Samsung Electronics and Naver recently released a jointly developed artificial intelligence (AI) semiconductor, which took a year to develop. The product is known for being about eight times more energy efficient than competing chips such as NVIDIA, and is expected to be used to provide computing power for Naver's ultra-large-scale artificial intelligence model HyperCLOVAX.
South Korea's Ministry of Science, Information and Communications announced that the "4th Artificial Intelligence Semiconductor High-Level Strategic Dialogue" was held at a hotel in Seocho District, Seoul on December 19 to showcase the results of domestic artificial intelligence semiconductor companies.
The artificial intelligence semiconductor developed by Naver and Samsung Electronics has been unveiled in the form of field programmable gate array (FPGA). FPGA is a semiconductor that allows developers to modify designs and is mainly used for prototyping before mass production. This semiconductor is used specifically for "inference," the process of using new data to produce logical results after an AI model has completed its "learning" phase.
Naver said the AI semiconductor is eight times more energy efficient than competing products from companies such as NVIDIA. This is achieved by integrating low-power, compact double data rate (LPDDR) DRAM to improve efficiency. Greater energy efficiency means the same computing performance can be achieved with less power, allowing large-scale computing tasks to be performed at a lower cost.
Previously, the two companies had signed a memorandum of understanding (MOU) on the development of artificial intelligence semiconductor solutions in December last year and launched a practical working group. The completed AI semiconductor is expected to be used to run Naver’s HyperCLOVAX.
Meanwhile, the Ministry of Science and Information and Communications Technology introduced on the same day the progress of the first phase of the "K-Cloud Project", a project to develop ultra-high-speed, low-power domestic artificial intelligence semiconductors and apply them to data centers.
The government plans to upgrade domestic AI semiconductors in three stages by 2030. The first phase involves applying domestic neural processing units (NPUs) to data centers and providing cloud-based artificial intelligence services.
Companies participating in the first phase, including NHNCloud, SapeonKorea and NOTA, demonstrated their "intelligent traffic control service" that can identify vehicles and pedestrians in crosswalk CCTV footage. KTCloud, Rebellion and SuperbAI demonstrated solutions for calculating nutrient intake, while NaverCloud, FuriosaAI and SimplePlatform demonstrated technology for identifying workers and measuring risk levels using CCTV footage from construction sites.