en.Wedoany.com Reported - SK Telecom (SKT) is going all-in on AI. From its self-developed large language model A.X and multiple AI agents, to partnering with OpenAI and establishing subsidiaries to accelerate AI innovation, the telecom operator's business footprint continues to expand. Now, it has embarked on an even more ambitious plan: building a sovereign AI stack.
SKT has formed an alliance with semiconductor company Arm and accelerator startup Rebellions to jointly design custom data center hardware dedicated to inference workloads, constructing a sovereign AI stack. The three parties will also simultaneously develop a complete software stack to accompany the hardware. Addressing the question of "why build your own instead of buying off the shelf," Lee Jaeshin, Head of AI Business Development at SKT, explained to the media that combining Arm's AGI CPU with Rebellions' RebelCard accelerator creates a heterogeneous architecture that achieves higher energy efficiency and lower operational costs compared to traditional GPU-based servers. He further pointed out that this translates to greater inference throughput under the same power constraints, a critical advantage as data centers are increasingly limited by power capacity rather than compute budgets.
The three parties plan to conduct technical trials in SKT's data centers to verify the performance and stability of the sovereign AI stack. Lee revealed that the operator's ultimate goal is to deploy its AI foundation model, A.X K1, to run on the self-developed stack, thereby strengthening autonomous control over AI infrastructure.
Beyond pursuing the efficiency gains of self-developed infrastructure, an SKT spokesperson stated that the growing demand for sovereign AI is also a significant driving force. Gartner predicts that as regulatory frameworks adapt to AI development and geopolitical tensions intensify, global spending on sovereign cloud infrastructure-as-a-service will exceed $80 billion in 2026 and surpass $100 billion in 2027. Sovereign AI, as a deeper branch of this trend, involves both the infrastructure and model layers.
The South Korean government launched the "AI Foundation Model" project last year, selecting five teams to build a sovereign foundation model for the nation. SKT was one of the selected parties and became one of three companies to enter the second phase after the initial review. Notably, SKT's Haein GPU cluster—one of the largest in South Korea—has been used as the compute layer for the government's foundation model project. Furthermore, SKT has committed to building 1 GW of AI data center capacity nationwide.
The spokesperson stated that SKT's work "can serve as a reference case for how telecom operators can find new growth engines through AI." He emphasized that telecom companies bring differentiated assets to sovereign AI: owning and operating the network, data centers, and the AI models running on them, achieving end-to-end integration—a capability that is difficult to replicate. "We believe this business prospect is real and vast, and SKT's sovereign AI stack, Haein, and A.X K1 practices are proving this," the spokesperson concluded.
This article is compiled by Wedoany. All AI citations must indicate the source as "Wedoany". If there is any infringement or other issues, please notify us promptly, and we will modify or delete it accordingly. Email: news@wedoany.com
