2026
- April 15 - NVIDIA Acquires Groq for $20 Billion: LPU Officially Enters the NVIDIA Ecosystem
- April 20 - Cerebras IPO Deep Dive: S-1 Filing, $22-25B Valuation, OpenAI $10B Mega-Deal
- April 22 - NVIDIA Vera Rubin Platform Deep Dive: 6-Chip Package, 288GB HBM4, 50 PFLOPS FP4
- April 22 - Google TPU 8t + 8i: The First TPU Generation to Split Training and Inference
- April 22 - AMD MI400 + Helios Rack: 432GB HBM4 + 260 TB/s UALoF Open Interconnect
- April 25 - Apple Silicon Comeback: M3 Ultra 192GB UMA Local LLM Revolution
- April 30 - Inference Optimization Technology Evolution: PagedAttention / FlashAttention / Speculative Decoding Deep Dive
- May 14 - Intel Cancels Falcon Shores, Pivots to Jaguar Shores: From Single-Chip Competition to Rack-Scale Systems
- May 20 - Rack-Scale AI Era: NVL72 vs Helios vs Groq 3 LPX vs Trn3 UltraServer — Four Major Solutions Compared
- May 25 - HBM Three-Way Battle: SK Hynix / Samsung / Micron Fight for AI Memory Supremacy
- May 28 - AI Chip Startup Survival Report: Tenstorrent / SambaNova / Graphcore in 2026
- May 30 - AI Cluster Power Crisis: 1MW Racks, Nuclear Plants, SMRs, and Green AI
- June 1 - NVIDIA Launches RTX Spark: AI Compute Enters the Personal Computer Era
- June 1 - 2026 Global AI Computing Report & Ten Major Computing Industry Trends Released
- June 1 - 2026 H2 Top AI Chip Selection Guide: From H100 to Rubin, MI400, TPU 8t, TPU 8i
- June 4 - Huawei Ascend 950 Mass Production and the Full Picture of China's AI Chip Ecosystem
- June 5 - Computex 2026 AI Compute Card Major Events: DGX Station for Windows, Intel Crescent Island, and More Major Launches
- June 6 - Computex 2026 Wrap-Up: AI PC Chip War Begins, NVIDIA RTX Spark Arrives Fall 2026
- June 7 - AI Hardware Enters the "Era of Deployment": Five Major Shifts of 2026 and the Rules for Survival
- June 10 - Kunlun P800 Deep Dive: Performance Data, Architectural Innovation, and SuperNode Deployment
- June 10 - Huawei Ascend 910C Deep Dive: Specs, Deployment, and Full Performance Overview
- June 10 - 2026 Domestic AI Chip Progress: Huawei Ascend 950, Baidu Kunlun M100, Alibaba T-Head M890 Fully Explained
- June 16 - NVIDIA Vera Rubin Enters Full Production: The Agentic AI Factory Era Begins
- June 16 - Intel Gaudi 4 / Jaguar Shores Latest Progress: Returning to the AI Race with HBM4 Memory
- June 16 - Milestone! Huawei Ascend 910C Completes Full-Parameter Training of a 1.6-Trillion-Parameter Model
- June 16 - Google TPU 8i/8t Officially Launched: Training and Inference Split for the First Time, 2nm Process Powers the Agentic Era
- June 16 - AMD MI455X Stuns at CES 2026: AI Chip Performance Up 1000x in 4 Years
- June 20 - 2026 H2 AI Chip Roadmap Major Update: Qualcomm Enters, AMD MI400 Three Models Unveiled, Huawei Three-Generation Roadmap
- June 23 - Cambricon MLU690 vs NVIDIA H100: In-Depth Comparison — Can a Domestic AI Chip Replace the H100?
- June 23 - China's Domestic AI Chip Triopoly (2026): Ascend, Cambricon, Moore Threads — Who Is the "China H100"?
- June 26 - OpenAI's In-House AI Chip Jalapeño Deep Dive: Taped Out in 9 Months, Inference Cost Cut 50%
- June 28 - June 2026 AI Chip Major Events Roundup: Ascend 910C Trains Trillion-Parameter Model, OpenAI Custom Chip, RTX Spark Launch
- July 1 - 2026 H1 AI Chip Industry Review: Blackwell Ultra, the Domestic Big Three, and the Inference Era
- July 10 - NVIDIA Vera Rubin Officially Ships: First VR200 NVL72 Delivered, Samsung HBM4 Mass Production, Rubin Ultra Cabinet Sky-High Price
- July 10 - Huawei Ascend 950 Series Capacity & Orders Deep Dive: 950PR Monthly Capacity Jumps 10×, ByteDance Locks In 350k Units for $5.6B
- July 11 - Domestic GPU IPO Wave: The "Four Little Dragons" Assemble on Capital Markets, Moore Threads MTT S5000 Benchmarks Against H100
- July 11 - AMD Advancing AI 2026 Opens Tomorrow: Three CDNA5 MI400 Models, Helios Rack Hits 3 exaFLOPS, OpenAI + Meta Lock 12GW Deal
- July 20 - WAIC 2026 Recap: Huawei Atlas 950 SuperPoD Live Hardware Wins SAIL Grand Award, Domestic Compute Enters the "System-Level" Showdown
- August 15 - Inference Accelerator Market 2026: 60%–70% of the Accelerator Market, GPU vs ASIC Share Inverts, Five Schools Clash
- August 18 - Domestic Big Three 2026 H2: Localization Rate Crosses 40% Toward 60%, Ascend 960 Roadmap, MLU690 and S5000 Ecosystems Ramp Up
- August 20 - HBM4 Mass-Production Year One: Samsung Yield Breaks 80%, Three Giants Pass NVIDIA Certification, the Last Bottleneck of AI Compute Supply
- August 22 - Hyperscaler Custom Silicon Wave 2026: OpenAI Jalapeno, Maia 200, MTIA, TPU v8 Together "De-NVIDIA-ize"
- August 25 - Hot Chips 2026 Full Recap: Rubin, MI455X, Crescent Island Together as AI Compute Delivery Enters the "System-Level" Era
- September 4 - Vera Rubin 全面量产：100% 全液冷 + 800V 直流供电，AI 数据中心基础设施范式重构
- September 4 - HBM4 从送样竞速进入量产爬坡竞速：SK 海力士率先交付 Rubin、美光产能翻倍、三星直逼王座
- September 4 - 国产 AI 芯片半年报大检阅：寒武纪净赚 23 亿、壁仞营收暴涨 20 倍、燧原登板，"抢芯大战"白热化
- September 4 - 昇腾 950 超节点 Q4 上市，中兴、新华三、曙光全线跟进：国产超节点从概念走向交付
- September 4 - AMD Helios 正式出货：MI455X 机架 525 万美元一台，OpenAI / Meta / 甲骨文 / Anthropic 12GW 订单进入兑现期
- September 5 - AWS Adds Another 2 Million NVIDIA GPUs: Vera CPU Debuts on AWS, 100,000 GPUs Reserved for the U.S. Government AI Factory
- September 5 - AI 算力周报（9.1-9.5）：DeepSeek 订 16 万颗昇腾 950DT、英伟达 129 亿美元收购 Hugging Face、Rubin Ultra 显存减配
- September 16 - Vera Rubin NVL72 MLPerf v6.1 Debut: Qwen3-VL Throughput 3.7x GB300, CoreWeave Brings Multi-Rack Cluster Online Same Day
- September 18 - 昇腾 960 官宣发布：FP4 算力翻倍、全球首个 NPO 超节点，华为确认一年一代
- September 19 - JD Cloud's 100,000-Card Cluster Bets on Moore Threads: Domestic GPUs Enter a Top AI Cloud's Core Compute Base for the First Time
- September 22 - Alibaba Zhenwu V900 Unveiled at Apsara Conference: 3x M890 Performance, 216GB Memory, 500,000-Card Cluster, Mass Production in 2027 Q1
- September 23 - Three Signals for the Domestic Compute Ecosystem in One Week: China Telecom Open-Sources Ascend-Trained Model, Inspur 128-Chip Inference Appliance, EVAS Raises RMB 2 Billion
- September 23 - DeepSeek Confirms Betting on Huawei Chips for LLM Training: From the 160,000-Unit 950DT Rumor to a "Must Succeed" Commitment
- September 25 - GPU Rental Rates Climb Again: Nebius Raises H100/B200 On-Demand Prices About 20% From October, and the Build-vs-Rent Balance Is Shifting
- September 26 - Groq 3 LPX Enters Full Mass Production: Samsung 4nm Foundry, 315 PFLOPS FP8 per Rack — NVIDIA Turns the LPU into the Seventh Chip of Vera Rubin
- September 26 - Intelligent Compute to Hit 9,800 EFLOPS by 2030: Five Hard Targets in the ICT Industry "15th Five-Year" Plan
- September 28 - NVIDIA's China Share Crashes to 8% as Ascend Rises to 50%: The Dramatic Reshuffle of China's AI Compute Market in a September Report
- October 1 - 高通与AWS达成40亿美元AI联盟：多代定制AI芯片+1.6T光互联
- October 1 - 燧原科技登陆科创板：开盘暴涨188%、市值1764亿元，国产AI芯片资本化再下一城
- October 2 - HBM短缺引爆国产AI芯片涨价潮：昇腾950DT报价超25万元、寒武纪690涨20%-30%
- October 2 - 平头哥公布倚天CPU路线图：倚天720/730明年推出，730首次全自研微架构
- October 3 - 昇腾950智算集群云服务正式商用：9月30日国内上线、11月30日全球商用
- October 3 - 5GW、10GW成为新单位：Anthropic与OpenAI的算力军备竞赛
- October 4 - 2nm时代开启：AMD MI455X成首款TSMC N2 AI GPU，三星/台积电2027量产竞速
- October 4 - OpenAI Jalapeño新细节：13.4 PFLOPS 4-bit算力、232GB内存、用LLM在20个月内造出芯片
- October 4 - AMD创MLPerf推理纪录：512卡MI350实现575万tokens/秒，NVIDIA垄断首次实质性松动
- October 5 - SK海力士发布AI存储"新频谱"：HBF、PIM、SALT-KV重新定义AI基础设施
- October 5 - 定制HBM军备赛：AWS联手NVIDIA开发NVHBM，超大规模厂商开始"造内存"
- October 5 - 摩尔线程"庐山"GPU年底亮相：3A游戏渲染提升15倍、光追性能提升50倍
- October 5 - 长鑫存储进军NAND：国产存储原厂格局生变
- October 6 - 机架即计算机：NVL72、Helios、昇腾超节点的系统级战争
- October 6 - 2026年Q4 AI芯片前瞻：Rubin放量、Helios交付、昇腾950DT上市与国产新窗口
- October 6 - NVIDIA N1X进军PC处理器：发布当日Intel、AMD股价各跌4.2%
- October 6 - Intel年内第三次涨价：CPU上调10%，AI需求下的产能博弈
- October 6 - AMD Helios进入量产：MI455X首批Q3末出货，OpenAI/Meta/Anthropic 12GW承诺兑现中
- October 6 - AI数据中心成为电网参与者：AEMA联盟成立、NVIDIA DSX实现24%吞吐增益
- October 6 - AI芯片周报（2026年第41周）：2nm量产竞速、存储新范式与机架级大战
- October 8 - RTX Spark 笔记本开订、DGX Station 登陆 Windows：NVIDIA 把 AI 超算搬上办公桌
- October 9 - OpenAI Jalapeño 完整规格披露：700W推理ASIC、9个月流片、单pod 27 EFLOPS
- October 9 - NVIDIA 投资竞争对手 d-Matrix、支持多厂商互联 Token Fabric："竞合"成为新打法
- October 9 - 内存超级周期全面爆发：台积电Q3营收创纪录、三星利润暴涨8倍、HBM4E获NVIDIA认证
- October 9 - 徐直军 HC2026 访谈：昇腾950DT年底规模供货，"昇腾中国份额已超英伟达"
- October 10 - xAI 与特斯拉 Dojo 2：非典型玩家的自研芯片路线
- October 10 - 晶圆级 vs 芯粒：两条极限算力路线的碰撞
- October 10 - 语音 AI 推理硬件：ASR 与 TTS 服务的算力配置
- October 10 - vLLM 部署实战：不同硬件上的适配矩阵与调参要点
- October 10 - 视频生成与世界模型：下一个算力黑洞
- October 10 - Vera CPU 深度：SCF + 1.5TB LPDDR5X，NVIDIA 自研 CPU 的野心
- October 10 - Vera CPU 智能体基准实测：10.8秒拉起千个沙盒，速度达 AMD EPYC 8倍
- October 10 - 二手 H100 保值之谜：AI 芯片为何成了"硬通货"
- October 10 - 高校实验室算力配置指南：30 万预算能买到什么
- October 10 - 统一内存大乱斗：128GB、192GB、748GB、288GB 谁才是最优解？
- October 10 - 台积电海外建厂与 AI 芯片供给：亚利桑那、熊本与德累斯顿
- October 10 - Triton Inference Server 还是自研网关：推理服务层的选型
- October 10 - 变压器与 UPS 短缺：数据中心扩容的隐形瓶颈
- October 10 - 训练与推理的算力占比之变：推理时代何时真正到来
- October 10 - 2026 训练 GPU 选型指南：从 H100 到 Rubin，预算百万到亿级怎么选
- October 10 - 每瓦 Token 数：AI 芯片能效的新标尺
- October 10 - Tenstorrent Blackhole 与 Jim Keller 的开源 RISC-V 路线
- October 10 - 从 Tensor Core 到矩阵引擎：AI 算力单元十五年演进
- October 10 - AI 算力 TCO 深度测算：3年总拥有成本的隐藏项
- October 10 - 初创公司第一台 8 卡服务器怎么配：预算、卡型与避坑
- October 10 - 投机解码深度解析：小模型如何给大模型"代打"
- October 10 - Spectrum-X 与 NVLink Fusion：英伟达的网络与开放策略读解
- October 10 - 主权 AI 全球圈地：从中东到东南亚的算力新地图
- October 10 - 算能 BM1684X：边缘 AI 盒子里的隐形冠军
- October 10 - 单卡功耗天花板：从 700W 到 1400W，散热路线到了十字路口
- October 10 - SGLang 高并发推理：RadixAttention 与生产环境实践
- October 10 - 服务器 CPU 变天：EPYC Venice 256核 2nm 量产 vs Intel 三度涨价仍失份额
- October 10 - 燧原、壁仞、登临、天数智芯：二线国产 AI 芯片厂商动态盘点
- October 10 - 三星的 HBM 追赶战：失位、反扑与 1c HBM3e 之后
- October 10 - SambaNova SN40L 深度解析：三层内存的另辟蹊径
- October 10 - 从 Rubin Ultra 到 Feynman：NVIDIA 2027-2028 路线图全解读
- October 10 - Rubin R200 深度解析：3360亿晶体管、288GB HBM4、50 PFLOPS FP4 的代际跨越
- October 10 - Rubin CPX：专攻长上下文推理的差异化设计，与它跌宕的 2026 年
- October 10 - RTX Spark 笔记本首发潮：微软 Surface 对标 MacBook Pro、联想 YOGA Pro 15 三年磨一剑
- October 10 - RTX PRO 6000 Blackwell：工作站级 AI 算力天花板的正确用法
- October 10 - Ray 与 Kubernetes：GPU 集群弹性调度的两条路线
- October 10 - Rail-optimized 网络架构：万卡集群的布线哲学
- October 10 - 量子计算与 AI：远水与近火的交汇点在哪里
- October 10 - 量化实战：GPTQ、AWQ 与大模型压缩的落地取舍
- October 10 - PyTorch 编译栈与自定义硬件后端：torch.compile 之外的世界
- October 10 - 制程竞赛 2026：N2 量产、18A 蓄力、High-NA EUV 百万晶圆
- October 10 - 电比卡贵：AI 数据中心的能源危机与自救
- October 10 - OpenAI 商业化全景：年化营收冲击700亿美元，Q2运营亏损123亿美元
- October 10 - 开源模型与国产芯片的飞轮：DeepSeek 效应的产业启示
- October 10 - CUDA 城墙下的三条隧道：CANN 开源、ROCm 补课与 PyTorch 中立层
- October 10 - 从 ONNX 到 TensorRT：视觉 AI 的部署流水线拆解
- October 10 - 端云协同新格局：手机 NPU 与云端推理的分工边界
- October 10 - AI 服务器 OEM/ODM 的收入海啸：广达、纬创、超微们的机遇与隐忧
- October 10 - 一颗 GPU 背后的七颗芯片：NVIDIA "七芯片堆栈"战略全拆解
- October 10 - HGX 与 MGX：英伟达双平台参考设计怎么选
- October 10 - NVIDIA H200 深度解析：141GB HBM3e 的长青主力与它在中国市场的曲折命运
- October 10 - 英伟达财报里的数据中心：收入结构与客户集中度解读
- October 10 - NVIDIA 承诺最高1000亿美元投资 OpenAI：10GW 算力、非控股股权、首吉瓦 Rubin 交付
- October 10 - 光进铜退：NPO/CPO 光互连如何重塑超节点架构
- October 10 - 多模态推理卡怎么选：VLM、语音与视频生成的硬件差异
- October 10 - 多租户 GPU 池化：算力切分、隔离与计费
- October 10 - 摩尔线程双线作战：MTT S5000 智算卡与"庐山"图形 GPU 的AI+游戏并举
- October 10 - 摩尔线程 S80 到 S4000：从消费级争议到企业级落地
- October 10 - MoE 架构对芯片意味着什么：稀疏激活的算力新账
- October 10 - 寒武纪 MLU690 深度解析：700+ TFLOPS FP16 与国产训练卡的寒武纪时刻
- October 10 - MIG 与算力切分：一张 A100/H100 如何服务多个租户
- October 10 - 微软 Maia 200 量产落地：Azure 自研推理芯片的关键一跃
- October 10 - 微软 Fairwater：为 AI 重新发明的数据中心
- October 10 - 美光的 HBM4 豪赌：产能、产地与市场格局
- October 10 - MI455X 深度解析：CDNA 5、432GB HBM4、40 PFLOPS MXFP4，AMD 数据中心旗舰全拆解
- October 10 - MI400 系列盘点：MI430X 驱动欧洲 LUMI-AI 超算，AMD HBM4 家族成型
- October 10 - Meta MTIA v3 进入量产：拆解式架构背后的自研推理芯片雄心
- October 10 - HBM-DRAM-NAND 三层重构：KV Cache 如何改变存储金字塔
- October 10 - 显存带宽 vs 容量：推理瓶颈的一体两面
- October 10 - MCP 与 AI Agent 基础设施：当智能体成为算力的调用方
- October 10 - 微软 Maia 100 架构复盘：云巨头的第一块自研 AI 芯片
- October 10 - LoRA 与 SFT 微调用什么卡：24GB 到 141GB 的梯度选择
- October 10 - 长上下文的硬件账本：稀疏注意力如何改写显存需求
- October 10 - 大模型 API 还是自建推理：一份可以套用的成本对比框架
- October 10 - 从风冷到液冷：AI 机柜功耗狂飙下的散热革命
- October 10 - Lightmatter 与硅光互连：光进入 AI 集群的三种方式
- October 10 - 互联网巨头自研芯片盘点：百度昆仑芯 R200 与阿里平头哥真武 V900
- October 10 - 昆仑芯 M100/M300 深度解析：百度三代自研的集大成
- October 10 - 韩国 AI 芯片整合样本：Rebellions 并购 Sapeon 后的进击路线
- October 10 - Rapidus 冲刺 2nm：日本半导体国家队的 AI 芯片雄心
- October 10 - 互联标准战争：NVLink、UALink、UEC 与灵衢的四方博弈
- October 10 - Intel AI 芯片十字路口：Gaudi 系列承压、Jaguar Shores 待发、Crescent Island 补位
- October 10 - 推理芯片创业浪潮：d-Matrix 获 NVIDIA 入股、Taalas/CScale/Volantis 融资全景
- October 10 - 推理服务 SLA 与弹性扩缩容：从 P99 延迟说起
- October 10 - 2026 推理 GPU 选型指南：Prefill/Decode 分离时代的卡怎么挑
- October 10 - 工业视觉质检的算力选型：从 Jetson 到国产边缘卡
- October 10 - 印度主权算力提速：从 4.5 万张 GPU 共享池到本土芯片制造
- October 10 - 天数智芯天垓 150 深度解析：国产 GPGPU 的实干路线
- October 10 - 海光 DCU 深度解析：CUDA 兼容路线的独特价值
- October 10 - Helios 机架架构全解：72颗 MI455X、31TB HBM4、UALink 开放生态对垒 NVL72
- October 10 - HBM4E 与 HBM5：2027 之后的存储路线图与定制化趋势
- October 10 - 从 HBM2e 到 HBM4E：一文读懂 AI 显存的四代演进
- October 10 - 三代同堂：H100、H200、B200 采购对比与代际提升
- October 10 - Groq LPU 并入 Rubin 平台：LPX 机架专攻智能体"最后一公里"解码
- October 10 - 绿色算力：AI 数据中心的能耗指标与碳账本
- October 10 - Graphcore IPU 的兴衰：被 SoftBank 收购后的现状
- October 10 - 拆一台 GPU 服务器：从供电、散热到 NVSwitch 托盘
- October 10 - GPU 故障与 RMA 实战：大集群的硬件运维日常
- October 10 - 万卡集群：自建、租用还是混合？三种路线的账本
- October 10 - Google TPU Ironwood（v7）深度解析：96GB→192GB 的推理怪兽
- October 10 - 谷歌第八代 TPU 拆分战略：TPU 8t 主攻训练、TPU 8i 专精推理
- October 10 - GB300 Grace Blackwell Ultra 全解：从 20 PFLOPS 桌面超算到数据中心旗舰
- October 10 - GB200 NVL72 架构拆解：72 卡为何是一个"GPU"
- October 10 - FP4、FP8、HiF8：低精度格式如何把 AI 算力翻倍
- October 10 - 算力口径扫盲：稠密与稀疏 TFLOPS 差多少、厂商怎么"注水"
- October 10 - 出口管制下的供应链重构：2026 年的版图变化
- October 10 - Etched Sohu：All-in Transformer 的专用 ASIC 押注
- October 10 - EPYC Venice 256核登陆 2nm：AMD 服务器 CPU 的 AI 时代打法
- October 10 - 燧原 S60 与产品线复盘：云燧系列的训练推理布局
- October 10 - 边缘 AI 芯片全景：从 DGX Spark 到具身智能的算力下沉
- October 10 - 边缘 AI 盒子选型：从 NPU 小主机到工控级方案
- October 10 - 国产 AI 芯片替代采购决策框架：昇腾、寒武纪、摩尔线程、昆仑芯怎么选
- October 10 - 国产化替代迁移清单：从 CUDA 迁到昇腾/寒武纪/海光要做什么
- October 10 - 国产芯片跑 DeepSeek 实测盘点：各家适配进展与性能口碑
- October 10 - 国产 AI 芯片的订单积压：交付周期与产能爬坡
- October 10 - 国产 AI 芯片资本化全景：从寒武纪到燧原，IPO 窗口与估值逻辑
- October 10 - 桌面 AI 实验室：DGX Spark 与 DGX Station 的使用场景拆解
- October 10 - DeepSpeed、FSDP 到 Megatron：大模型分布式训练框架演进
- October 10 - CUDA 与驱动版本兼容矩阵：部署工程师的避坑手册
- October 10 - CoWoS 之困：先进封装如何成为 AI 芯片最隐蔽的瓶颈
- October 10 - 算力资产金融化：芯片回租、GPU 抵押与"卡本位"融资
- October 10 - 存算一体芯片新势力：Fractile、Untether AI 等技术路线
- October 10 - xAI Colossus 2 与全球十万卡级训练集群竞赛
- October 10 - 集合通信科普：AllReduce 如何吃掉你的训练时间
- October 10 - 冷板 vs 浸没式液冷：一张改造成本对照表
- October 10 - 芯粒与 UCIe：封装内互连的"USB-C"时刻
- October 10 - 2026 国产 GPU 一级市场融资热图：钱都流向了哪里
- October 10 - Cerebras 晶圆级引擎架构解析：把整个晶圆做成一颗芯片
- October 10 - Cerebras 晶圆级引擎：号称30倍于GPU的速度与100亿美元 OpenAI 电力协议
- October 10 - 寒武纪思元 590 深度解析：承前启后的关键一代
- October 10 - 博通与 Marvell：AI 定制 ASIC 双雄的订单之争
- October 10 - 壁仞 BR100 的启示：一度刷新纪录的国产 GPU 为何曲折
- October 10 - B300 与 GB300 有何不同：Blackwell Ultra 的两种打开方式
- October 10 - Atlas 300I Duo：国产推理部署最广的卡之一
- October 10 - 昇腾960 超节点：全球首个 NPO 超节点的 4096 卡野心
- October 10 - 昇腾950 双轨战略：950PR 主攻推理 Prefill、950DT 押注训练超节点
- October 10 - 昇腾 310B：被忽视的边缘推理基石
- October 10 - 百万卡级 TPU 协议：Anthropic 与谷歌重新定义 AI 算力合作规模
- October 10 - MI355X 到 MI455X：AMD CDNA 两代跳跃解析
- October 10 - AMD "Gorgon Halo" 来袭：Ryzen AI Max 400 系列发布，192GB统一内存撞上15年最高DRAM价格
- October 10 - AMD 数据中心业务的增长曲线：从追赶者到第二供应商
- October 10 - 买下未来：AMD 收购 MEXT、Taalas、World Labs 的算力版图拼图
- October 10 - AWS Trainium3 架构解析：Project Rainier 之后的性能跃升
- October 10 - AI PC 与桌面超算选型：RTX Spark、Gorgon Halo、DGX Station、Mac 怎么选
- October 10 - 5000亿美元涌向 AI 财团：这轮融资狂潮与2000年有何不同
- October 10 - AI 芯片周报（2026年第42周）：Gorgon Halo、千亿投资 OpenAI 与算力金融化
- October 10 - AI 芯片测试全流程：从晶圆 CP 到系统级老化
- October 10 - Agentic AI 正在重塑基础设施：从 token 到沙盒的新算力经济学
- October 10 - 800V 高压直流：AI 数据中心的电力架构革命
- October 10 - 2027 AI 芯片前瞻：Rubin Ultra、Helios II、昇腾970 与真武量产元年
- October 13 - OpenAI 算力版图最新进展：26 吉瓦采购协议、自研 Jalapeño 落地与百亿级融资
- October 13 - NVLink Fusion 生态 2026 进展：开放标准能否撼动 NVLink
- October 13 - 英伟达最新财报拆解：数据中心占比超九成，客户集中度正在悄悄变化
- October 13 - 摩尔线程 MTT S5000 出货与生态最新进展
- October 13 - 昆仑芯 M100 部署进展与百度自研芯片路线
- October 13 - 每百万 token 成本实测框架：2026 年 Q4 价格环境下的自建与 API 选型
- October 13 - 四大云厂 2026 年资本开支与算力采购指引汇总：Q4 该盯什么
- October 13 - HBM4E/HBM5 供应路线图与 2027 产能格局
- October 13 - 2026 年四季度企业 GPU 采购实操手册：缺货周期、锁价策略与预算重构
- October 13 - FP8 训练从尝鲜到主流：2026 年的真实成熟度
- October 13 - 国产卡 vs 进口卡：2026 年四季度选型决策树
- October 13 - 国产 AI 芯片 2026 年 10 月融资与上市动态：配售潮背后的量产竞速
- October 13 - 寒武纪 MLU690 规模部署与订单交付最新进展
- October 13 - 定制 ASIC 与通用 GPU：2026 年的分工与边界
- October 13 - 昇腾 960 超节点进展：4096 卡 NPO Hi-ONE 的现实与挑战
- October 13 - 昇腾 950 系列最新报价：950PR 与 950DT 的价格曲线
- October 13 - 2026 年 AI 创业公司的算力策略：从租到自建的时机
- October 13 - AI 算力总拥有成本 2026 版：电力、折旧、运维的全口径拆解
- October 13 - 四季度 AI 芯片供应与 HBM 涨价传导：真正的瓶颈已经不是晶圆
- October 13 - 国产 AI 芯片 vs 国际 AI 芯片：2026 年终全面对比
- October 13 - Agent 智能体负载下的推理调度：2026 的新挑战