AI model developers at SK Telecom’s Euljiro headquarters hold a remote meeting with NVIDIA representatives. / Courtesy of SK Telecom

SK Telecom said on the 22nd that it will continue collaboration with NVIDIA in AI model development, unveiling the results of their technological cooperation at the technical panel session of “NVIDIA Nemotron Developer Days Seoul 2026,” held for the first time in Korea on the 21st.

The two companies have closely collaborated across data, infrastructure and training environments since the construction of SK Telecom’s A100 GPU-based supercomputer “Titan,” unveiled in 2021. They have continued active technical cooperation throughout the development of SK Telecom’s proprietary AI models.

In particular, SK Telecom adopted NVIDIA’s Nemotron dataset for training its hyperscale model A.X K1 (A-dot X K1), which has 519 billion parameters, as part of the government-led “sovereign AI foundation model project” conducted last year.

In the process, SK Telecom and NVIDIA enhanced the stability of large-scale AI training by utilizing Megatron-LM, a distributed training framework, and NVIDIA NeMo Curator for data preparation and refinement.

This collaboration enabled SK Telecom to improve AI model performance, while NVIDIA enhanced the completeness of its software development frameworks, creating a virtuous cycle. SK Telecom’s model development division and NVIDIA’s technical support teams communicate biweekly to discuss infrastructure stability, performance improvements and optimization strategies.

The companies have also shared experience in building hyperscale models using new architectures such as mixture of experts (MoE), along with related development infrastructure insights, establishing a foundation for more precise and stable large-scale training.

SK Telecom plans to continue leveraging NVIDIA’s solutions for training its next-generation model A.X K2, currently under development, while maintaining technical exchanges to further enhance capabilities on both sides. The companies are also conducting joint research in next-generation technologies, including multimodal models and vision-language models (VLMs).

Kim Tae-yoon, head of foundation model development at SK Telecom, said, “SK Telecom and NVIDIA have advanced together by proactively adopting new technologies and exchanging feedback,” adding, “Through this partnership, we will contribute to the development of the domestic AI ecosystem.”

SK Telecom said on the 22nd that it will continue collaboration with NVIDIA in AI model de …

SK Telecom, NVIDIA accelerate A.X K2 development through data and infrastructure collaboration