KB Kookmin to lead record W10.7tr loan for Naver AI project

A syndicated loan is one large loan arranged for a borrower or project by a lead financial institution and funded by several lenders. Each participant commits part of the total. This structure matters when a project needs more money than one lender wants to provide alone. It can also spread credit exposure across the lender group.

Here, KB Kookmin Bank is arranging a 10.7 trillion won syndicated loan for the AI infrastructure project. Other domestic banks and securities firms are participating, although their final commitments and loan terms have not been disclosed. KB Kookmin is handling the structure and key negotiations.

The arrangement supports a project valued at about 13 trillion won. It also gives lenders a shared claim on the project’s financing rather than forcing one institution to fund everything. KB Kookmin called it South Korea’s first GPU financing and largest-ever syndicated loan arrangement.

The financing covers a combined AI computing platform, not just a shipment of chips. Asset Co. will acquire GPUs and other AI computing assets, while also owning the data center infrastructure. Together, these assets create the physical capacity needed for the project.

The article describes a project valued at about 13 trillion won, backed by a 10.7 trillion-won syndicated loan. Asset Co. will supply the resulting computing capacity to Naver Cloud under a long-term agreement. Naver Cloud will operate the AI factory and sell computing services to customers.

This division links the equipment to its business use. Asset Co. owns the GPUs and infrastructure, while Naver Cloud turns their capacity into services and revenue. The financing therefore supports both the expensive computing hardware and the facility that houses and supports it, rather than either component in isolation.

The infrastructure project is valued at about 13 trillion won. Its syndicated loan totals 10.7 trillion won, equivalent to $8 billion in the article. Separately, Naver, Nvidia, and Brookfield announced a broader $10 billion AI factory project, with Brookfield providing up to $9 billion for infrastructure and Nvidia planning a $1 billion investment in Naver.

The computing site is centered on GAK Sejong, Naver’s hyperscale data center. Capacity is planned to reach 55 megawatts in the first half of 2027, 100 megawatts by the end of 2027, and 200 megawatts in 2028. The expanded facility is expected to house about 100,000 Nvidia GPUs.

These figures show the project’s unusual scale. It combines massive equipment spending with a staged capacity rollout. The loan provides the main disclosed financing arrangement, while the long-term infrastructure agreement and planned investments support the broader AI factory buildout.

Brookfield will establish a special purpose company, tentatively named Asset Co., for the project. Asset Co. will acquire and own the GPUs and other AI computing assets, along with the data center infrastructure. Brookfield agreed to provide up to $9 billion for the infrastructure under the broader AI factory plan.

Naver Cloud will operate the AI factory and sell computing services to customers. It will receive computing capacity from Asset Co. under a long-term agreement. Usage fees will be linked to the AI factory business’s revenue, reducing Naver Cloud’s upfront investment burden. KB Kookmin will arrange the syndicated loan.

Naver guaranteed Naver Cloud’s payment obligations under the project contracts. The nominal guarantee is capped at 18.74 trillion won if the full $9 billion investment limit is used. That guarantee does not cover the syndicated loan itself, creating a clear distinction between contract payments and loan repayment.

Lenders need to know how a project will generate cash to meet its obligations. Here, the long-term agreement between Asset Co. and Naver Cloud creates a contracted revenue stream for the infrastructure business. That makes the scheduled service fees central to judging repayment capacity.

The project’s GPUs are expensive assets, but their resale value is difficult to predict, according to earlier media reports cited in the article. If lenders relied mainly on selling the GPUs later, they could face uncertainty about how much those assets would recover. Regular payments offer a more direct source of cash.

The arrangement still depends on Naver Cloud’s ability to earn enough from selling computing services. Usage fees are linked to AI factory revenue, and Naver guaranteed Naver Cloud’s project payment obligations, subject to the stated cap. Lenders will therefore examine both the contracted terms and the business’s ability to sustain those payments.

The project is designed around Naver Cloud paying usage fees for computing capacity. If its AI factory business does not generate enough revenue, those payments could become difficult to make. That would weaken the contracted revenue stream supporting Asset Co. and the infrastructure financing.

The article says Naver guaranteed Naver Cloud’s payment obligations under the project contracts, with a nominal cap of 18.74 trillion won under full use of the investment limit. However, the guarantee does not cover the syndicated loan itself. Lenders would therefore still face loan repayment risk if project cash flow proved insufficient.

The article does not specify the exact consequences of missed payments, such as restructuring, enforcement, or a shutdown. Based on the stated financing structure, weaker payments could put pressure on lenders, Asset Co., and planned AI factory expansion. The project’s ability to sell computing services would remain central to its financial stability.

The source article does not define a GPU. In established computing usage, a graphics processing unit is a processor designed to perform many calculations in parallel. That makes it useful for the matrix and numerical operations common in AI systems. GPUs can support both training models and running them for users.

One GPU offers limited capacity compared with a large AI service. Many GPUs can work together as a computing cluster, allowing larger models and more simultaneous workloads. In this project, Asset Co. will acquire the GPUs and related infrastructure, while Naver Cloud will operate the AI factory and sell computing services.

The planned scale reflects the demand for this capacity. The expanded GAK Sejong facility is expected to house about 100,000 Nvidia GPUs. Its planned capacity rises from 55 megawatts in early 2027 to 200 megawatts in 2028. The article presents this hardware as the foundation of the factory’s computing services.