Techday
This page was machine-translated and may differ from the original. View original

National AI Data Center Begins Full-Scale Support for Building AI Ecosystem

Google 우선 소스Published2023.10.16 14:44

▲ AI Semiconductor Innovation Summit
Supporting domestic AI semiconductor references through a demonstration server farm
Floating operations are important; Rebellion Atom chip demonstration introduced.
Furiosa and Sapion scheduled for release in 2024, expanding the ecosystem

Faced with the era of global competition in artificial intelligence, the government is proactively responding to enhance industrial competitiveness. It is focusing on strengthening the self-reliance of the domestic industry by constructing a national AI data center and establishing a system for the demonstration and verification of AI semiconductors.

The AI Industry Convergence Group held the AI Semiconductor Innovation Summit at COEX in Samseong-dong, Seoul on the 13th. The event featured presentations on the nation's first AI semiconductor testing and verification system, demonstration plans for the public sector, and utilization strategies for AI data centers.

The Artificial Intelligence Industry Convergence Business Group, established to build an AI ecosystem, is equipped with 77 types of demonstration equipment and a driving center capable of simulation and updating. Operating AI startups and various support programs, the group is centered in Gwangju and supports the creation of an AI ecosystem.

Center Director Oh Byung-doo emphasized, “The AI Business Group is preparing for nationwide expansion starting next year, and services are currently available for application and use in all regions,” adding that “the core of the National AI Data Center is AI semiconductors.”

The growth of fabless companies is paramount for the development of the domestic system semiconductor ecosystem. In particular, for startup-level fabless companies to achieve domestic self-sufficiency and enter the global market, not only is chip design and development essential, but references where the chip is directly applied to projects and clients are also required.

Introducing the project to establish an AI semiconductor testing and verification system, Researcher Kim Ji-woong stated, "We expect to lay the foundation for commercialization through such testing and verification." While domestic fabless companies manufacture semiconductors, it is not easy for their products to be selected for reference projects.


▲ The AI Industry Convergence Business Group booth at the 2023 AIoT International Exhibition


Reflecting these industry demands, the reliability of product utilization can be verified through basic verification and demonstration of AI semiconductors. It is significant in that it serves as a foundation for domestic fabless companies to enter the global market and move to the next stage.

Through AI semiconductor server farms and public sector demonstration services, KT Cloud, NHN Cloud, and Naver Cloud formed a consortium and collaborated with domestic fabless companies developing AI semiconductors and AI service companies to build a 19.95 PF AI semiconductor farm.

We have commercialized a package consisting of NPU, server farm, and platform SW, and are currently conducting demonstrations of five public AI services in the fields of health, control, education, and national defense.

Specifically, KT Cloud is collaborating with Rebellion and Superb AI to build a 4.45 PF server farm to demonstrate food monitoring services in the healthcare sector, while NHN Cloud is collaborating with Sapion Korea, Human ICT, and IVS to build an 11 PF server farm to demonstrate fire prevention control and intrusion identification in the defense sector.

Naver Cloud is collaborating with Furiosa AI, Elice, and Simple Platform to build a 4.5 PF server farm and conduct demonstrations of AI learning mentoring in the fields of industrial accident prevention monitoring and education.

Currently, it is known that only the KT Cloud server farm built with Rebellion's Atom chip is capable of floating computation. Furiosa AI is reportedly planning to develop and release a chip capable of floating computation in 2024. Sapion Korea also plans to introduce the X330 chip in 2024 to enable floating computation.

With no global standards for AI semiconductors in place, the domestic consortium's proactive establishment of a standard and testing verification system is expected to enhance global competitiveness and expand the domestic AI ecosystem.
본 기사에 대한 정정·반론·추후보도 청구는 보도 청구 안내를, 그간 게재된 보도문은 정정·반론보도 모아보기를 참고해 주세요.
명세환 기자
명세환 기자