How to Build AI Data Centers Amid Soaring Demand... PwC Consulting Releases Report

Recommendations for Regional AI Data Center Construction
Challenges in Building Hyperscale Data Centers in Urban and Metropolitan Areas
The Need for Diverse Business Models Amid Infrastructure and Regulatory Hurdles
Proposals for Utilizing Intermediate and Edge Data Centers Across Multiple Regions

There has been a recommendation to build various types of data centers by region (urban, greater metropolitan area, and non-metropolitan area) in order to prepare for the increasing domestic demand for artificial intelligence (AI) data centers.

On May 21, PwC Consulting announced the publication of a report titled "New AI Data Center Business Opportunities in the Era of Generative AI." The report examines the current state of AI data centers in Korea, whose necessity is growing with the advancement of generative AI technologies, and explores new business opportunities.

According to the report, the market for AI semiconductors for data centers is growing at an average annual rate of 94.9% from 2022 to 2025. Unlike conventional data centers, which are built to store, use, and share data among users in the cloud, AI data centers require the ability to process and transmit large volumes of data for AI model training and inference, as well as advanced real-time response capabilities.

To minimize latency, AI data centers require two types: hyperscale and edge AI data centers. While hyperscale data centers focus on large-scale data learning and computation, edge data centers are located near users who generate or utilize information, enabling rapid processing of real-time generated data or quickly transmitting information that requires advanced computation to hyperscale data centers.

The report stated, "As AI models become more complex, it is necessary to newly build hyperscale data centers connecting thousands of servers to meet the needs of AI service providers, in order to attract global AI companies to Korean data centers." However, it also pointed out, "In Korea, electricity demand is concentrated in Seoul and the greater metropolitan area, and the transmission and distribution network for delivering electricity from power plants to demand centers has reached its limit, making it realistically difficult to build hyperscale AI data centers in urban or metropolitan areas."

Furthermore, the report expressed concern that "in provincial areas, where there is relatively more available power, a lack of infrastructure is causing demand for data center construction to flow overseas," and added, "Due to regulations such as the Enforcement Decree of the Electric Utility Act and the Special Act on Distributed Energy Promotion, which are not aligned with the needs of global data center customers, it is difficult to develop, build, and attract tenants to domestic data centers."

Therefore, the report emphasized the need for multifaceted data center construction business models. It suggested, "It is necessary to build various types of data centers not only in urban and metropolitan areas, but also in non-metropolitan regions." Representative proposals included leasing surplus space in third-party intermediate-region data centers to operate intermediate computing centers, or building or outsourcing the operation of small to mid-sized edge data centers in multiple regions.

Yoo Wonseok, Head of AI Data Center Strategy and Implementation Services at PwC Consulting, stated, "In response to the recent surge in generative AI usage, the construction and expansion of domestic AI data centers will continue," and emphasized, "Going forward, a differentiated approach is needed to build various types of AI data centers by region."

This content was produced with the assistance of AI translation services.