Samsung plans to combine digital twins, AI, and robotics throughout its manufacturing infrastructure, beginning with a brand new ‘Megafactory’.
For COOs and CIOs in industrial sectors, the good manufacturing unit announcement – inbuilt collaboration with NVIDIA – is an instance of the shift from contained AI pilots into manufacturing. The purpose is a completely clever and predictive setting, beginning with the complicated calls for of semiconductor, cell system, and robotics manufacturing.
Samsung will deploy greater than 50,000 NVIDIA GPUs to embed AI all through the manufacturing move. This goes additional than typical automation. The manufacturing unit will use a single clever community that can let AI repeatedly analyse, predict, and optimise manufacturing environments in real-time.
Digital twins for bodily positive factors
Operationally, a key element is the widespread use of digital twins, powered by NVIDIA Omniverse libraries.
Samsung is constructing digital twins able to visualising total fab operations nearly. The enterprise software is to make use of these digital environments to establish anomalies, carry out predictive upkeep, and optimise manufacturing earlier than adjustments are utilized within the bodily world. This strategy targets a discount in downtime and permits testing course of enhancements with out risking bodily line disruption.
In a high-value use case, Samsung detailed effectivity positive factors in its computational lithography course of. Through the use of NVIDIA cuLitho and CUDA-X libraries for its optical proximity correction (OPC) course of, the corporate achieved a 20x acquire in computational lithography efficiency. As OPC is a key step in correct wafer patterning, the enhancement permits AI to foretell and proper circuit sample variations with far higher pace and precision, thereby decreasing growth cycles.
This ‘AI Manufacturing facility’ idea displays a wider business push. Many enterprises are actually trying to consolidate AI growth, leveraging platforms like Google Vertex AI or IBM watsonx to handle fashions and information. Samsung’s strategy, constructed on a 25-year collaboration with NVIDIA, is a hardware-centric mannequin for attaining this integration at scale.
Samsung places AI on the core of a full manufacturing ecosystem strategy
The implementation extends to next-generation {hardware} and robotics. The businesses are working collectively on HBM4, with Samsung’s design utilizing Sixth-generation 10-nanometer-class DRAM and a 4nm logic base die. Samsung states its HBM4 processing speeds can attain 11Gbps, far exceeding the 8Gbps JEDEC customary. This superior reminiscence is meant to type a basis for the AI-driven manufacturing infrastructure.
On the manufacturing unit ground, the plan entails bodily automation. Samsung is utilizing the NVIDIA RTX PRO 6000 Blackwell Server Version platform to advance manufacturing automation and humanoid robotics. Additionally it is utilizing the NVIDIA Jetson Thor robotic platform to speed up real-time AI reasoning, activity execution, and security controls in clever robots.
This isn’t a single-site venture, both. Samsung is planning to increase its AI manufacturing unit infrastructure to its international manufacturing hubs to convey higher intelligence and agility to its worldwide semiconductor operations.
Past the manufacturing unit partitions, the collaboration is tackling the community infrastructure required to assist widespread bodily AI. The companions are working with Korean telecom operators and researchers on AI-RAN growth.
AI-RAN is a next-generation communication expertise that integrates AI computing energy into cell community capabilities. For a CTO, this issues as a result of it permits robots, drones, and industrial automation tools to course of information and run inferences in real-time on the community edge. This AI-powered cell community is positioned as a neural community that’s important for widespread adoption of bodily AI.
For enterprise decisionmakers, Samsung’s blueprint reveals that an efficient ‘AI manufacturing unit’ should be an end-to-end system. It requires unifying {hardware} (like HBM4), proprietary AI fashions, operational expertise (robotics), and edge networking (AI-RAN) right into a single and predictive information move.
See additionally: ASUS IoT platform makes use of NVIDIA tech for edge AI robotics
Wish to be taught extra about IoT from business leaders? Try IoT Tech Expo happening in Amsterdam, California, and London. The excellent occasion is a part of TechEx and is co-located with different main expertise occasions together with AI & Massive Knowledge Expo. Click on right here for extra data.
IoT Information is powered by TechForge Media. Discover different upcoming enterprise expertise occasions and webinars right here.