NVIDIA signs South Korea AI partnership cluster aimed at gigawatt-scale “factories,” as valuation debate follows

NVIDIA is expanding its South Korea push for artificial intelligence infrastructure with a new set of partnerships that, according to reporting, targets gigawatt-scale AI “factories” and support for government-aligned, or “sovereign,” and “physical” AI projects. The company’s effort links semiconductor and memory supplier SK hynix, mobile and connectivity provider SK Telecom, and NAVER, the country’s large internet and technology platform, under a single compute-building umbrella.

The partnerships were reported as covering cooperation intended to increase the availability of specialized AI compute systems in South Korea, with the “gigawatt-scale” framing emphasizing the power and data center scale required for training and running large AI models. In practical terms, the emphasis on factories indicates a shift from isolated chip supply toward full-stack ecosystem building, where compute, networking, and data-center capacity are developed together rather than in separate silos.

NVIDIA’s role in the arrangement appears to be centered on providing the graphics processing unit (GPU) and accelerated computing platform technology that powers most modern AI training workloads. While specific hardware configurations were not detailed in the available coverage, the thrust is consistent with NVIDIA’s long-running approach of aligning suppliers and system integrators to deliver complete AI infrastructure packages.

The announcement coverage also described the collaboration as intended to support “sovereign” projects, a term commonly used in policy discussions to mean AI capabilities developed for domestic strategic control rather than relying solely on offshore capacity. It also referenced “physical AI” projects, a phrase typically used for AI systems tied to the physical world, such as robotics, manufacturing optimization, and other use cases where sensing and real-world controls matter as much as raw model performance.

Analysts and investors have focused on the economics of AI infrastructure buildouts, and the reporting highlighted a “valuation tension,” suggesting that even as demand for AI compute remains robust, market expectations for how quickly infrastructure spending converts into earnings and cash flow can run ahead of what is contractually secured. That tension is often most visible when plans are framed in ambitious terms, such as gigawatt-scale capacity, while the industry still depends on power availability, permitting, and data-center construction schedules that can introduce delays.

Beyond NVIDIA, the partner set reflects each company’s likely contribution to the buildout. SK hynix brings memory and storage technologies, which are key bottlenecks for AI data throughput and model training. SK Telecom’s inclusion points to networking and communications capabilities, which are essential for linking large clusters of AI servers. NAVER’s participation suggests a practical layer aimed at turning infrastructure into working services, models, or AI-enabled applications.

Sector context matters because AI infrastructure spending is increasingly a “whole ecosystem” game. GPUs alone do not deliver usable capacity, since deployments also require high-density power systems, cooling, fast networking between servers, and software stacks for training and inference. Partnerships like this are designed to reduce friction across those layers by aligning major suppliers and deployment channels, potentially shortening the time from procurement to production workloads.

What is not clear from the available reporting is how quickly the gigawatt-scale capacity plans will translate into signed volumes of hardware, long-term purchase commitments, or specific project milestones. The coverage referenced the partnerships and the broad goal, but it did not provide, at least in the accessible text, concrete numbers such as megawatts scheduled by date, total capital expenditure, contract sizes, or the expected duration of any related supply agreements.

Looking ahead, investors and industry watchers will likely focus on follow-on disclosure: whether NVIDIA and its partners outline timelines for data-center deployments in South Korea, disclose any commercialization targets tied to sovereign or physical AI initiatives, and clarify how these collaborations map to measurable revenue drivers for NVIDIA’s data center business. Additional detail on system shipments, customer commitments, and the resulting power and compute capacity would help determine whether this effort reinforces the current demand trajectory or extends it into a longer, more uncertain build cycle.