Samsung Electronics announced plans to establish a new AI Megafactory in partnership with NVIDIA. The initiative will embed AI throughout Samsung’s manufacturing processes, using over 50,000 NVIDIA GPUs to accelerate the development and production of semiconductors, mobile devices, and robotics.

The AI Factory will integrate all aspects of semiconductor production, including design, operations, equipment, and quality control, into a single intelligent network. AI will continuously analyze, predict, and optimize production in real time, going beyond conventional automation to interpret the data generated across manufacturing.

AI Integration in Semiconductor and Robotics Production

Samsung and NVIDIA have a collaboration that spans more than 25 years, starting with Samsung DRAM in NVIDIA’s early graphics cards and extending to foundry partnerships. They are now working together on HBM4, a high-bandwidth, energy-efficient memory solution expected to accelerate AI application development and support AI-driven manufacturing.

Samsung HBM4 is built on 6th-generation 10-nanometer DRAM and a 4nm logic base die, delivering speeds up to 11 Gbps, above the JEDEC standard of 8 Gbps. Samsung will also continue providing next-generation memory solutions, including HBM, GDDR, and SOCAMM, along with foundry services to support innovation across the AI value chain.

Scaling Intelligent Manufacturing with NVIDIA Technologies

Samsung plans to implement NVIDIA accelerated computing and Omniverse libraries to expand its AI Factory and enable digital twin manufacturing across memory, logic, foundry, and advanced packaging.

Using NVIDIA cuLitho and CUDA-X libraries for optical proximity correction (OPC), Samsung achieved a 20x increase in computational lithography performance. This improvement allows AI to predict and correct circuit pattern variations faster and more accurately, reducing development cycles. Digital twins visualize factory operations, detect anomalies, and optimize production before changes are applied physically.

Advanced AI Models and Robotics for Smart Production

Samsung integrates proprietary AI models across over 400 million devices and internal manufacturing systems to enhance intelligence and innovation in production. Built on NVIDIA accelerated computing and the Megatron framework, these models provide advanced reasoning for real-time translation, multilingual conversations, and intelligent summarization.

In robotics, Samsung uses NVIDIA RTX PRO 6000 Blackwell Servers and the Jetson Thor platform to advance automation and humanoid robotics. AI connects virtual simulations with real-world robot data, enabling robots to understand surroundings, make decisions, execute tasks, and operate safely.

Expanding AI into Networks and Industries

Samsung and NVIDIA are collaborating on AI-RAN, a next-generation network integrating AI computing into mobile networks. AI-RAN allows robots, drones, and industrial automation to sense, process data, and perform inferences in real time at the network edge.

This initiative builds on last year’s AI-RAN proof-of-concept, combining Samsung’s software-based network with NVIDIA GPUs. The companies will continue collaborating with Korean telecom operators, academia, and research institutions to strengthen AI-RAN for broader adoption of physical AI.

PHOTO: SAMSUNG SDS

This article was created with AI assistance.