SK telecom is stepping up efforts to strengthen its large-scale artificial intelligence (AI) capabilities through an expanded partnership with Nvidia, aiming to build a competitive sovereign AI ecosystem in South Korea.
The collaboration goes beyond conventional technology integration, covering the full AI development cycle—from data processing and infrastructure to model architecture. The move reflects a broader strategy to reduce reliance on external AI platforms and establish domestic capabilities in foundation model development.
At the “Nvidia NeMoTron Developer Days Seoul 2026” event held on April 21, SK telecom outlined progress in its joint initiatives with Nvidia, highlighting advancements in its proprietary AI foundation model series, branded “A.X.”
Following the development of its 519-billion-parameter model, A.X K1, the company is now accelerating work on its next-generation model, A.X K2. The new model is expected to incorporate more advanced architectures and expanded capabilities.
The partnership between the two companies began at the infrastructure level. Nvidia played a key role in supporting the design and optimization of SK telecom’s A100 GPU-based supercomputing system, known as “Titan,” which was introduced in 2021. Since then, the collaboration has extended into model development, contributing to improvements in training performance and scalability.
During the training of large-scale AI models, SK telecom has leveraged Nvidia’s AI software ecosystem, including the Megatron-LM framework for distributed training and NeMo Curator for data processing. This combination has enabled the company to enhance both training efficiency and model stability.
The two companies have maintained ongoing technical coordination, holding regular discussions to optimize infrastructure and refine model performance. They have also shared expertise in next-generation model design, including mixture-of-experts (MoE) architectures, which are increasingly used to scale AI systems efficiently.
Looking ahead, the scope of collaboration is expected to expand further. SK telecom plans to continue utilizing Nvidia’s technologies in the development of A.X K2, while also exploring joint research in emerging areas such as multimodal AI and vision-language models.
Industry analysts view the partnership as part of a broader competition over AI sovereignty. As global tech giants consolidate their dominance in AI platforms, countries and companies are increasingly seeking to secure their own models and infrastructure. In this context, SK telecom’s strategy is seen as an effort to narrow the technological gap through strategic alliances.
Ultimately, the collaboration between SK telecom and Nvidia is likely to serve as a test case for strengthening Korea’s AI ecosystem while enhancing the global competitiveness of its technology sector.
