Gwangju City announced on the 25th that it will launch a consulting project next month in partnership with the Ministry of Science and ICT (MSIT) to establish the “National NPU Center.”
The city has secured a budget of 600 million KRW for this project. The results will serve as the foundational data for establishing the center in Gwangju, covering aspects such as the scale of the facility, total required budget, and strategic utilization plans. An NPU (Neural Processing Unit) is a device that functions like the human brain, learning and processing information through numerous neural-like cells. It derives optimal values by training AI products—such as autonomous vehicles, robot vacuum cleaners, intelligent cameras, and industrial logistics systems—with accumulated data from fields like medicine, law, automotive, and public safety. As the global competition for dominance in the AI industry rapidly shifts from the “training” phase (building large language models) to the “inference” phase (creating value through real-world applications), the NPU Center is regarded as a cornerstone of AI technology transition alongside GPU computing centers.
Furthermore, the Gwangju NPU Center offers significant advantages in terms of low power consumption and high efficiency compared to GPU centers. NPUs consume up to three times less power than GPUs and maximize efficiency by focusing on computational functions while eliminating unnecessary circuitry. This makes them ideal for battery-dependent devices like smartphones and laptops. Significantly, NPU chips can now be manufactured domestically, allowing for self-sufficiency without relying on imports. While the GPU market is dominated by global giants like NVIDIA, South Korean fabless companies based in Gwangju—such as ASICLAND, AIM Future, FuriosaAI, and Rebellions—are accelerating the advancement and commercialization of NPU technology.
Son Du-young, Director of the Artificial Intelligence Industry Bureau, explained the synergy between the two technologies using an aviation analogy: “If we compare the center to an airplane, the NPU is like a fuel-efficient engine for long-distance cruising, while the GPU acts as a supplementary engine responsible for complex takeoffs, landings, and maneuvering.” He added, “The global trend is the harmony between NPU and GPU. He further expressed confidence, stating, “The NPU Center in Gwangju will be a vital infrastructure for the national AI strategy and a key hub to help domestic companies secure global competitiveness. With the localization of NPU chips, the combination of the National AI Data Center and the NPU Center will draw companies and researchers undergoing AI transitions to Gwangju.”
