媒体报道称 NVIDIA 与富士康合作，把其全球 200 多个工厂改造为自学习智能工厂。富士康公布初步成果：组装线缺陷率降低超 90%，注塑换模工作量减少 50%，通过自动排产减少 25% 工时。其 1200 亿参数大模型 FoxBrain 用 NVIDIA DGX SuperPods 训练速度较 H100 提升 2.48 倍，12 天处理 1 万亿 token。
AI 生成摘要 · 以原文为准
Nvidia has been working with key supplier Foxconn—the world’s biggest AI server and iPhone assembler—to turn its 200+ global manufacturing plants into self-learning smart factories using AI, media report. Foxconn has revealed major initial results:
-Slashed assembly-line defect rates by over 90%
-Cut injection molding changeover work by 50%
-Reduced labor hours by 25% via automated production scheduling
Foxconn also trained its 120-billion parameter LLM, FoxBrain, 2.48x faster using Nvidia DGX SuperPods (vs. H100), processing 1 trillion tokens in 12 days. Across its facilities, Foxconn uses Nvidia Omniverse for digital twin factory planning, Isaac for production robotics, alongside Mission Control, NIM microservices, NVIDIA Run, more. $NVDA $AAPL #Foxconn #FoxBrain #DGXSuperPod https://money.udn.com/money/story/5612/9838449?from=edn_subcatelist_cate
来源：Dan Nystedt · x.com