New micro data center facilities will rely on locally developed AI chips to provide low-latency AI services to SMEs, hospitals, and public institutions in South Korea.
In sum – what to know:
Local chips – the initiative supports the construction of regional edge data centers using domestic AI chips.
Regional AI – modular micro data centers will be located in non-metro areas to deliver low-cost AI to SMEs.
Real references – project aims to help the likes of Furiosa AI and Rebellions to build real-world references.
South Korea’s government is moving forward with a KRW27.3 billion ($19.9 million) plan to build micro data centers (MDCs) based on locally manufactured AI chips to serve small and medium-sized enterprises (SMEs), hospitals, and public institutions. The strategy is to expand national AI chip production, drive regional edge deployments of AI services, and to support the national SME sector.
The Ministry of Science and ICT will lead the initiative. The plan is to construct small container-based data centers in non-metropolitan regions across South Korea. These facilities will rely on locally developed AI chips and provide low-latency and cost-effective AI services to SMEs, and local hospitals and public institutions. The project seeks to support domestic chipmakers like Furiosa AI and Rebellions, providing a platform to secure real-world deployments, and also to reduce depedence on US firms like global GPU leader Nvidia.
A ministry official said the project’s goal is to establish a proof-of-concept environment where domestic AI chips can be tested and applied in real data center settings – an area currently dominated by Nvidia. MDCs integrate all core infrastructure (compute, storage, power, cooling, and security) into modular units that can be deployed close to edge locations such as factories or clinics.
Currently, over 90% of commercial data centers in Korea are concentrated in the Seoul metro area, and are largely controlled by global cloud providers like AWS and Microsoft Azure. These global CSPs set strict hardware and operational requirements, leaving little room for infrastructure experimentation by local firms.
As a result, domestic AI chipmakers have faced hurdles in commercializing their solutions. The new MDC project aims to address this by mandating participating companies to create a scalable, cloud-like architecture that links multiple MDCs across the country. The project will run from 2025 through 2029. Regions such as Daegu and Gwangju are currently under consideration for initial deployments, according to the report.
South Korean conglomerate SK Group recently confirmed it is partnering with AWS with the main aim of building what it would be the largest and most powerful AI data center in the Asian nation. Local press said that the construction of the AI-driven facility will begin in August in Ulsan’s Mipo National Industrial Complex, where the 103-megawatt data center will eventually house 60,000 GPUs – the largest GPU deployment ever in a Korean data center.
Reports also highlight that the phased rollout will see 41 MW of capacity come online by November 2027, expanding to the full 103 MW by February 2029. The Korean conglomerate plans to further scale the site to a gigawatt-class (1 GW) facility. Sources, variously cited, said AWS will invest $4 billion, while SK’s key subsidiaries – SK Telecom, SK Broadband, SK Gas and SK Hynix – will contribute substantial resources to the initiative. As part of its broader AI strategy, SK Telecom and SK Broadband plan to invest 3.4 trillion won ($2.5 billion) in AI-related initiatives by 2028.
The data center’s proximity to SK Gas’s LNG combined-cycle power plant ensures stable power delivery, critical for the AI infrastructure.
Chinese company Alibaba Cloud recently said it was opening a second data center in South Korea by the end of last month, further expanding its regional footprint to meet rising demand for cloud computing and artificial intelligence (AI) services. Alibaba Cloud originally entered the South Korean market with the launch of its first data center in March 2022, after announcing the project in late 2021, according to local press reports.
