百度 Kunlunxin M100 如期发布，M300 定档 2027
百度把自研芯片路线图直接摆上台面：M100 现在就能买来跑推理，M300 等 2027 年。做推理业务、关心算力成本的朋友可以看看，这是国内少有的能对齐英伟达的选项。
Robin 在 Baidu World 2026 上表示，随着 AI 智能体使用量增长，推理将成为 AI 芯片长期增长的主要驱动力。百度自研芯片 Kunlunxin 的 M100 已按计划发布，主打高性价比推理。路线图上，旗舰芯片 M300 预计 2027 年出货，在算力、内存和互连方面带来升级；2028 年还将推出基于新架构、面向不同应用场景的更多芯片。
这款芯片的定位是承接智能体规模化落地后的推理算力需求，而非训练场景。百度将 Kunlunxin 的迭代节奏明确到了 2027 年的 M300 和 2028 年的新架构产品，等于对外公布了三年的自研芯片路线图。
对开发者与企业用户而言，这意味着推理成本曲线将由百度自研硬件直接影响；与英伟达 GPU 方案相比，Kunlunxin 主打的是推理侧性价比，可作为国内推理算力采购中的备选路线。
"We believe inference will become the primary driver of sustained, long-term growth in AI chips," Robin said at Baidu World 2026, as the growing use of AI agents drives demand for inference.
Kunlunxin's M100 launched as planned, delivering cost-effective inference. Next on the roadmap: - The flagship M300 is expected to ship in 2027, bringing advances in compute, memory and interconnection - Additional chips built on a new architecture are expected to roll out in 2028, tailored to different application needs
This report is categorized under AITOP's "Industry" desk, focusing on China's AI industry and domestic foundation model ecosystem. Within our cross-regional analytical framework, it directly maps to US-China foundation model benchmarks and commercialization comparison. Key entities and technical topics under continuous tracking include Kunlunxin, 百度, AI芯片, 推理.
Source attribution: curated from 百度 AI Baidu (@Baidu_Inc).
