KT, KT Cloud, and Samsung Electronics Join Forces for 'Korean-style AI Full Stack'

KT announced on the 29th that it has signed a Memorandum of Understanding (MOU) for ‘Next-Generation Memory Technology Cooperation to Implement a Korean-Style AI Full Stack’ together with KT Cloud and Samsung Electronics.

The three-party agreement ceremony held the previous day at the KT Cloud headquarters located in Gangnam-gu, Seoul, was attended by key officials including Song Jae-ho, Vice President of KT, Yoon Dong-sik, President of KT Cloud, Kim Jae-jun, Vice President of Samsung Electronics, and Shim Eun-soo, Vice President of Samsung Electronics.

(From left) Song Jae-ho, Vice President of KT; Kim Jae-jun, Vice President of Samsung Electronics; Yoon Dong-sik, President of KT Cloud; and Shim Eun-soo, Vice President of Samsung Electronics, are taking a commemorative photo after signing the MOU. [Photo by KT]

Through this agreement, the three companies plan to collaborate to strengthen AI full-stack capabilities using KT’s super-large-scale AI ‘Mideum’ in Samsung Electronics’ PIM and PNM environments, focusing on ▲memory impact analysis of super-large-scale AI models, ▲AI semiconductor platform research, and ▲next-generation cloud solutions.

The Korean-style AI full stack refers to products and services encompassing domestic AI semiconductors, software, cloud infrastructure, and AI application services. PIM (Processing-in-Memory) is an intelligent semiconductor that adds processor functions performing computations to memory semiconductors that handle storage tasks. It can perform simple computations within the memory, reducing data transfer frequency and alleviating bottlenecks. PNM (Processing-near-Memory) is a technology that places computing functions near the memory to minimize data transfer time with the CPU.

When developing AI services using existing AI infrastructure, inefficiencies occurred in computation speed and power consumption. To address this, KT and KT Cloud commercialized the NPU (Neural Processing Unit), an AI semiconductor optimized for AI computation, in collaboration with semiconductor design company Rebellion at KT Cloud in May.

Furthermore, KT and KT Cloud will utilize KT’s super-large-scale AI model ‘Mideum’ to conduct demonstration research and technical cooperation on Samsung Electronics’ PIM and PNM infrastructure to complete the Korean-style AI full stack. To resolve the ‘memory bottleneck’ issue, such as interrupted responses when using AI services, they plan to enhance the Korean-style AI full stack by adopting Samsung Electronics’ smart infrastructure technology.

Song Jae-ho, Head of KT AI/DX Convergence Business Division (Vice President), said, “Through next-generation memory technology development cooperation with Samsung Electronics for super-large-scale AI development, we will build a competitive infrastructure base,” adding, “With the completed AI full stack, we will expand the AI ecosystem across industries and contribute to strengthening South Korea’s AI competitiveness.”

Yoon Dong-sik, President of KT Cloud, stated, “Through joint cooperation combining KT Cloud’s super-large-scale AI infrastructure, Samsung Electronics’ next-generation semiconductor technology, and KT’s AI solutions and technology, we will continuously advance AI semiconductor sophistication and complete the Korean-style AI full stack.”

Kim Jae-jun, Head of Strategy Marketing Office, Memory Business Division at Samsung Electronics (Vice President), said, “Through technological cooperation with KT and KT Cloud, we will resolve the memory bottleneck of super-large-scale AI models that require large-scale data processing,” and added, “We will expand next-generation memory technologies through smart technologies such as PIM and PNM that reflect the market needs of cloud and AI companies.”

This content was produced with the assistance of AI translation services.