Vera Rubin NVL72 Global Supply Chain
Vera Rubin NVL72 — Global Supply Chain Network
Key contributing regions, companies, and their roles in building a single AI Factory rack
🇺🇸
United States
Design & Software
  • NVIDIA (Santa Clara) GPU/CPU/NIC chip design, CUDA software stack, platform architecture
  • Microsoft RTX Spark PC co-development, Windows AI integration
🇹🇼
Taiwan
Manufacturing Core
  • TSMC 3nm fab, CoWoS/SoIC advanced packaging
  • Foxconn / Quanta / Wistron Full-rack ODM assembly, liquid cooling integration
  • Delta Electronics 800V HVDC power conversion, solid-state transformers
  • AVC / Auras / Cooler Master Liquid cooling manifolds, cold plates, QD valves
🇰🇷
South Korea
Memory & AI Ecosystem
  • SK Hynix HBM4 (6th-gen) primary supplier, base die R&D
  • Samsung Electronics HBM4 supply partner, advanced packaging co-dev
  • Naver Cloud Sovereign AI, Korean-language LLM deployment
  • LG Electronics / Doosan Autonomous factory robotics, smart manufacturing
Critical Dependency: HBM4 + TSMC Fab Co-Production
Korea: HBM4 Stack Design
+
Taiwan: TSMC 3nm Base Die
→
TSMC CoWoS Packaging
→
Rubin GPU Module
→
Taiwan ODM Final Rack
→
AI Factory Worldwide
⚠ Both Taiwan and South Korea must cooperate technically for HBM4 production — Korea's memory design requires TSMC's advanced foundry process for the base die. Neither can produce Vera Rubin independently. Source: NVIDIA Korea Partner Night (Computex 2026); Chosun Biz analysis.