-Logic architecture, significantly improving performance and reducing costs for AI model inference [9][10] - Key innovations in the accelerator include a locally rotating unit that enhances inference speed by 3.82-3.93 times while saving 92.7% chip area, and a stacked architecture that eliminates external memory access delays [10][11] - The accelerator achieves a peak performance of 2.33 TOPS with a power consumption of only 49.54mW per ReRAM chip, demonstrating its capability to meet the demands of LLM inference [11][12] Group 3: Edge AI Applications - A fully analog intelligent vision SoC developed in collaboration with local research teams has achieved significant efficiency in edge AI applications, marking a breakthrough in end-to-end processing without the need for A/D conversion [14][16] - This SoC integrates various components, including a PWM image sensor and ReRAM memory, to deliver high-speed, low-power performance suitable for edge devices [15][16] - The chip's energy efficiency reaches 345.54 TOPS/W, with substantial improvements over previous solutions, making it ideal for applications in smart wearables and autonomous driving [16][18] Group 4: Industry Implications - The advancements made by Hefei Reliance Memory not only signify a technological leap for China in the semiconductor industry but also provide a viable path for domestic AI inference solutions, reducing reliance on foreign technologies [17][18] - The combination of ReRAM technology with mature processes aligns with the industry's trend towards cost reduction and efficiency, supporting the broader goal of achieving self-sufficiency in semiconductor manufacturing [18][19] - The success at ISSCC 2026 serves as a model for other domestic companies to follow, emphasizing the importance of innovation and collaboration in advancing the semiconductor landscape [17][19]

Core Viewpoint - The investment by Zhaoyi Innovation in Hangzhou Weina Core is seen as a strategic move to position itself in the integrated storage and computing sector, indicating potential future collaborations in NOR Flash in-memory computing technology [1][3]. Group 1: Investment Details - Zhaoyi Innovation officially acquired a stake in Hangzhou Weina Core on March 3, 2026, increasing its registered capital to 1.5753 million yuan, with a nominal increase of 9,490 yuan [1]. - The small capital increase is interpreted as a compliance measure rather than a significant financial investment, suggesting that further investments may follow [1]. Group 2: Company Background - Zhaoyi Innovation has over ten years of experience in the semiconductor industry, focusing on storage chips and holding a leading global market share in NOR Flash [3]. - The company has developed advanced technologies, including 3D stacked DRAM, which has achieved HBM functionality with a bandwidth of 1.2 TB/s, a 40% reduction in power consumption, and a 50% decrease in costs [3]. Group 3: Weina Core's Technology - Weina Core, established in 2021, is recognized for its innovative 3D-CIM™ technology, which integrates in-memory computing to eliminate data transfer overhead and address bandwidth, energy efficiency, and computing power challenges [3]. - The company has a strong R&D team with a high density of top-tier talent, having published numerous record-breaking chip test results in the past six years [3]. Group 4: Strategic Implications - The collaboration between Zhaoyi Innovation and Weina Core is expected to enhance Zhaoyi's capabilities in AI computing, completing its ecosystem and accelerating the deployment of integrated storage and computing chips [7]. - Zhaoyi Innovation's recent announcement to invest 400 million yuan in an integrated circuit industry fund reflects its commitment to deepening its involvement in the semiconductor sector and promoting domestic alternatives [7]. Group 5: Industry Trends - The IEEE is expected to release standards for integrated storage and computing systems in 2026, which will accelerate industry standardization [9]. - The partnership between Zhaoyi Innovation and Weina Core is anticipated to stimulate domestic supply chain collaboration in the integrated storage and computing sector, aiding local companies in overcoming AI chip patent barriers [9].

Core Viewpoint - The article discusses the advancements in the semiconductor industry, particularly focusing on the introduction of embedded RRAM technology by Shengxianwei, which aims to reduce costs and improve efficiency in AMOLED display driver chips [2][3]. Group 1: RRAM Technology and Its Advantages - The new Ramless+RRAM solution integrates storage units directly into the chip, eliminating the need for external NOR Flash chips, thus significantly reducing packaging costs and PCB space [3]. - The embedded architecture of RRAM enhances efficiency by eliminating delays associated with external reading, resulting in faster parameter retrieval and improved system responsiveness [3]. - RRAM technology operates at lower voltage and power, extending the battery life of end devices, making it a favorable solution amid current supply chain uncertainties [3]. Group 2: RRAM's Technical Superiority - RRAM features a simplified process structure, requiring only two additional masks in 28nm and more advanced logic processes, compared to over ten masks for traditional eFlash, leading to shorter production cycles and lower wafer manufacturing costs [4]. - The energy efficiency of RRAM is significantly higher, with write power consumption being a fraction of that of Flash, making it ideal for applications sensitive to power consumption [4]. - RRAM has inherent scalability, maintaining stable storage performance even at sizes below 10nm, overcoming physical limitations faced by traditional storage technologies [5]. Group 3: Market Position and Future Prospects - Reliance Memory, established in 2018, has rapidly grown in the RRAM sector, holding over 300 patents and providing comprehensive RRAM technology support to various semiconductor design companies [7]. - The demand for compact, low-power, high-performance storage solutions is expected to surge with the rise of AI computing at the edge, positioning Reliance Memory's RRAM technology as a key player in the market [7]. - The successful application of RRAM in display driver chips marks the beginning of its potential across various sectors, including ultra-low power MCUs and future integrated storage-computing chips [7].

Core Insights - A research team from Peking University's School of Electronics has developed the smallest and lowest power-consuming ferroelectric transistor to date, which is expected to support the enhancement of AI chip computing power and energy efficiency [1] - The physical gate length of the ferroelectric transistor has been reduced to the 1-nanometer limit, breaking traditional physical constraints and achieving energy consumption levels an order of magnitude lower than the best international standards [1] - The "storage-computation integration" feature of this transistor allows it to perform data storage and computation simultaneously, aligning well with the evolution direction of AI chips [1] Industry Impact - This breakthrough is anticipated to eliminate the efficiency bottleneck caused by the separation of "storage" and "computation" in traditional computing architectures, laying a critical technological foundation for the development of the next generation of high-performance, low-energy AI chips [1] - The ultra-low operating voltage and energy consumption characteristics of the nanoscale ferroelectric transistor provide a core device solution for constructing high-efficiency data centers, further establishing a technological basis for the next generation of high-performance AI chips [1] - This technological advancement fills a gap in the domestic atomic-scale ferroelectric transistor field and breaks the monopoly of international giants in the core device sector of AI chips [1] Related Companies - Relevant A-share concept stocks include Guolin Technology and Yidao Information, as reported [1]

Core Viewpoint - The research team at Peking University has developed the smallest and lowest power ferroelectric transistor, potentially enhancing AI chip performance and energy efficiency [1][2] Group 1: Technological Advancements - The physical gate length of the ferroelectric transistor has been reduced to the atomic scale of 1 nanometer, allowing for the creation of a high-strength electric field within the ferroelectric layer with minimal external energy (0.6V) [2] - This innovation overcomes traditional limitations of ferroelectric transistors, reducing energy consumption by an order of magnitude compared to the best international standards [2] Group 2: Implications for AI and Computing - The ferroelectric transistor integrates storage and computation, akin to human brain neurons, which could eliminate the efficiency bottleneck caused by the separation of storage and computation in traditional architectures [1] - The ultra-low operating voltage and energy characteristics of the new ferroelectric transistor provide a core component solution for building high-efficiency data centers and lay the technical foundation for the next generation of high-performance AI chips [2]

Core Insights - The rapid development of artificial intelligence is challenging traditional chip architectures due to the "power wall" and "storage wall," which limit efficiency and increase energy consumption [1] - A breakthrough has been achieved by a research team led by Professor Peng Hailin at Peking University, who developed the world's first wafer-scale ultra-thin and uniform bismuth-based two-dimensional ferroelectric oxide, enabling the construction of high-speed ferroelectric transistors with ultra-low operating voltage and high durability [1][2] Group 1 - The new ferroelectric oxide exhibits a dielectric constant of up to 24 and structural stability at temperatures exceeding 600°C, while maintaining excellent ferroelectric properties at a single crystal cell thickness of approximately 1 nanometer [2] - The research team has created a high-performance array of ferroelectric transistors that outperform other storage technologies by 1 to 2 orders of magnitude, demonstrating 32 stable multi-level storage states and over 10 years of data retention capability [3] Group 2 - The devices can withstand 1.5 trillion cycles under conditions of 0.8 volts ultra-low voltage and 20 nanoseconds high-speed writing, exceeding the stringent standards for cloud AI computing [3] - The team has also developed dynamically reconfigurable in-memory logic circuits, allowing the same device to perform logic operations and switch to non-volatile storage under conventional CMOS voltage below 1 volt, paving the way for a new paradigm in adaptive intelligent chips [3]

Core Viewpoint - The third phase of Yangtze Memory Technologies Co., Ltd. (YMTC) is set to be completed and put into production this year, which will attract around 200 upstream and downstream enterprises to the region [1] Group 1: Company Overview - Yangtze Memory was established in July 2016 and is a leading company in China's storage chip manufacturing sector, primarily providing 3D NAND flash wafers, embedded storage chips, and solid-state drives [1] - The second phase of YMTC was established in December 2021 with a registered capital of 60 billion yuan, while the third phase was established in September 2025 with a registered capital of 20.72 billion yuan [1] Group 2: Market Position and Projections - By 2025, YMTC is expected to hold approximately 7% to 8% of the global market share in storage capacity, with projections indicating it could exceed 10% by 2026, potentially surpassing Micron Technology to become the fourth-largest storage chip manufacturer globally [1] - Market research firm Omdia estimates that YMTC's capital expenditure will be more aggressive than its global peers, accounting for about 20% of total global NAND flash investment by 2025, with expectations for continued growth [3] Group 3: Industry Context - The global demand for storage chips is on the rise due to the ongoing AI computing power boom, marking a new growth cycle for the industry [4] - NAND flash prices are projected to increase by over 40% in the first quarter of this year, and the global storage chip market is expected to reach 1.584 trillion yuan by 2031, with a compound annual growth rate of 9.3% from 2025 to 2031 [4] Group 4: Strategic Initiatives - The company aims to enhance its product matrix and expand both domestic and international markets while focusing on technological innovation and the development of next-generation flash memory chips [3] - The recent establishment of a joint-stock company indicates that YMTC has completed its shareholding reform, and it has entered the list of China's top ten unicorns with a valuation of 160 billion yuan [4]