KT (Acting Representative Park Jong-wook) announced on the twenty ninth that it had signed an MOU with kt Cloud and Samsung Electronics for ‘Technical Cooperation in Next-Generation Memory to Realize a Korean-Style AI Full-Stack’.

KT Vice President Song Jae-ho, KT Cloud President Yoon Dong-sik, Samsung Electronics Vice President Kim Jae-jun, and Samsung Electronics Vice President Shim Eun-soo attended the signing ceremony held at kt cloud headquarters in Gangnam-gu, Seoul on the afternoon of the twenty eighth.

Through this agreement, the three corporations ▲analyze the memory impact of the super-giant AI model ▲AI semiconductor through KT’s super-giant artificial intelligence (AI) “belief” in Samsung Electronics’ PIM (Processing in Memory) and PNM (Processing Near Memory) environments They plan to cooperate to strengthen AI full-stack capabilities reminiscent of platform research and next-generation cloud solutions.

Korean AI full stack refers to services that cover every thing from infrastructure reminiscent of domestic AI semiconductors, software, and cloud to AI application services.

PIM (Processing-in-Memory) is an intelligent semiconductor that adds a processor function that performs computation to a memory semiconductor that performs storage. It has the advantage of reducing the number of information movements and reducing bottlenecks because easy operations could be performed in memory. PNM (Processing-near-Memory) is a technology that minimizes data transfer time with the CPU by arranging calculation functions near memory.

Within the means of developing AI services, when using existing AI infrastructure, problems reminiscent of low efficiency in computation speed and power consumption have arisen. To resolve this problem, KT and kt Cloud commercialized an AI semiconductor Neural Processing Unit (NPU) optimized for AI operation in May last yr at kt Cloud along with Riot, a semiconductor design company.

Moreover, KT and kt Cloud will use KT’s super-giant AI model belief to finish Korean-style AI full-stack and begin empirical research and technical cooperation on Samsung Electronics’ PIMž and PNM infrastructure. It plans to introduce Samsung Electronics’ smart infrastructure technology to unravel the ‘memory bottleneck’ wherein answers are cut off when using AI services.

Vice President Song Jae-ho of KT’s AI/DX Convergence Business Division said, “We are going to construct a competitive infrastructure foundation through cooperation with Samsung Electronics to develop next-generation memory technology for the event of super-large AI.” It’s going to expand and contribute to strengthening Korea’s artificial intelligence competitiveness.”

Yoon Dong-sik, president of kt Cloud, said, “We are going to proceed to advance AI semiconductors and complete a Korean-style AI full-stack through joint cooperation with super-large AI infrastructure reminiscent of HAC and NPU, Samsung Electronics’ next-generation semiconductor technology, and KT’s AI solutions and technologies.”

“We are going to solve the memory bottleneck of super-large AI models that require large-scale data processing through technological cooperation,” said Kim Jae-joon, vp of strategic marketing at Samsung Electronics’ memory division. “Smart technologies reminiscent of PIM and PNM reflect the market needs of cloud and AI corporations. Through this, we’ll expand next-generation memory technology.”

Reporter Juyoung Lee juyoung09@aitimes.com