← today今日

Archive往期

2026-07-26 NewsNational AI Stacks Pull Supply Chains Inward国家级AI体系牵引供应链内聚
2026-07-25 NewsAI Infrastructure Becomes a Full-Stack CommitmentAI基础设施转向全栈投入 ResearchMemory Traffic Becomes the Optimization Target内存流量成为优化主目标
2026-07-09 NewsAI Demand Reprices the Chip StackAI需求重定价芯片供应链 ResearchInference Efficiency Moves Into Memory and Tools推理效率转向存储与工具链
2026-07-08 NewsAI Build-Out Meets Physical LimitsAI扩张撞上物理约束 ResearchInference Workarounds Move Closer to Hardware推理优化逼近硬件边界
2026-07-07 NewsAI Memory Turns Demand Into CapexAI存储把需求转成资本开支 ResearchInference Efficiency Meets Physical-Layer Constraints推理效率遇上物理层约束
2026-07-06 NewsMemory Scarcity Sets the Fab Agenda存储紧缺重写晶圆厂议程 ResearchEfficiency Moves Up the Compute Stack效率优化上移到计算栈
2026-07-05 NewsAI Demand Presses the Whole StackAI需求压紧整条芯片栈 ResearchInference Efficiency Moves Below the Model推理效率下沉到模型之下
2026-07-04 NewsAI Demand Tests the Supply StackAI需求考验供应链全栈 ResearchMemory-Centric Compute Moves Closer to Silicon以存储为中心的计算走近硅片
2026-07-03 NewsAI Demand Pulls the Supply Chain TighterAI需求继续收紧供应链 ResearchInference Bottlenecks Move Into Memory and Packaging推理瓶颈转向存储与封装
2026-07-02 NewsMemory Pressure Rewrites the AI Stack存储压力重写AI堆栈 ResearchMemory Movement Defines the Research Agenda数据搬运主导研究议程
2026-07-01 NewsInference Spreads Across Packaging, Memory, and Policy推理需求牵动封装、存储与政策 ResearchNeuromorphic Work Moves From Models To Blocks神经形态从模型走向模块
2026-06-30 NewsAI Demand Pulls the Memory Stack TightAI需求拉紧存储供应链 ResearchHardware Workloads Move Closer to the Metal硬件工作负载贴近底层
2026-06-29 NewsAI Demand Forces Supply Chain DecisionsAI需求逼出供应链决策 ResearchEfficient Hardware Meets Reliability Pressure高效硬件遇上可靠性压力
2026-06-28 NewsMemory Scarcity Pulls Policy Into the Stack存储短缺把政策拉进供应链 ResearchData Movement Sets the Research Agenda数据搬移主导研究议程
2026-06-27 NewsAI Constraints Move Through the StackAI约束沿堆栈扩散 ResearchHardware Research Chases Memory Locality硬件研究追逐存储局部性
2026-06-26 NewsCapacity Pressure Spreads Beyond GPUs产能压力越过GPU向外扩散 ResearchEfficiency Work Moves Up the Stack效率研究沿计算栈上移
2026-06-25 NewsCustom Silicon Pushes Against Memory Limits定制芯片逼近内存边界 ResearchMemory Paths Shape AI System Design内存路径重塑AI系统设计
2026-06-24 NewsAI Demand Reprices the Entire Chip StackAI需求重估整个芯片栈 ResearchMemory Movement Becomes the Design Target数据搬移成为设计靶心
2026-06-23 NewsAI Infrastructure Redirects Capital Across the StackAI基础设施重塑全栈资本流向 ResearchHardware Co-Design Moves Toward Measured Constraints软硬件协同设计转向实测约束
2026-06-22 NewsAI Bottlenecks Spread Across the Hardware StackAI瓶颈蔓延至整个硬件栈 ResearchScaling Moves Into Devices, Memory, and Context扩展转向器件、存储与上下文
2026-06-21 NewsAI Scale Tests Capacity, Power, and ControlAI规模化考验产能、电力与管控 ResearchHardware Specialization Spreads Beyond Matrix Multiplication硬件专用化走出矩阵乘法
2026-06-20 NewsAI Demand Reprices the Semiconductor StackAI需求重估整条半导体链 ResearchWorkloads Redraw the Hardware Boundary工作负载重画硬件边界
2026-06-19 NewsAI Demand Reprices Control of CapacityAI需求重估产能控制权 ResearchReliability Moves Into the Accelerator Stack可靠性进入加速器堆栈
2026-06-18 NewsAI Demand Reorders the Chip StackAI需求重排芯片供应链 ResearchHardware-Aware AI Moves Into the Loop硬件感知AI进入闭环
2026-06-17 NewsPackaging and Memory Set AI's Pace封装与存储决定AI节奏 ResearchDevice Physics Meets Deployable AI Hardware器件物理走向可部署AI硬件
2026-06-16 NewsAI Demand Stresses Every LayerAI需求拉紧每一层供应链 ResearchMemory Bottlenecks Shape Architecture Research存储瓶颈塑造架构研究
2026-06-15 NewsAI Demand Pushes the Stack to DiversifyAI需求推动全栈分散供应 ResearchHardware Models Meet Real-System Constraints硬件模型直面真实系统约束
2026-06-14 NewsAI Demand Meets Physical and Political LimitsAI需求撞上物理与政策边界 ResearchEfficiency Moves Up the Systems Stack效率优化上移至系统栈
2026-06-13 NewsCapacity Pressure Reprices the Chip Stack产能压力重估芯片供应链 ResearchAdaptive Computation Replaces Fixed Resource Budgets自适应计算取代固定资源预算
2026-06-12 NewsAI Demand Tightens Every Layer of SupplyAI需求收紧芯片供应链各环节 ResearchArchitectures Push Work Closer to Data架构把计算推向数据所在处
2026-06-11 NewsCapacity Pressure Rewrites the Chip Map产能压力正在重绘芯片版图 ResearchData Movement Becomes the Design Variable数据移动成为核心设计变量
2026-06-10 NewsCapacity Expands as Supply Controls Tighten产能扩张与供应管控同步收紧 ResearchMove Less Data, Use Hardware Better少搬数据，更高效地使用硬件
2026-06-08 NewsAI Capacity Becomes an Ecosystem ConstraintAI产能成为全产业链约束 ResearchEfficiency Moves From Models Into Systems效率优化从模型延伸至系统
2026-06-07 NewsCapacity Pressure Spreads Across the Chip Stack产能压力蔓延至芯片全栈 ResearchInference Efficiency Moves Into System Design推理效率进入系统级设计
2026-06-06 NewsAI Hardware Meets Heat, Materials and PermitsAI硬件撞上散热、材料与审批 ResearchPackaging Becomes a Compute Architecture封装正在变成计算架构
2026-06-05 NewsThe AI Memory Wall Goes SystemwideAI内存墙蔓延到全系统 ResearchInference Systems Meet New Memory Devices推理系统遇上新型存储器件
2026-06-04 NewsAI Demand Meets Supply DisciplineAI需求撞上供应纪律 ResearchTensor Cores, Thermal Pixels and Quantized Experts张量核、热像素与量化专家
2026-06-03 NewsMemory and Links Become the AI System存储与互连成为AI系统本体 ResearchWide-Bandgap Devices Move Toward System Limits宽禁带器件逼近系统级极限
2026-06-02 NewsAI Capacity Moves Into Packaging and MemoryAI产能瓶颈转向封装与存储 ResearchLow-Bit Inference Meets Real Edge Silicon低比特推理遇上真实边缘硅片
2026-06-01 NewsComputex Opens; Nvidia Lands on the PC; US Closes the LoopholeComputex开幕：英伟达登陆PC，美国关闭芯片豁免漏洞 ResearchMemristive SNNs Get Practical; FP4 Variants Multiply; Physical AI Reframes the Bottleneck忆阻SNN走向实用、FP4变体不断分化、物理AI重新定义推理瓶颈
2026-05-31 NewsSamsung Snaps HBM4E First; Nvidia Bets $6.5B on US Photonics三星抢发HBM4E，英伟达豪赌65亿美元美国硅光 ResearchMemory Goes Magnetic; MoE Serving Splits Apart存储走向磁性，MoE服务架构走向分裂
2026-05-28 NewsInference Money Lands; Chip Diplomacy Whiplashes推理融资落地，芯片外交剧烈摇摆 ResearchOpen Buses, Ferroelectric Attention, Mechanical Logic开放总线、铁电注意力、机械逻辑
2026-05-27 NewsMemory Crosses the $1 Trillion Line; Taiwan Cracks the Smuggling Pipe存储跨入万亿俱乐部，台湾撬开走私管道 ResearchAfter Scaling: 2D Channels, Orbital Currents, and Photonic Tensorcores工艺微缩之后：2D沟道、轨道电流与光子张量核
2026-05-26 NewsMemory's Heat Wall Cracks Open as Taiwan's AI Rally Eclipses India存储热墙松动，台湾AI行情盖过印度 ResearchThe Memory Hierarchy Goes Analog, Ferroelectric, and Photonic存储层级转向模拟、铁电与光子化
2026-05-25 NewsHuawei Bypasses EUV (On Paper); Micron Says the Memory Crunch Has Years To Run华为宣称'绕开EUV';美光称存储紧张还要持续多年 ResearchThe MoE Serving Stack Hardens — From NAND-CIM to Ascend-Native SchedulingMoE服务栈成型——从NAND CIM到Ascend本地化调度
2026-05-24 NewsAI Chip Boom Continues: Micron Hits Record Margins, Nvidia Eyes ChinaAI芯片热潮持续：美光创纪录利润率，Nvidia押注中国 ResearchData Movement Is the New Compute数据搬移就是新的算力
2026-05-23 NewsUS Memory Fabs Light Up as Huawei Bypasses NAND Sanctions美国存储产能上线,华为绕开NAND管制 ResearchSpecialization Sharpens: Near-Memory, Per-Component, Per-Expert专用化加深:近存储、按部件、按专家
2026-05-22 ResearchPhysics Limits Below, Scaling Infrastructure Above下层逼近物理极限,上层架设扩展基础设施
2026-05-21 NewsAMD Doubles Down on Taiwan as Nvidia Yields ChinaAMD加码押注台湾,英伟达让出中国 ResearchCompute-in-Memory Comes of Age While LLM Serving Treats Hardware as a First-Class Variable存算一体走向系统,LLM服务把硬件当作一等变量