dram

Sam Altman 说 DRAM 厂商虽将进入每年约 1000 亿美元自由现金流的时代，但因过去多次濒临破产的记忆而不敢激进扩产，导致芯片供应链的'看不见的手'失灵，内存短缺无法自我修正。他同时谈到 OpenAI 垂直整合芯片、数据中心与机器人，以及数据中心成本与个人 AI 智能体的前景。

Sam Altman argues that local inference is an inefficient allocation of resources in a constrained environment: equipping everyone's computer with large amounts of scarce DRAM that is only used about 5% of the time is far less effective than keeping cloud resources at high utilization.

Sam Altman argues that running AI models locally is an inefficient use of scarce global DRAM resources and that cloud-based, high-utilization data centers are a better allocation. The post adds a skeptical TL;DR framing the remark as a vendor pitch for renting over owning.

Micron is reportedly dropping its 4F² DRAM cell architecture while competitors push ahead with the technology, marking a notable shift in memory chip strategy.

Samsung has signed a massive five-year deal reserving 70% of its DRAM supply and 100% of its 2027 HBM capacity to big tech firms, with HBM prices expected to jump 150–200% and DRAM prices seeing double-digit increases, likely raising GPU and device costs.

长鑫存储（CXMT）宣称在4F² DRAM定价上取得突破，引发三星、美光等海外存储厂商股价下跌，但长鑫自身股价近期反而走低，引发市场质疑。

Elon Musk 预言中国到2028年将大量生产 RAM，援引韩国财经专栏警告：新产能上线后内存可能在2029年出现供过于求，韩国仅有2-3年时间发展新增长产业。

Micron and Samsung executives say the memory shortage, driven by surging AI demand for HBM and server DRAM, will persist through 2028, with production capacity prioritized for AI over consumer devices.

Samsung Electronics is expected to more than double output of its HBM4 and HBM4E DRAM chips next year, with glass carrier demand rising 2.5 times, indicating significant production growth for high-bandwidth memory in response to AI and computing demands.

Samsung presented a 16 GB LPDDR5X memory package with integrated compute units at Hot Chips 2026, offering 614 GB/s internal bandwidth to significantly improve AI inference performance by overcoming memory bandwidth limitations.

At Hot Chips 2026, Samsung presented opportunities to utilize unused area on HBM base dies after moving to a logic node, with phases including integrating memory controllers to optimize PHY area and power.

The article analyzes the progress of China's memory industry in next-generation architectures such as 4F² DRAM and Xtacking, arguing that CXMT and YMTC have transformed from followers into active participants in new architectures, but still lag behind Korean manufacturers in HBM, advanced packaging, and high-end customer certification.

Report says Samsung, SK Hynix, and Micron have sold out all 2027 DRAM and HBM capacity to AI companies, worsening the RAM shortage and driving up consumer prices for memory, SSDs, and consoles.

This academic paper bridges the gap between experimental characterization and device-level modeling of DRAM read disturbance phenomena (RowHammer and RowPress), presenting TCAD simulations to explain bitflip directions, counts, and minimum aggressor activations, with implications for mitigation techniques.

Chinese memory chipmaker CXMT shares surged nearly 470% on its IPO debut, making it mainland China's most valuable listed company amid strong investor demand and government push for tech self-reliance.

SK Hynix, a major supplier of memory chips for Nvidia's AI accelerators, had a successful US IPO, opening at $170 per share and reaching a $1 trillion valuation amid surging demand for DRAM and HBM driven by AI data center buildout.

AI infrastructure investment is driving up demand for memory (DRAM, NAND), causing aggressive price increases. UBS forecasts DRAM and NAND price hikes in Q3 and Q4, with the DRAM market expected to remain undersupplied until at least 2028.

According to Bernstein, SK Hynix has a 90% profit margin on DRAM, highlighting high profitability in the memory chip market.

A new lawsuit accuses Samsung, SK Hynix, and Micron of colluding to fix DRAM prices, with potential impacts on AI demand for memory chips.

Micron's VP and GM of Cloud Memory Products discusses how AI demand is driving a memory super cycle, with technologies like HBM, LPDDR6, and PCIe Gen 6 SSDs becoming critical for data center architectures.