Naver signs first AI factory customer in 2 trillion won computing deal

Naver said on November 12 that it had secured an initial customer for its AI factory business through a computing services contract worth about 2 trillion won. On the third-quarter earnings call, Chief Executive Choi Soo-yeon said the customer is a U.S. AI inference company and that the service includes next-generation graphics processors. The deal gives the customer computing capacity for inference, the work a model does when it generates a response to a request.

Naver expects to begin recording AI factory revenue in the first half of 2027. The company described the first contract as about 20 megawatts of capacity, according to coverage of the earnings call by Aju Business Daily and Yonhap Infomax. The supply period is four to five years. The 2 trillion won figure covers that period. The published results did not identify the customer or break the contract value into annual amounts.

Naver secured infrastructure before signing the customer

According to EBN’s coverage of the call, the equipment for the customer includes Nvidia B300 and Vera Rubin GPUs. The computing service gives the buyer access to the equipment for the contract period without building a data center of its own for the job. The AI factory combines graphics processors and data center operations to provide computing as a service to businesses. Servers need electricity and cooling, as well as a network to move data between machines. A customer contracts for capacity on infrastructure that Naver operates. Naver has identified its GAK Sejong data center as the hub for the project.

The infrastructure arrangement predates the customer announcement. On October 27, Naver disclosed a deal with Brookfield worth up to $9 billion to secure computing infrastructure. Under that arrangement, a Brookfield-backed entity will acquire assets such as graphics processors and data centers and supply computing capacity to Naver Cloud. Naver Cloud will operate the AI factory. Nvidia is expected to provide computing technology and support for the buildout.

The October agreement gives the operator access to the infrastructure it needs to provide computing services. The November agreement brings in a customer for those services. Naver previously described a payment structure that links usage fees for the underlying infrastructure to the business's revenue. It said it would also guarantee Naver Cloud's fee obligations for up to eight years. The contracts have different counterparties, cover different transactions and are denominated in different currencies.

The 20-megawatt contract sits within a wider buildout

Naver and its partners plan to begin operating at 55 megawatts in the first half of 2027. Nvidia and Brookfield have described a plan to expand the deployment around GAK Sejong to 200 megawatts by 2028. In this context, megawatts describe power capacity for the equipment and facility. The amount of computing that fits within that capacity depends on the hardware and how the facility is operated.

The roughly 20 megawatts discussed on the earnings call describes the size of the first customer contract. The 55-megawatt figure describes the initial operating plan for the project as a whole. Naver’s published materials did not specify how many GPUs it would allocate to the customer or give a delivery schedule for individual phases. EBN and Aju Business Daily reported that the company is discussing further supply agreements with more than 10 global AI companies.

Naver reported third-quarter consolidated revenue of 3.4724 trillion won, up 10.7% from a year earlier. Operating profit fell 20.9% to 451.3 billion won. The company attributed the decline in part to spending on AI competitiveness, including infrastructure. Its earnings slides showed infrastructure expenses rising 29.1% year over year to 282.3 billion won and cited the arrival of computing equipment for internal use and the opening of data center space.

The earnings slides list the deal as securing a customer for initial 2027 capacity. Naver expects the AI factory business to start contributing to reported revenue as it begins service in the first half of 2027. Chief Financial Officer Kim Hee-cheol said on the call that the company expects the AI factory business to generate an operating profit in 2027, according to EBN.