Core Insights - South Korea is accelerating its AI research projects, but is facing increasing electricity supply pressures [1] - The competition in the AI field is intensifying, leading to power supply constraints in universities [1] - The country is focusing on renewable energy sources like solar and wind to meet the growing electricity demand [1] Group 1: AI Research and Electricity Demand - AI research consumes a significant amount of electricity, causing universities to temporarily shut down other lab servers to meet AI project needs [1] - NVIDIA announced it will supply 260,000 GPUs to South Korea, but the required power to operate these GPUs, including cooling systems, is approximately 600 megawatts [1] - Due to power supply limitations, there are concerns about the ability to utilize these GPUs effectively [1] Group 2: Renewable Energy and Cost Challenges - South Korea's President emphasized the urgent need for substantial electricity to support data centers, stating that renewable energy is the only system that can quickly meet this demand [1] - Analysts point out that the current cost of renewable energy generation in South Korea is relatively high, suggesting a focus on other power generation methods like natural gas [1] - Bloomberg New Energy Finance estimates that the levelized cost of electricity (LCOE) for solar energy in South Korea will reach $98 per megawatt-hour in 2024, significantly higher than the global average of $35 [1] Group 3: Transmission Infrastructure Issues - The electricity supply pressure is evident not only in generation but also in transmission, with bottlenecks in the power grid construction [2] - A survey indicated that out of 54 transmission projects listed in South Korea's 11th Electricity Supply and Demand Basic Plan, 30 projects have been delayed or are expected to be delayed [2] - For instance, a transmission line connecting several power plants on the east coast to the capital region, originally scheduled for completion in 2019, is now postponed to the earliest by the end of 2026 [2]
韩国AI研究面临电力“卡脖子”
Xin Hua She·2025-11-17 07:24