SK Group and NVIDIA have announced an expanded strategic partnership that covers two of the most consequential segments of the AI infrastructure supply chain: AI factories and next-generation memory technology.

The agreement broadens a relationship that already spans multiple layers of the AI hardware stack. SK Group, the South Korean conglomerate whose affiliates include major memory suppliers, supplies components that sit at the heart of NVIDIA's accelerated computing platforms. Memory, in particular, has become a critical bottleneck as AI models grow larger and inference workloads demand faster data movement between compute and storage.

The expanded cooperation addresses that pressure directly. Under the partnership framework, the two companies will work together on AI factories — large-scale, purpose-built data centers assembled around NVIDIA computing platforms — and on the memory technologies those facilities require.

For SK Group, the deal deepens its exposure to the capital spending wave now reshaping the data center industry. AI factory construction requires not only GPUs but also high-bandwidth memory, advanced packaging capacity, networking equipment, power infrastructure and integration services. SK's affiliates operate across several of these categories, which gives the group multiple entry points into each project cycle.

For NVIDIA, the partnership secures closer alignment with one of the few suppliers capable of delivering memory at the scale AI factories demand. High-bandwidth memory remains one of the tightest constraints in the accelerated computing supply chain. Closer coordination with SK Group gives NVIDIA a more predictable path to the memory capacity its roadmap depends on.

The memory component of the agreement points toward next-generation products. As model sizes increase, the volume of data that must move between processors and memory rises in step. Memory bandwidth, capacity and power efficiency have therefore become primary design constraints for AI systems, not secondary considerations. Cooperation on next-generation memory signals that both companies intend to develop these technologies in tandem rather than sequentially.

The AI factory side of the partnership reflects a structural shift in how computing capacity gets built. Instead of general-purpose data centers, AI factories are designed specifically for training and inference at scale, with power delivery, cooling and networking architected around accelerated computing from the start. NVIDIA has promoted this model as the standard for future AI infrastructure, and partnerships with major industrial groups are the mechanism through which these facilities get financed, built and equipped.

For the broader market, the announcement carries several implications. It tightens the link between memory suppliers and platform vendors, a trend that has already reshaped procurement dynamics across the semiconductor industry. It also signals that competition in AI infrastructure now runs at the level of entire ecosystems — chip designers, memory manufacturers, integrators and builders — rather than individual components.

SK Group affiliates involved in memory production stand to gain sustained demand visibility from the arrangement. That visibility matters for capacity planning, since memory fabs require multi-year investment commitments and demand cycles in AI hardware remain volatile.

The two companies have not framed the partnership as exclusive. Instead, the expansion establishes a broader working relationship across the AI infrastructure stack, spanning factory-scale deployments and the memory technologies that underpin them.

Details of the announcement appeared on the NVIDIA Newsroom. The partnership builds on prior cooperation between the two companies and extends it into the two categories where demand pressure is highest: complete AI factory infrastructure and the next generation of memory products designed to feed it.