Naver seals $9b Brookfield deal to finance AI factory

Naver’s AI factory is a planned $10 billion computing infrastructure business developed with Nvidia and Brookfield. Its purpose is to provide the large-scale computing capacity needed for AI work, while also creating a business that can serve external customers. Naver will lead the project and operate the business.

The factory will secure GPUs, data centers and related infrastructure. Naver Cloud will access this capacity through a GPU-as-a-service arrangement. It will also offer AI data-center services to outside customers, using Naver’s experience operating GPU clusters and data centers.

Nvidia will provide technical support, including its DSX high-performance computing platform. Brookfield will finance much of the infrastructure through a special-purpose vehicle. This structure aims to make the project more capital-efficient for Naver while allowing computing capacity to grow as demand increases.

The financing structure could provide up to $9 billion, against the project’s total announced value of $10 billion. That scale gives Naver access to substantial computing infrastructure without having to fund the entire build from its own balance sheet at the start. Brookfield, a global asset manager managing more than $1 trillion, is providing the investment structure.

The money will be used to secure graphics processing units, data centers and other AI infrastructure. These assets will sit within a special-purpose vehicle tentatively called Asset Co. Naver Cloud will then access the resulting computing resources and use them to serve customers.

The arrangement is designed to reduce Naver’s upfront capital burden. It also supports staged expansion. As demand grows, Naver Cloud can use more capacity and pay usage fees linked to business revenue, rather than purchasing all infrastructure immediately.

Asset Co is a special-purpose vehicle that Brookfield will establish for the infrastructure financing. Brookfield’s investment vehicle will wholly own Asset Co, giving it responsibility for funding the GPUs, data centers and related assets. Naver, meanwhile, will lead the wider AI factory business and provide operational expertise.

Naver Cloud will not buy all the infrastructure upfront. Instead, it will access the computing resources through a GPU-as-a-service arrangement. It will pay usage fees linked to revenue from the business and provide AI data-center services to outside customers. This links costs more closely to actual business activity.

The arrangement shifts much of the initial financing burden to Brookfield. Yet Asset Co is expected to appear in Naver’s consolidated financial statements as a subsidiary under Korean reporting standards, because Naver is expected to control it. The final structure therefore separates funding ownership from accounting control.

GPU-as-a-service means customers access graphics processing capacity as a service rather than owning all the hardware themselves. In this project, Naver Cloud will use computing resources secured through Brookfield’s infrastructure financing. It can then offer AI data-center services to customers without paying the full construction cost at the beginning.

The key mechanism is usage-based payment. Brookfield finances the infrastructure through Asset Co, while Naver Cloud pays fees linked to revenue from the business. For example, Naver Cloud can access more computing capacity as customer demand rises and pay for that use through the operating arrangement.

This matters because GPUs and data centers require large capital commitments. The structure reduces Naver’s upfront burden and supports staged expansion. It does not remove operating costs, but it allows Naver to align spending more closely with demand and revenue as the AI factory grows.

If demand for AI computing rises, Naver could expand the factory’s capacity in stages rather than building everything before customers arrive. The article describes this as a capital-efficient structure. It could help Naver pursue growth while limiting the amount of money it must commit upfront to GPUs, data centers and related infrastructure.

The mechanism is the GPU-as-a-service arrangement. Brookfield finances the assets, while Naver Cloud pays usage fees linked to revenue from the business. Naver Cloud can therefore access additional computing resources as demand grows and use them to serve external customers. Naver’s operating experience would support the business.

The project could create new growth opportunities in the global AI infrastructure market. However, higher demand would also mean greater usage and potentially higher fees. The article does not quantify future revenue or profits, but it presents staged expansion and lower initial capital needs as the main advantages.

Nvidia’s importance comes from its role in supplying the project’s core computing technology and technical support. The article identifies graphics processing units as essential infrastructure for the AI factory. Nvidia will also support the system with its DSX high-performance computing platform, connecting the project to its advanced computing capabilities.

In practical terms, the GPUs provide the processing capacity that Naver Cloud will access and offer to customers. Nvidia’s DSX platform is intended to improve how that high-performance infrastructure operates. The article says this support should improve efficiency and competitiveness, while Naver contributes experience in building GPU clusters and managing data centers.

The partnership is therefore broader than a hardware purchase. Naver leads the business, Brookfield finances its infrastructure, and Nvidia supplies technical expertise. Nvidia is also set to invest $1 billion in newly issued Naver shares, potentially becoming Naver’s third-largest shareholder, which would deepen the relationship.

Modern AI systems require enormous computing power because they perform vast numbers of calculations when models learn from data and when they generate results. Specialized GPUs are well suited to handling many calculations at once. Training can involve repeated processing across large datasets, while running the finished systems still requires capacity for users and applications.

That workload depends on more than chips alone. GPUs must be connected into high-performance clusters, supported by data centers, storage, networking, power and cooling. In this project, Naver will draw on experience operating GPU clusters and data centers, while Nvidia will provide its DSX platform to improve infrastructure efficiency.

The article’s financing structure reflects this scale. Brookfield will secure GPUs, data centers and other infrastructure with up to $9 billion. Naver Cloud will access that capacity as a service and expand it in stages. The article does not quantify specific workloads, energy use or data volumes.