市场研究机构Counterpoint Research 11日发布数据：人形机器人行业的DRAM内存需求将增长约20倍，从今年约8300 Tb（太比特）增至2030年的172000 Tb。
该预测反映两大趋势：人形机器人开始搭载机器人基础大模型——依托物理AI打造的高性能“大脑”，这提升了单台设备所需DRAM容量；同时机器人出货量本身也开始正式进入增长通道。Counterpoint预计，单台人形机器人所需DRAM容量将翻倍以上，从今年19 GB提升至2030年的39 GB。
Smart Analytics Global的数据显示，今年全球人形机器人出货量将达到6万台，为去年的3倍；到2030年，出货量将增长至约50万台。
业内人士表示，这一趋势将大幅拉动低功耗DRAM（即LPDDR，智能手机常用内存）需求，因为能效水平很大程度决定机器人性能。大容量NAND闪存也将在机器人领域开辟新市场。当前HBM面临容量瓶颈，NAND闪存作为AI存储配套方案，正受到市场关注。
内存厂商已着手布局。三星电子7月曾表示，公司正前瞻性研发面向机器人的下一代半导体整体方案，涵盖高带宽低功耗内存、图像传感器、晶圆代工服务以及先进封装。其设备体验事业部（Device eXperience）牵头推进该市场，并专门设立机器人体验业务部，开展人形机器人相关研发。
SK海力士计划为英伟达Jetson Thor机器人芯片平台联合开发下一代内存产品。美光在最新财报电话会上提到，L4及以上级别自动驾驶车辆通常搭载200 GB以上内存与数TB存储空间，并补充称，人形机器人所需内存与存储容量将达到相近水平。
本文由 AI图鉴自动抓取并翻译，版权归原作者所有；如需查看原文请点下方链接。
Market Research Institute Counterpoint Research 11, publish: The DRAM memory demand in the human robotics industry will grow by about 20 times, from about 8,300 TB this year to 172000 TB by 2030.
This projection reflects two major trends: the human robots began to carry the robotic base LLM, a high-performance “brain” built by Physical AI, which increased the DRAM capacity required for single-stage equipment; and the robotics themselves began to formally enter the growth path. Counterpoint expects that the DRAM capacity required for single-person robots will more than double, from 19 GB this year to 39 GB by 2030.
Data from Smart Analytics Global show that global human robots will be delivering 60,000 units this year, three times as much as last year; by 2030, the volume will have grown to about 500,000.
This trend, according to industry sources, will significantly boost the demand for low-capacity DRAM (i.e., LPDDR, a commonly used memory of smartphones) because energy efficiency determines robotic performance to a large extent. The large-capacity NAND flash memory will also open up new markets in robotics. The current HBM is facing capacity bottlenecks and NAND flashing as an AI storage package is receiving market attention.
The memory manufacturer has started to layout. In July, Tristar Electronics indicated that the company was forward-looking in developing the next generation semiconductor package for robotics, covering high-bandwidth low-capacity memory, image sensors, crystal round-generation services and advanced sealing. Its Department of Equipment Experiences (Device EXperience) leads the market and specializes in robotic experience operations for human robot-related research and development.
KK Hercules plans to jointly develop the next generation of memory products for the Jetson Thor robot chip platform in Weida. As mentioned at the latest financial conference, automatic driving vehicles at the L4 level and above usually carry memory above 200 GB and several TB storage spaces, adding that the memory and storage capacity required for human robots will be similar.
Auto-fetched and translated by AI Atlas. All rights belong to the original author.