Moore Threads / 摩尔线程

Moore Threads / 摩尔线程 is named in 国产 AI 算力能凭「超节点」弯道超车吗？｜WAIC 深度观察 S10E23 as one of the Chinese AI-chip vendors with its own [[ScaleUpAIInterconnect|Scale Up]] approach. The source does not benchmark Moore Threads directly; it uses the company to illustrate Proprietary AI Interconnect Fragmentation inside the domestic supernode market.

The page should be read as a source-scoped marker. Moore Threads matters here because cloud customers and large model teams may not want to learn too many separate software ecosystems, so vendors need strong proof beyond launch announcements.

177: 详解Kimi K3：强到冲击Anthropic估值的模型什么样？ adds a more concrete adaptation signal. Zhao Chenyang / 赵晨阳 says Moore Threads quickly supported [[KimiK3|Kimi K3]] through a MUSA/SGLang-style ecosystem and uses the example to argue that Kernel Development Agents can help alternative accelerator vendors adapt faster, though not necessarily overturn Nvidia’s lead.

Connections

• Domestic AI Chip Catch-Up and Domestic AI Chip Order Validation — broader substitution and customer-order test.
• Biren Technology / 壁仞科技, MetaX / 沐曦, Cambricon / 寒武纪, and Kunlunxin / 昆仑芯 — domestic AI-chip peer set.
• CUDA and AI Infrastructure Full-Stack Moat — incumbent ecosystem comparison.
• Kimi K3, Kernel Development Agents, and Model-Infra Co-Design — K3 adaptation and kernel-agent branch added by LateTalk episode 177.