Naver AI Data Center: Sejong GPU Capacity Resale Plan

Naver plans to sell AI computing capacity directly through a wholly-owned subsidiary that would rent out GPU-equipped infrastructure from its Gak Sejong facility in South Korea. The Naver AI data center project is financed by Brookfield Asset Management and equipped with Nvidia hardware. The subsidiary launch is planned for late 2026, and no pricing or capacity-sale terms have been disclosed.

The plan changes Naver's role at the site. Instead of only operating infrastructure for its own use, the company would sell compute to outside customers. Korean AI server distributors such as Kolon Benit and Etevers report record revenues as firms scramble for GPU capacity, which shows the demand the subsidiary would target. Those distributor figures measure hardware sales, not rental demand, so they support the demand signal without confirming it.

Naver AI data center financing and scale

According to Naver's official announcement, the planned funding will scale up the first Nvidia DSX AI factory site in Sejong, growing its capacity from 55 megawatts to 200 megawatts. A July 2026 summary, a secondary source, adds a 2028 target for the 200 MW build and a longer gigawatt-scale roadmap. The 200 MW target is about 3.6 times the initial capacity.

A deal note from a lower-credibility source puts project finance at $10 billion, with Brookfield providing $9 billion and Nvidia $1 billion. No filing or official announcement in the material confirms that split, so the breakdown should be treated as unverified.

Competitive position and open questions

A seller of capacity carries utilisation risk that a hardware distributor does not. Naver would need to price rented GPU time against the cost of the supply it secures at Sejong. Customers short on GPU access stand to gain the most from a domestic option, while server distributors that sell hardware outright face pressure if buyers shift toward rented capacity. This is analysis, not a reported outcome.

Several questions remain open. The material does not include pricing, customer terms, or named customers. The 2028 timeline comes from a secondary summary, and no manufacturer press release for the subsidiary was retrieved. Whether the late 2026 launch holds will show how quickly the Sejong capacity reaches paying users.

Why this matters

For enterprises in South Korea, a domestic capacity seller could shorten waits for GPU access, but the terms that would deliver that benefit are not yet public. Scaling a site to 200 MW brings costs beyond the compute itself. Many AI data centers run on on-site gas turbines that are loud for nearby residents, gas-fired power adds CO2 emissions, and cooling consumes large volumes of water. The Sejong material does not address these effects, so residents and regulators should ask for them to be disclosed as the build-out proceeds.

Sources

NAVER, NVIDIA and Brookfield to Expand Korea’s

AI-generated image.

✔Human Verified

Researched and cross-referenced against primary sources by the Bytevyte editorial team. This article was generated with the assistance of artificial intelligence and reviewed by the Bytevyte editorial team.