NAVER, NVIDIA and Brookfield plan to scale South Korea’s national AI factory to 200 megawatts by 2028

NAVER said it, NVIDIA and Brookfield are planning to expand South Korea’s “national AI factory” infrastructure buildout, scaling an initial multi-tenant AI factory program built around NVIDIA technology to 200 megawatts of capacity by 2028. The announcement was reported as part of a broader effort to provide shared, data-center-like AI capacity for developers rather than requiring each company to build infrastructure from scratch.

In the reported plan, the initial phase centers on NAVER’s multi-tenant NVIDIA DSX AI Factory. DSX is NVIDIA’s data center software and systems stack intended to help customers deploy and manage AI workloads in production environments, typically by combining accelerated computing with the operational tooling around it. NAVER is positioning the AI factory concept as a standardized platform that can be used by multiple AI teams.

The expansion target of 200 megawatts by 2028 underscores how electrical power and cooling capacity are increasingly the binding constraints for AI deployment. Rather than focusing only on chips, these “AI factory” initiatives aim to increase the amount of power available for high-density servers and accelerate timelines for new AI model training and inference deployments.

Brookfield is named in the effort as the infrastructure partner, reflecting growing investment interest in data-center scale and buildouts tied to AI workloads. The reported structure also highlights a common pattern in major AI infrastructure projects: technology providers help define the compute and software stack, while major asset managers or infrastructure firms support facility expansion and capital deployment.

For NVIDIA, the DSX AI Factory framing ties its data center ecosystem to concrete capacity builds. The company’s hardware and software platforms are typically used by enterprise and cloud customers, but national or shared AI factory projects can expand the addressable base by giving local developers access to a consistent deployment environment.

For South Korea’s AI industry, the reported buildout target positions NAVER’s factory model as a national capacity layer that can serve multiple AI innovators, including organizations that may not want to undertake large, individualized data-center construction. Shared infrastructure can also reduce friction around standardized performance, security practices, and operational support when multiple teams are developing models simultaneously.

Still, key operational details remain unclear from the report. It does not specify, in the information provided, how capacity will be allocated among participants, the timeline for incremental phase additions before 2028, or what mix of workloads the expanded facilities are expected to prioritize. The report also does not lay out pricing, tenancy terms, or whether capacity will be reserved primarily for NAVER partners versus broader external customers.

What to watch next is how NAVER and its partners translate the power target into a phased rollout, including any milestones for facility commissioning and how NVIDIA DSX capabilities are configured for multi-tenant use. Future disclosures could also clarify which AI developers gain access, and whether additional technology partners join as the AI factory scales toward the 200-megawatt goal.