Samsung, Naver Partner on Hyperscale AI Semiconductors

The South Korean companies signed an MoU to collaborate on next-gen AI solutions.

Samsung Electronics and Korean Internet search provider Naver have struck a partnership to jointly develop chips that can accommodate large artificial intelligence (AI) workloads. On Monday the companies signed a memorandum of understanding that will entail expanding Naver’s hyperscale technology to the semiconductor industry. The efforts involve forming a working-level task force to develop semiconductor solutions for AI.

AI is advancing rapidly and larger, more complex models require large amounts of data to be processed quickly, pushing the limits of standard chips. As a result, semiconductor companies are doubling down on efforts to design chips that can handle escalating AI workloads. Improvements to chip architectures will help speed up AI data processing as well as increase power and memory storage systems.

With the global AI market expected to swell to $1.8 trillion by 2030, according to Grand View Research, semiconductor companies that develop AI-capable chips are positioned to benefit. That’s why Naver and Samsung are looking to push Korea’s AI prowess to the next level through better semiconductor solutions.

The partnership will require the convergence of innovation in hyperscale AI, software, services, and hardware. With both companies providing their own expertise in the given areas, the goal is to develop new semiconductor solutions that can address bottlenecks in current AI systems by maximizing efficiency in power consumption and considering the needs of AI end users starting from the initial stages of development while enabling optimization at the system level.

“Combining our acquired knowledge and know-how from HyperCLOVA [Naver’s AI language model system] with Samsung’s semiconductor manufacturing prowess, we believe we can create an entirely new class of solutions that can better tackle the challenges of today’s AI technologies,” said Naver’s Suk Geun Chung in a news release.

Naver will supply software expertise to help develop semiconductor solutions for AI with its hyperscale AI based on its supercomputer infrastructure and services that it makes available to over 36 million users daily. Additionally, the company will provide expertise in compression algorithms with performance reportedly two to three times as fast as existing models currently running on GPUs.

Samsung, a leading global player in the semiconductor arena, develops and manufactures chips and memory technologies such as computation storage, processing-in-memory (PIM), processing-near-memory (PNM), and Compute Express Link (CXL)—all helpful in addressing bottlenecks in large-scale AI systems.

The first phase of the partnership will involve working on the development and technological verification of a compression solution critical to the application and expansion of hyperscale AI models. The two companies also plan to continue their collaboration on next-generation memory solutions that support HPC, such as SmartSSD, HBM-PIM and CXL.

Given the U.S. federal government’s restrictions on advanced semiconductor exports to China, a policy intended to curb Beijing’s development of AI systems for possible militaristic purposes, Samsung and Naver may have to vet buyers of their hyperscale AI semiconductor solutions or risk harming trade and diplomatic ties.