Samsung Plans AI Megafactory With 50,000 Nvidia GPUs

Samsung will build an AI Megafactory deploying more than 50,000 Nvidia GPUs to embed artificial intelligence across its semiconductor manufacturing flow for mobile devices, robotics and research while creating digital twins of fabs. The partners will use Nvidia platforms — including Omniverse, CUDA‑X and cuLitho — to simulate workflows, train internal models, enable predictive maintenance and accelerate computational lithography (reported up to 20x performance gains) to improve yields and speed. The agreement expands a decades‑long Samsung–Nvidia partnership into foundry and manufacturing services and includes cooperation on next‑generation HBM4 memory; Samsung plans to market HBM4 in 2026 and is in discussions to supply Nvidia. The announcement is part of a broader South Korean push, with SK Group and Hyundai also working with Nvidia on large GPU-backed “AI factory” initiatives and manufacturing AI clouds. Samsung has not disclosed the megafactory’s location or construction timetable. The news lifted Samsung’s shares by more than 4% following the announcement.