Naver’s $10 Billion AI Data Center Plan Tests Korea’s Financing Market

Naver’s plan to expand a major artificial intelligence data center in Sejong is becoming an early test of whether South Korea’s financial market can support the vast capital needs of next-generation computing infrastructure.

The Korean technology company and global asset manager Brookfield are discussing financing of up to $9 billion for the first expansion phase of the Gak Sejong AI factory. The broader project is valued at about $10 billion, with Nvidia planning to invest $1 billion. The initial negotiating period is due to expire next week, but discussions are expected to continue while the parties assess the financing structure and commercial viability.

The plan would increase the facility from an initial capacity of 55 megawatts to 200 MW and equip it with roughly 100,000 Nvidia graphics processing units. Operations at 55 MW are scheduled to begin in the first half of 2027, followed by an increase to 100 MW by the end of that year and 200 MW in 2028. Naver ultimately envisions a much larger 1-gigawatt facility.

Several major Korean financial institutions are considering participation, including KB Kookmin, Shinhan, Hana and Woori banks, as well as the state-run Korea Development Bank. Securities companies are also examining potential roles in arranging and underwriting parts of the package before distributing exposure to other investors.

However, the involvement of Naver, Nvidia and Brookfield does not remove the core risks for lenders. Unlike conventional data centers, AI facilities require heavy spending not only on buildings, electricity and cooling systems but also on GPUs that can become outdated relatively quickly. Financiers must judge whether customer demand, equipment utilization and recurring revenue will remain strong enough to support the debt over time.

The project’s scale adds further uncertainty. A move from 200 MW to 1 GW would require substantially more capital and electricity, along with a reliable supply of advanced processors. Korea has limited experience financing a single AI data center at that size, making a staged funding approach more likely.

Despite the challenges, AI data centers are attracting attention from Korean lenders as traditional property project financing faces high construction costs, elevated borrowing expenses and weak development economics. Data centers can resemble long-term infrastructure investments when they have strong tenants and predictable cash flow, but AI adds uncertainty over hardware cycles and future computing demand.

A successful financing package could provide a model for other large Korean AI infrastructure projects. Difficulty raising the funds, even with three prominent sponsors, would instead underscore how cautious investors remain about turning rapid AI investment into sustainable returns.