Naver and Samsung Electronics to Develop 'Hyperscale AI' Specialized Semiconductor

Launch of AI Semiconductor Solution Development TF
Bottleneck Issues in Large-scale Data Computation for Massive AI
Development of Next-generation Memory Solutions Combining Memory Semiconductors and AI Processors

[Asia Economy Reporter Yuri Choi] Naver and Samsung Electronics have joined forces to secure a leading position in the 'hyper-large AI' market, which is regarded as the future of the artificial intelligence (AI) industry. Samsung Electronics' next-generation semiconductor technology will be introduced to solve the technical challenge of system bottlenecks in Naver's hyper-large AI 'HyperCLOVA' and to develop optimized semiconductors.

Naver and Samsung Electronics Launch Hyper-Large AI Task Force

On the 6th, Naver announced that it signed a memorandum of understanding (MOU) with Samsung Electronics to develop AI semiconductor solutions and launched a practical task force (TF). Following their collaboration in June on building a 5G specialized network at Naver’s second office building 1784, the two companies have agreed to jointly develop next-generation semiconductor solutions.

Hyper-large AI performs complex computations using massive amounts of data. This process requires system lightweighting technology and next-generation memory technology. Even if the processor completes calculations at high speed, a slow memory speed causes system bottlenecks.

Besides resolving bottlenecks, the two companies will jointly develop memory with AI processors combined, known as high-bandwidth processing memory (HBM-PIM). Using HBM-PIM improves speed and computational efficiency compared to configuring processors and memory separately. This enables hyper-large AI to perform more computations at faster speeds, playing a key role in building next-generation AI systems.

Additionally, Samsung Electronics will develop Computer Express Link (CXL), a next-generation memory solution technology optimized for hyper-large AI. CXL is one of the interface standards connecting processors, memory, and storage devices in computers. It allows multiple devices and memory to be connected without speed degradation, significantly enhancing computational performance when applied to hyper-large AI.

Expecting Synergy from Naver’s Software and Samsung Electronics’ Hardware Competitiveness

The reason the two companies have partnered is that developing AI semiconductor solutions requires integrated capabilities including hyper-large AI technology, software, hardware, and service operation.

Naver possesses the software capabilities necessary for AI semiconductor development. Based on its self-built supercomputer infrastructure, it developed the hyper-large AI 'HyperCLOVA.' Naver has operational know-how from applying it to various services such as search query correction and shopping review summarization. It also holds lightweight algorithms based on graphics processing units (GPU) that deliver 2 to 3 times faster performance than previous models. Samsung Electronics has technological competitiveness not only in semiconductor development and manufacturing but also in memory-based solutions. It has developed numerous solutions to resolve system bottlenecks in hyper-large AI.

Jung Seok-geun, CEO of Naver CLOVA CIC, said, "By combining the knowledge and know-how Naver has secured through servicing HyperCLOVA with Samsung Electronics’ advanced semiconductor manufacturing technology, we will be able to create unprecedented new solutions that solve the challenges faced by the latest AI technologies. Starting with this collaboration, we will expand the scope of technology and strengthen domestic AI technological competitiveness."

Han Jin-man, Vice President and Head of Strategic Marketing at Samsung Electronics Memory Business Division, stated, "Through semiconductor solutions reflecting the needs of AI service companies and users, we will expand our next-generation memory lineup leading the market, including PIM and computational storage."