NAVER seals $9 billion Brookfield AI infrastructure deal, clearing condition for NVIDIA investment

A $10 billion AI project moves toward construction

NAVER has completed an agreement with Brookfield to secure up to $9 billion in artificial intelligence computing infrastructure, clearing a condition for NVIDIA's planned $1 billion equity investment in the South Korean technology company. The project will begin at NAVER's GAK Sejong data center, with an initial operating target of 55 megawatts in the first half of 2027.

The agreement, disclosed on October 27, 2026, turns a partnership announced in July into a more detailed arrangement for buying equipment, supplying computing services and guaranteeing payments. NAVER Cloud will operate the AI factory business, using infrastructure financed through Brookfield to serve companies and public institutions.

NAVER plans to reach 100 megawatts by the end of 2027 and 200 megawatts in 2028, before pursuing a longer term ambition of one gigawatt. The July announcement put the 200 megawatt deployment at approximately 100,000 NVIDIA graphics processing units, or GPUs, the chips used to train and run many AI systems.

The financial commitment extends beyond construction. NAVER has disclosed a payment guarantee of 18.74 trillion won for its subsidiary NAVER Cloud, equivalent to 64.73% of NAVER's equity. That guarantee connects the parent company's finances to the payments its cloud business will make for access to the infrastructure.

For customers, the proposed service offers access to powerful computing equipment without having to buy and operate every component themselves. For NAVER, the commercial task is to secure enough paying users to cover infrastructure charges while turning the expansion into a profitable business.

What the partners have agreed to fund

NAVER, Brookfield and NVIDIA described a $10 billion project in their July 24 joint partnership announcement, released at the K-AI Summit in San Francisco during South Korean President Lee Jae Myung's visit. Brookfield was named the exclusive capital partner, providing up to $9 billion, while NVIDIA committed $1 billion and NAVER agreed to fund the remaining amount.

Brookfield's contribution is a ceiling rather than a fixed drawdown. Although its maximum commitment and NVIDIA's planned investment together equal the announced $10 billion project size, NAVER's eventual contribution will depend on how the financing and expenditure are completed.

Brookfield's investment vehicle will establish a Korean company provisionally called Asset Co. to hold infrastructure assets. The arrangement provides for investment in GPUs, data center facilities and related equipment, with NAVER Cloud receiving computing services from that company.

NAVER will lead site acquisition, construction and operations. NVIDIA will supply GPUs and provide technical cooperation through its computing platforms. NAVER Cloud will manage the platform and customer services, drawing on its experience operating large GPU clusters and data centers.

How the agreement developed

NAVER's July partnership account described the Brookfield agreement signed on July 22 as preliminary and said a definitive contract still needed to be signed. The October disclosure marks the later completion of the infrastructure agreement.

The main milestones show how the project has progressed from an expansion plan to contracts and scheduled deployment:

• June 8, 2026: NAVER disclosed plans for the AI factory expansion while infrastructure financing discussions were continuing.
• July 22, 2026: NAVER signed a preliminary agreement worth $9 billion with Brookfield.
• July 24, 2026: NAVER signed a $1 billion strategic investment agreement with NVIDIA and announced the expanded partnership in San Francisco.
• October 27, 2026: NAVER disclosed completion of the Brookfield infrastructure agreement and the related payment guarantee.
• First half of 2027: NAVER targets initial operations at 55 megawatts.
• End of 2027 and during 2028: Planned cumulative capacity rises to 100 megawatts and then 200 megawatts.

NVIDIA's share subscription payment is scheduled for October 30. The completed infrastructure agreement satisfies the prerequisite that NAVER secure the computing resources needed for the expansion, allowing the company to move toward finishing the share issuance.

Why NVIDIA is also becoming a shareholder

NVIDIA plans to buy 7,241,564 newly issued NAVER shares at 204,500 won each. The subscription would raise about 1.48 trillion won, described as approximately $1 billion, and give NVIDIA a stake of about 4.5%. NAVER described the agreement as NVIDIA's first strategic investment of this kind in Asia.

The transaction is a third party allotment, meaning NAVER issues new shares to a selected investor rather than offering them through a general public sale. NAVER receives the subscription money, while NVIDIA becomes a shareholder alongside supplying technology for the project.

NVIDIA's operational role has also become more specific. A July correction to NAVER's disclosure removed earlier wording about NVIDIA finding global customers and sharing revenue and business risks, and instead identified it as the GPU supplier. Customer acquisition therefore remains a central task for NAVER's operating business.

The AI factory will use NVIDIA's DSX architecture, which brings together the computing infrastructure needed for large AI workloads. Blackwell GPUs and the next generation Vera Rubin platforms are to be introduced in phases, rather than through a single equipment installation.

What 200 megawatts means for customers

GAK Sejong, which opened in 2023, will be the first base for the project. NAVER previously opened GAK Chuncheon in 2013, giving it experience running its own data centers before this larger partnership.

An AI factory combines GPUs with networks, storage, electricity, cooling and software. The name describes a facility that repeatedly produces AI outputs, such as trained models or responses to users, using large amounts of computing power.

Megawatts measure electrical power, which helps describe the scale of a computing facility. NAVER's planned increase from 55 to 200 megawatts would make the deployment about 3.6 times its initial size. The eventual one gigawatt ambition equals 1,000 megawatts, five times the 2028 target.

NAVER plans to provide GPU as a Service, often shortened to GPUaaS. Customers pay to use computing resources instead of purchasing the chips and supporting equipment themselves. The July announcement described a dedicated resource pool offering computing capacity, software and support for emerging AI companies in South Korea and the United States.

The guarantee puts NAVER behind the payments

NAVER's 18.74 trillion won guarantee covers NAVER Cloud's payment obligations under the project agreements. The disclosed period runs from October 27, 2026, through December 31, 2034. Company descriptions refer to a guarantee of up to eight years; the specified calendar dates extend slightly beyond eight years.

NAVER calculated the amount as the nominal total of expected supply payments through the end of 2034, assuming the entire $9 billion infrastructure investment limit is used. Nominal total means the payments are added together at their stated future amounts.

NAVER said no supply payment or other payment obligation had become fixed at the disclosure date. Actual amounts will depend on the GPU investment schedule, interest rates, exchange rates and cost settlements. The guarantee itself requires NAVER to stand behind covered obligations rather than pay the entire stated sum immediately.

The new commitment brings NAVER's total outstanding guarantees to approximately 19.49 trillion won. With the latest guarantee equivalent to nearly two thirds of equity, payment performance at NAVER Cloud becomes a substantial financial consideration for its parent.

NAVER Cloud is also expected to pay NAVER 670.725 billion won in guarantee fees across the disclosed period. The estimate uses an annual fee rate of 0.4375% and assumes full deployment of the investment limit. The eventual fee total can change with service commencement, the scope of the guarantee and tax review.

Outside financing still carries responsibilities

Brookfield's investment vehicle will provide 100% of the equity in Asset Co., but NAVER expects the company to become its subsidiary and affiliate under Korean International Financial Reporting Standards. NAVER says the accounting treatment reflects its control over the entity.

Ownership and accounting control can differ. A company financed entirely by one investor can still be included in another company's consolidated accounts when the accounting rules recognize that other company's control. Asset Co. is expected to fall within NAVER's group accounts even though Brookfield supplies its equity.

NAVER Cloud will pay infrastructure usage fees linked to business revenue. This structure spreads the initial equipment financing burden across external capital and future service payments, while NAVER's guarantee provides financial backing for the contractual obligations.

The commercial terms still needing public detail include usage fee levels, payment formulas, financing rates, customer commitments and any NAVER minority investment in the vehicle. Those terms will determine how closely the business's infrastructure expenses move with the money it earns from customers.

Customer prices will determine the margin

Eugene Investment & Securities analyst Jeong Ui-hoon estimated an operating profit margin of 5.9% under a scenario in which one gigawatt of IT power capacity is fully operating and GPU rental prices average $3 per hour. His calculation put the operation at a loss if the hourly price fell below $2.80.

The gap between those prices is just $0.20, or about 6.7% of the $3 assumption. The calculation illustrates how a relatively small decline in rental prices could absorb the operator's margin under that scenario. It applies to a fully operating one gigawatt business, far larger than NAVER's initial deployment.

Customer contract length will also matter. Shorter agreements give NAVER opportunities to change prices as the market moves, while longer agreements can provide steadier revenue. Infrastructure payments that continue for years create a need to manage both the number of customers and the duration of their commitments.

Korea Investment & Securities analyst Jeong Ho-yoon identified delays in customer contract announcements as a source of investor uncertainty. He interpreted NAVER's approach as an effort to improve profitability by combining customers on contracts of different lengths, and expected substantial earnings contributions from 2028 as capacity expands.

NAVER's overseas plan starts with Asia and the Pacific before extending to Europe and the Middle East. The company has described deployment dates as provisional: securing sites, permits and electricity infrastructure, along with changes in equipment supply, can alter investment amounts and schedules.

Computing capacity connects to NAVER's own AI plans

NAVER is pursuing software and model development alongside the infrastructure expansion. The July announcement said it was advancing HyperCLOVA X using NVIDIA's Nemotron 3 Ultra open models, combined with NAVER's proprietary data and training expertise.

NAVER also became the first Korean company to join the NVIDIA Nemotron Coalition. Its work covers initial model training, later refinement and reinforcement learning, a process in which feedback helps a model improve its decisions or responses.

NAVER plans to launch an AI agent platform in South Korea in the second half of 2026 using NVIDIA Agent Toolkit software, including NemoClaw blueprints. AI agents are systems designed to carry out tasks through a sequence of actions. NAVER is also developing a Seoul World Model using its urban street imagery and spatial data with NVIDIA Cosmos models.

NAVER founder and board chairman Lee Hae-jin framed the partnership around supporting a wider range of AI systems for different communities. Speaking at the July summit, he said:

I believe that in order to preserve diversity in this world, diverse kinds of AI need to be created and allowed to operate for each country, group, and individual.

The infrastructure project gives that ambition a deployment schedule and a financing structure. Its commercial progress will depend on turning the planned computing capacity into services that businesses and public institutions pay to use.

Key Points

• NAVER completed a Brookfield agreement covering up to $9 billion in AI infrastructure, clearing a condition for NVIDIA's planned $1 billion share investment.
• GAK Sejong will be the first base, targeting 55 megawatts in early 2027, 100 megawatts by that year's end and 200 megawatts in 2028.
• NAVER disclosed an 18.74 trillion won payment guarantee, equivalent to 64.73% of its equity, covering obligations through December 2034.
• Brookfield will finance infrastructure through a dedicated company, while NAVER Cloud operates services and pays usage fees.
• Customer demand, rental prices, financing terms and access to electricity will shape profitability and the pace of expansion.