NVIDIA inks multi-year memory deals for AI “factories” in South Korea, deepening ties with local tech giants

NVIDIA is taking additional steps to secure the hardware backbone for the next wave of artificial intelligence buildouts in South Korea, linking its expansion plans for large-scale AI data centers, often described as “factories,” to multi-year memory supply agreements.

According to a report citing the company’s moves in the region, NVIDIA has entered multi-year agreements with SK hynix to secure next-generation AI memory. Memory is a critical constraint in large AI deployments because training and inference systems require fast, high-capacity memory to keep processors fed with data. The reported agreements are positioned as supply support for NVIDIA’s broader push to scale AI capacity in Korea.

The same report places the memory arrangements inside a wider ecosystem effort. NVIDIA is working with SK Telecom and NAVER on plans intended to support what the companies describe as gigawatt-scale AI cloud infrastructure in Korea. The “gigawatt-scale” framing refers to extremely large power envelopes for data centers, which implies long lead times for power, cooling, networking, servers, and the memory and storage components that can become bottlenecks during rapid capacity ramps.

While details of the memory partnership were not fully spelled out in the Yahoo report, the company has also publicly tied memory development to AI factory expansion. In a separate NVIDIA Newsroom item, NVIDIA and SK hynix announced a multiyear technology partnership aimed at advancing next-generation memory for the global AI factory buildout. That announcement indicates that the collaboration is not limited to near-term supply and includes forward-looking work to improve memory suitability for AI workloads.

Beyond chips and data centers, the broader deal activity in South Korea points to NVIDIA’s strategy of embedding its technology across multiple layers of the AI stack. Other reporting on NVIDIA’s recent engagements in the country described a broader set of partnerships spanning AI infrastructure and robotics, and suggested an expanded cooperation with Hyundai Motor Group that targets industrializing robotics and scaling AI across mobility and manufacturing. NVIDIA has long positioned robotics and automation as future demand engines for AI computing, although the extent and timeline of specific deployments were not detailed in those accounts.

In market terms, the key implication is that NVIDIA’s ability to deliver AI “factories” at scale depends not only on selling GPUs (graphics processing units), but also on securing the surrounding components that make the systems work reliably and efficiently. Memory supply agreements with a major DRAM vendor like SK hynix can help reduce the risk that AI datacenter expansion is constrained by the slower-to-allocate part of the supply chain, especially when data center power and construction schedules are already under strain.

NVIDIA did not disclose in the Yahoo report the financial terms of the SK hynix agreements, the exact memory generations involved, or the expected capacity outputs tied to the partnerships. It also did not quantify timelines for the gigawatt-scale AI cloud effort with SK Telecom and NAVER in the excerpted reporting. Those gaps matter because memory pricing, production ramp schedules, and data center build-out timing will determine whether the partnerships translate into faster revenue growth or mostly function as capacity-risk mitigation.

Investors and industry observers will likely watch for follow-on confirmation on the scale of Korea deployments, including which cloud services and data-center operators will host the AI infrastructure, whether the projects expand beyond pilot phases, and how quickly NVIDIA-linked systems move from development to production. Another watch item is whether memory constraints ease as SK hynix’s next-generation offerings reach higher volumes, since that can influence both cost and delivery schedules for AI platforms globally.