Copyright Wccftech
Samsung and NVIDIA are now significantly enhancing their 25-year collaborative relationship by undertaking a project to build a full-fledged AI factory, equipped with over 50,000 NVIDIA GPUs. In addition, the two behemoths continue to collaborate with each other in diverse spheres, including HBM4, EDA tools, and AI-RAN. Samsung to deploy 50,000 NVIDIA GPUs within its AI Megafactory Samsung has partnered with NVIDIA to deploy 50,000 of NVIDIA's GPUs: "By deploying more than 50,000 NVIDIA GPU’s, AI will be embedded throughout Samsung's entire manufacturing flow, accelerating development and production of next-generation semiconductors, mobile devices and robotics." Samsung intends to use this AI megafactory to integrate its labyrinthine manufacturing processes into "a single intelligent network, where AI continuously analyzes, predicts, and optimizes production environments in real time." To scale its AI factory in the years to come, Samsung also plans to leverage NVIDIA accelerated computing and implement digital twin manufacturing with the help of NVIDIA Omniverse libraries. For those who may not be aware, NVIDIA Omniverse is a simulation platform for building and operating real-time 3D applications, services, and virtual worlds. And, digital twin manufacturing allows for the creation of a virtual replica of a physical manufacturing process, enabling companies to monitor and analyze the entire process in real-time. Samsung plans to build digital twins to digitally visualize its fab operations, using these "virtual environments to identify anomalies, perform predictive maintenance, and optimize production before changes are applied in the physical world." Eventually, Samsung plans to expand its AI factory infrastructure to its global manufacturing hubs, including its US base in Taylor. Other ways in which Samsung and NVIDIA are collaborating with each other: Both companies are collaborating on HBM4, which leverages Samsung's 6th-gen, 10nm-class DRAM and a 4nm logic base die, reaching speeds of up to 11Gbps. Samsung has achieved a 20x performance gain in its computational lithography by leveraging NVIDIA cuLitho and CUDA-X libraries for its optical proximity correction (OPC) process, which uses AI to predict and correct circuit pattern variations with speed and precision. Samsung and NVIDIA are also collaborating with other Electronic Design Automation (EDA) partners to develop next-gen GPU-accelerated EDA tools and design technologies. As a refresher, EDA is used to automate the design, verification, and manufacturing preparation of ICs and microprocessors. Samsung is building its bespoke AI models by leveraging NVIDIA accelerated computing and the Megatron framework for LLMs. The two companies are also collaborating with each other in the sphere of intelligent robotics via the NVIDIA RTX PRO 6000 Blackwell Server Edition platform and the NVIDIA Jetson Thor robotic platform. Finally, Samsung is working with NVIDIA, Korean telecom operators, the academia, and research institutions to foster collaboration on AI-RAN, a next-generation communication technology that integrates AI into cellular radio access networks to improve their performance, efficiency, and capabilities.
