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 sensible manufacturing facility announcement – inbuilt collaboration with NVIDIA – is an instance of the shift from contained AI pilots into manufacturing. The objective is a totally clever and predictive surroundings, beginning with the complicated calls for of semiconductor, cell gadget, and robotics manufacturing.
Samsung will deploy greater than 50,000 NVIDIA GPUs to embed AI all through the manufacturing circulate. This goes additional than typical automation. The manufacturing facility will use a single clever community that may let AI repeatedly analyse, predict, and optimise manufacturing environments in real-time.
Digital twins for bodily features
Operationally, a key part is the widespread use of digital twins, powered by NVIDIA Omniverse libraries.
Samsung is constructing digital twins able to visualising complete fab operations just about. The enterprise utility 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 method 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 features 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 velocity and precision, thereby lowering growth cycles.
This ‘AI Manufacturing facility’ idea displays a wider trade push. Many enterprises are actually trying to consolidate AI growth, leveraging platforms like Google Vertex AI or IBM watsonx to handle fashions and knowledge. Samsung’s method, constructed on a 25-year collaboration with NVIDIA, is a hardware-centric mannequin for reaching this integration at scale.
Samsung places AI on the core of a full manufacturing ecosystem method
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 commonplace. This superior reminiscence is meant to kind a basis for the AI-driven manufacturing infrastructure.
On the manufacturing facility ground, the plan includes bodily automation. Samsung is utilizing the NVIDIA RTX PRO 6000 Blackwell Server Version platform to advance manufacturing automation and humanoid robotics. It’s also utilizing the NVIDIA Jetson Thor robotic platform to speed up real-time AI reasoning, job execution, and security controls in clever robots.
This isn’t a single-site mission, both. Samsung is planning to increase its AI manufacturing facility infrastructure to its world manufacturing hubs to deliver higher intelligence and agility to its worldwide semiconductor operations.
Past the manufacturing facility partitions, the collaboration is tackling the community infrastructure required to help 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 knowledge 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 exhibits that an efficient ‘AI manufacturing facility’ have to 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 knowledge circulate.
See additionally: ASUS IoT platform makes use of NVIDIA tech for edge AI robotics
Need to study extra about IoT from trade leaders? Try IoT Tech Expo going down 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 Information 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.
