memory

Trader degentrading (@degentradingLSD) pre-market analysis shows the following major signals in the current market:Macroeconomic Level: The 10Y yield has risen to 5.2%, and the 30Y has risen to 5.5%, approaching historical highs; the South Korean stock market opened down 2.5%, with the memory and semiconductor sectors (SK Hynix, Samsung, MU, SNDK) generally down 2% to 5%. It is expected that price movements will continue to fluctuate before MU's earnings report is released.New Cloud Computing: The investment payback period for computing projects is now less than a year, and market perceptions are changing. It is expected that the new cloud sector will undergo rapid revaluation, and the issue of power bottlenecks is also receiving increasing attention.Tech Stocks: META is weak pre-market, focusing on buying at the 690 support level or opportunities for breaking historical highs. If the large-scale data center sector breaks through as a whole, META, MSFT, and others will face paradigm-level repricing.Cryptocurrency Market: Mainstream coins and altcoins are generally weak, with BTC possibly testing the 82K support level; last week, TAO and DOGE were reduced, and CRDO was taken for profit to release risk capital. During the KBW period, crypto sentiment is usually somewhat positive, and this week's strategy focuses on finding short-term bullish opportunities.

According to Slow Fog Security Alert, the AI memory toolchain under MemTensor has been compromised. The open-source long-term memory library MemoryOS (PyPI) aimed at LLM and AI Agent, as well as the official plugin memtensor/memos-cloud-openclaw-plugin (npm) that connects to the OpenClaw runtime, have been implanted with a cross-platform Go binary, which triggers execution when the package is loaded or imported.Affected versions include: MemoryOS==2.34 on PyPI, and npm plugins 0.1.21, 0.1.23, 0.1.25. Among them, the affected npm plugins may also cause user prompt content leakage. Slow Fog recommends taking the following measures: uninstall or downgrade to known safe versions (npm 0.1.20, PyPI 2.33); terminate the sckit process; block related infrastructure; check network activity; and rotate credentials in the affected environment.

In episode 166 of the BidClub podcast, chip designer Bubble boi stated that high-bandwidth flash is technically attractive but lacks decisive data center use cases. The information he provided indicated that the main customers he heard about are Chinese clients who find it difficult to obtain or produce high-end HBM.Bubble boi mentioned that HBF and high-end AI flash have a latency of about 5 microseconds, lower bandwidth than HBM, a lower cost per bit, but the total cost may be comparable. He believes the clearest customers are entities unable to obtain HBM, which aligns with the Chinese demand he heard from two to three people, as they cannot adequately manufacture accelerators or are willing to accept non-leading specification products.

The global memory market continues to face severe shortages, with AI and cloud applications driving demand and prices for DRAM and NAND Flash. China's two major memory manufacturers, Yangtze Memory Technologies and Changxin Memory Technologies, are set to break the current duopoly.Yangtze Memory Technologies will begin trial production of LPDDR5 DRAM by the end of 2026, while Changxin Memory Technologies aims to transition from DRAM to NAND Flash. Industry speculation suggests that the two companies may face off as early as 2028. Yangtze Memory Technologies has completed the development of 294-layer 3D NAND using Xtacking technology, and Samsung will enter into a patent licensing agreement with them in 2025, starting with the 10th generation V-NAND.Yangtze Memory Technologies' phase three plan is set to commence production by the end of 2026, with an initial capacity of approximately 25,000 wafers of DRAM, ramping up in 2027, targeting LPDDR5, and planning for subsequent expansions as well as HBM and 3D DRAM. Changxin is reported to be setting up a NAND research and production line in Beijing, with mass production expected as early as 2028 to 2029.

Micron and the Taiwan Micron Memory Union held their first mediation meeting on September 18. Micron stated that both parties engaged in constructive discussions that helped advance communication and agreed to hold a second mediation meeting on October 22.Micron pointed out that representatives from both sides discussed several topics during the meeting, including the overall compensation and bonus system framework, No Free Rider, and Non-Retaliation issues. Regarding Non-Retaliation, Micron reiterated the company's commitment that according to the Micron Global Team Member Handbook, team members have the right to peacefully assemble, freely associate, form and join unions or labor organizations of their choice, and seek collective bargaining or other forms of representation without fear of retaliation.Micron expressed gratitude to the Central Science Park Administration and the mediation committee for their assistance and support during this mediation process. Moving forward, both parties will continue to uphold a spirit of mutual respect, support employees, serve customers, and jointly promote the company's ongoing success through positive and pragmatic communication.

At the 2026 World Manufacturing Conference, memory chip manufacturer Changxin Technology announced the official mass production of its fifth-generation technology platform.According to reports, Changxin Technology has achieved a half-pitch of 11.95 nanometers in the active area of the memory array, with a capacitor aspect ratio of 45:1, and the height of the core kinetic area has been reduced to 6762 nanometers. The storage density of the fifth-generation process platform has significantly improved, with each wafer producing over 50% more than the previous generation platform under the same conditions.The 24GB LPDDR5X product developed based on this platform has entered formal mass production and is being used in mainstream domestic flagship smartphones.

Acer Chairman Chen Junsheng stated on the 19th that there is no shortage of memory, as suppliers in mainland China are continuously releasing quantities, acting as price destroyers. The contract prices are currently experiencing high fluctuations, with chaotic ups and downs. Some components like LPDDR5 or DDR5 9600 paired with N1 or N1X are still in short supply, but more common components like DDR4 are in oversupply. Currently, there is also no shortage of CPUs, except for certain low-priced licensed models from Microsoft, such as those with Small Core specifications, which are in tight supply.Chen Junsheng pointed out that the price increases for key components this year are unprecedented, with customers going from complaining to scrambling for goods. Acer is increasing its inventory of low-priced components in response, facing pressure to raise PC prices each quarter, with different products still seeing price increases of 5% to 20% in the fourth quarter. The three major memory suppliers have expressed optimism about price increases until 2027, but the unlimited production capacity released by mainland China means that sustained price increases are unlikely.He expects that by 2027, Acer's costs for SSDs and memory will decline compared to this year, with a chance for prices to stabilize in the second half of next year. There may still be opportunities for price increases in the first quarter of 2027, but the increases should become narrower. The expansion of semiconductor production is expected to begin in mid-2027, after which prices may peak and enter a stable phase.

Micron Technology will announce its fiscal fourth-quarter results on September 30. NVIDIA CFO Colette Kress stated during the fiscal second-quarter conference call on August 26 that the company is seeing "extreme pricing conditions in memory," with price increases exceeding previous expectations, and that "will further rise next year." Pricing pressure has been reflected in NVIDIA's margins, with Kress guiding that gross margins will bottom out in the fourth quarter in the range of 71%-72%, before recovering to 72%-73% in fiscal year 2028.Kress indicated that the current memory shortage is largely driven by AI infrastructure itself. SK Hynix expects the memory market to reach supply-demand balance no earlier than 2030 and has invested approximately $38.3 billion to expand its two domestic bases in South Korea through 2031. SK Hynix holds about 50% of the high-bandwidth memory market share, leading Samsung's 33% and Micron's 18%.

According to Reuters, citing informed sources, Changxin Storage (CXMT) is preparing to enter the NAND flash memory chip market, expanding beyond its core DRAM business and competing with Samsung Electronics, SK Hynix, Micron, and Yangtze Memory Technologies.Changxin Storage plans to establish a NAND flash research and production line at its new factory in Beijing and set up a research institute in Beijing that includes NAND development projects. The company has discussed related plans with clients, including a startup that intends to procure chips for AI systems and supercomputers. The timeline for the operation of the R&D line or large-scale commercial manufacturing remains unclear.The global storage shortage is driven by demand for AI servers and is expected to last at least until 2027. Manufacturers prioritize DRAM and HBM spending, limiting the increase in NAND capacity. Changxin Storage raised 57.92 billion yuan in its IPO in July. Yangtze Memory Technologies was placed on the entity list in 2022.

Vitalik posted on the X platform, expressing the hope to include the recursive STARK memory pool proposal EIP-8288 in the Ethereum I-star upgrade, stating that it can be seen as the next step after Frames.This proposal allows quantum-safe signature data to be placed off-chain, reducing the cost of signatures; the fuel cost for quantum-safe privacy transactions is also expected to drop from about 10 million to tens of thousands. The proposal relies on frames and recursive STARK aggregated transaction dependencies, with nodes sending a STARK of about 100 to 300KB each time period; the on-chain overhead is one STARK and 96 bytes for each statement to be proven.

According to the Financial Times, citing informed sources, Chinese memory chip manufacturers Changxin Storage and Yangtze Memory Technologies have purchased enough deep ultraviolet lithography machines (DUV) from ASML to meet their expansion plans for the next three years.

KB Securities warned that the inventory of memory semiconductors at Samsung Electronics and SK Hynix has fallen to less than 10 days of supply, and the available supply next year will be significantly insufficient. The agency reported on Monday that investment in artificial intelligence infrastructure is expanding at an unprecedented pace, leading to a severe supply shortage in the memory chip market. A key factor is the transition to HBM4, which requires about three times the wafers of traditional DRAM, and limited wafer capacity will reduce the available capacity of traditional DRAM.KB Securities expects that next year, the demand for DRAM and NAND will exceed supply by more than 10 percentage points. Research director Kim Dong-won stated that artificial intelligence servers will consume HBM, server DDR5, and enterprise-grade solid-state drives, potentially leading to a historic shortage. Global hyperscale data center operators have raised their expectations for artificial intelligence infrastructure investment next year to $1.3 trillion, a 60% increase from the previous year. The agency anticipates that memory semiconductors will account for 57% of total investment in artificial intelligence infrastructure next year, up from 14% last year, with TrendForce estimating this ratio could reach as high as 68%. Samsung Electronics' stock price has dropped nearly 28% from its peak, while SK Hynix has fallen nearly 40%.

According to Blue Whale News, from Changxin Technology, ByteDance and ZTE have jointly developed a new generation of Doubao phone------Nubia NaviX Ultra, which will be equipped with Changxin's highest speed LPDDR5X memory at 10667Mbps. The official statement indicates that this is also the first time that domestically produced 10667Mbps high-speed LPDDR5X memory chips have achieved mass production and application.Previous news reported that the Nubia "Doubao phone" has completed network filing and will be mass-produced and launched in September.

The IPO review status of Yangtze Memory Technologies Co., Ltd. for its initial public offering and listing on the Sci-Tech Innovation Board has been changed to "Inquired.""Inquired" means that Yangtze Memory has officially entered the listing review stage at the Shanghai Stock Exchange, where a round or multiple rounds of written Q&A will be conducted regarding the company's information disclosure, Sci-Tech attributes, and other issues.Yangtze Memory is China's largest 3D NAND flash memory manufacturer and the only 3D NAND original factory. It has independently developed the Xtacking architecture and is a memory IDM company that integrates chip design, production manufacturing, and packaging testing.

Google's Senior Director of Supply Chain Infrastructure, Nikhil Cherian, revealed that to overcome memory bottlenecks, Google is developing software and hardware solutions and dismantling retired servers to recycle DDR4 components to establish an internal recycling supply chain. Google has designed special hardware adapters to connect the previous generation DDR4 to the new generation of AI servers while importing retired servers to remove their DDR4 modules for recycling.Cherian stated that the AI industry has rapidly shifted from being compute-constrained to memory-constrained, with high-performance memory accounting for about 75% of the bill of materials cost for a given AI server. A Goldman Sachs report indicated that memory prices will continue to rise in the third quarter, with personal computer DRAM prices expected to increase by 18% to 23% and server DRAM prices expected to rise by 13% to 18%. Trendforce data shows that in August, the spot market prices for DDR4 8GB and DDR5 8GB rose to $142 and $133, respectively.The two TPU ASICs launched by Google this year have been optimized for memory design, claiming to reduce memory demand to one-sixth of the original. The TPU8i chip features a dedicated layered memory design that relies on a high-speed DDR5 memory architecture to perform host-level tasks, with each chip equipped with 288GB of HBM3e high-bandwidth memory.

The SEMICON Taiwan 2026 Memory Summit took place on the 1st, where Nikhil Cherian, Senior Director of Supply Chain Infrastructure at Google Cloud under Alphabet, pointed out that with the popularity of multimodal and mixed expert architectures, AI computation has shifted from being power-limited to memory-limited, with high-performance memory accounting for over 75% of the cost of AI server hardware bill of materials. In the face of capacity, bandwidth, and power consumption bottlenecks, Google is breaking through the AI memory bottleneck through a dual-track strategy of hardware offloading for inference and training, and lossless quantization software algorithms.Google adopts an offloading strategy in hardware architecture, launching TPU 8i for low-latency inference and TPU 8t specialized for large-scale training. The TPU 8i is equipped with 288 GB of high-bandwidth memory, with SRAM capacity on the chip increased threefold to 384 MiB, placing dynamic conversation states and key-value caches on the chip itself to achieve zero chip-off latency. The TPU 8t forms a super-large computing cluster with 9600 chips, achieving a shared pool of HBM at a scale of 2 PB, eliminating chip-off data transfer bottlenecks, along with TPU Direct Storage technology.Google has developed the training-free TurboQuant lossless quantization algorithm, compressing the key-value cache of large models from 32 bits to 3 bits, reducing memory usage by six times without loss of accuracy, resulting in an eightfold acceleration in attention computation, and integrating old-generation DRAM technology to extend the lifecycle of components.

Google announced two new quality requirements for Android applications to address the industry memory chip shortage triggered by the AI data center boom. One focus is on reducing in-app memory usage and optimizing code. Google stated that the mobile industry is facing significant hardware supply constraints, which are changing the availability of device memory and, in turn, affecting user experience.To this end, Google has set new performance thresholds in areas such as dynamic memory usage and bitmap usage, and has increased code optimization requirements to prevent app stuttering and crashes. The company will launch new tools to alert developers to optimize when applications exceed the thresholds. More diagnostic tools will be released later this year, including a Memory Limiter feature that prevents applications from excessively occupying device memory.Developers must meet the new thresholds by February 2027. Additionally, all Play Store applications must comply with the Zero Tap Sign-In requirement by April 2027, which automatically restores user login status during device migration via the Android Restore Credentials API. These changes aim to help app and game developers continue to provide high-quality experiences in a market with limited memory supply, especially targeting price-sensitive low-end devices.

According to Nikkei Asia, SK Hynix CEO Kwak Noh-Jung stated that the global shortage of memory chips is expected to last until the end of 2030, with a low risk of oversupply. He pointed out that in the AI era, memory chips are no longer just "commodities," and there are currently no signs of oversupply or industry downturn; demand from AI customers remains strong.SK Hynix is a major supplier to NVIDIA. The report stated that the company remains open to further expanding its investments in the U.S. after launching its first HBM chip packaging plant in Indiana.

Samsung Electronics and SK Hynix plan to increase the supply of 8-layer HBM4 memory to NVIDIA in the second half of this year.This move is influenced by NVIDIA's supply strategy, aiming to ensure a stable supply of HBM memory while also considering product heat generation.