SK telecom bets big on AI data centers, eyes 15GW lead

The gist
SK Telecom is gunning for AI data center supremacy in South Korea, targeting a jaw-dropping 15GW capacity by 2035 and leaving rivals in the dust.
What to know
- SK Telecom has spun off SK Horizon, backed by $2.2 billion from KKR and IMM, to turbocharge its AI data center push while keeping 51% control.
- The company’s AI ‘factories’ will leverage in-house models like the 688B-parameter A.X K2 and SK Group’s semiconductor muscle for maximum efficiency and profit.
- South Korea’s AI data center boom is triggering a national energy rethink—with power demand set to jump by 260 TWh by 2030 and urgent calls for better tax breaks and grid upgrades.
SK Telecom’s 15GW Power Play
SK Telecom is outpacing rivals with a 15GW AI data center vision, leveraging dedicated subsidiaries and customer-first financing to dominate South Korea’s next-gen digital infrastructure.
SK Telecom is aggressively positioning itself as the dominant player among South Korea's Big 3 telcos in the AI data center race, targeting a massive 15GW capacity by 2035 with phased 5GW openings starting in 2029. This ambitious scale dwarfs competitors KT and LG Uplus, who aim for 1GW and 600MW-plus respectively, underscoring SK Telecom’s commitment to lead the market through large clustered hubs in regions like Yeongnam and the southwest. To manage the immense capital demands of this expansion, SK Telecom has established SK Hyper, a dedicated subsidiary responsible for site acquisition, infrastructure development, and customer acquisition, supported by a planned capital injection of up to 750 billion won by 2030.
To mitigate the heavy upfront investment required for its AI data center projects, SK Telecom employs a customer-first financing model that secures long-term contracts before inviting strategic and financial investors or project financing partners. This approach reduces direct capital burden and aligns with the company’s strategy to scale efficiently while maintaining operational control. Complementing this, SK Hyper adopts a dual business model offering both co-location facilities and fully integrated 'AI factories'—the latter including servers and enabling higher profitability and optimization of AI workloads, leveraging SK Group’s semiconductor expertise, particularly from SK hynix.
Further streamlining its AI data center ambitions, SK Telecom launched SK Horizon, a specialized infrastructure company spun off from SK Broadband, backed by a $2.2 billion investment partnership with KKR and a Korean private equity consortium. SK Horizon begins with eight operational data centers across South Korea and is actively expanding capacity and submarine cable infrastructure to support AI workloads. SK Telecom retains majority control with a 51% stake, while KKR and IMM hold significant minority shares, reflecting a strategic alliance to accelerate growth and secure funding in this fiercely competitive sector.
SK Telecom’s AI data center strategy capitalizes on South Korea’s unique advantages, including a stable power grid, advanced semiconductor manufacturing, and cooling technologies, positioning the country as a promising alternative to the U.S.-centric AI infrastructure market. By integrating group synergies such as SK hynix’s memory production, SK Telecom emphasizes memory’s critical role in AI token output beyond GPU capacity, aiming to build comprehensive 'AI factories' that optimize computing, memory, and software. This strategic positioning not only addresses the monumental investment challenge—estimated at 500 trillion won—but also leverages South Korea’s potential as a global AI test bed amid growing regulatory and grid constraints elsewhere.
Private Equity Fuels Expansion
By spinning off SK Horizon and selling a 49% stake to KKR and IMM, SK Telecom secures billions in growth capital while retaining control over the country’s largest AI data center platform.
SK Telecom's strategic financial maneuvering is epitomized by the spin-off of SK Horizon from SK Broadband, creating a dedicated AI data center infrastructure company backed by a substantial $2.2 billion equity investment from global investment giant KKR and the Korean private equity consortium IMM. This restructuring not only streamlines governance to accelerate decision-making and execution but also allows SK Horizon to focus sharply on expanding AI data center capacity, starting with eight operational centers and an ambitious target of 318MW. By isolating AI infrastructure operations under SK Horizon, SK Telecom positions itself to deploy capital more effectively and scale aggressively in the burgeoning AI data center market.
The ownership structure of SK Horizon reflects a carefully balanced strategic partnership: SK Telecom retains majority control with a 51% stake, while KKR and IMM hold significant minority shares of 29% and 20%, respectively. This arrangement injects critical external capital while preserving SK Telecom’s decisive influence over the company’s direction, enabling the telco to leverage private equity resources without ceding control. The $2.2 billion infusion from selling 49% of the AI data center business to these financial partners underscores SK Telecom’s commitment to accelerating its AI infrastructure growth through collaborative investment rather than solely relying on internal funding.
AI Factories and Talent Pipeline
SK Telecom’s proprietary AI models and industry alliances are driving a shift from research to real-world AI execution, backed by a talent pipeline built through targeted university partnerships.
SK Telecom is aggressively scaling its proprietary AI foundation models, exemplified by the A.X K2 model with 688 billion parameters, which delivers top-tier performance in South Korea, notably achieving 97.1% accuracy on the Math Arena AIME 2026 leaderboard and excelling in long-context understanding and agent-related tasks. This technological leap extends beyond text to multimodal AI capabilities, including image and speech processing models like A.X K2 VL Light-Preview and A.X K2 ALM, enabling diverse industrial applications from call centers to manufacturing and defense sectors, where AI agents are actively deployed in companies such as KG Steel and Conec as well as military branches.
To sustain competitive advantage, SK Telecom is pioneering an 'intelligence improvement loop' that tightly integrates real-world service data and user feedback from its AI platform 'A.'—which boasts over 10 million monthly active users—back into continuous model retraining and system enhancements. This approach is supported by a sophisticated mixture-of-experts architecture that dynamically routes queries between lightweight and large models to optimize inference costs and performance, while rigorous safety protocols including ongoing red-teaming and real-time monitoring ensure reliability as AI penetrates critical services.
Shifting its AI R&D focus from pure model performance to actionable execution, SK Telecom orchestrates a consortium of about 20 organizations across sectors such as finance, mobility, media, and healthcare to link user intent with real-world AI-driven actions, embodying its vision of 'AI that executes.' This ecosystem development is bolstered by a strategic partnership with Seoul National University, where SKT leads a 15-week joint AI program emphasizing practical, industry-based knowledge and hands-on experience, reflecting CEO Jung Jae-heon's commitment to cultivating future AI talent and securing long-term competitive advantage amid intensifying talent competition.
SK Telecom’s AI ecosystem integrates model development, data infrastructure, semiconductor utilization, service verification, and industrial application into a unified system managed by an elite internal consortium, ensuring seamless collaboration across research, development, and commercialization. This holistic approach enables rapid iteration and deployment of AI technologies, positioning SKT not just as a telecom giant but as a central orchestrator in South Korea’s burgeoning AI landscape.
Energy and Investment Crossroads
Surging AI data center demand is forcing South Korea to rethink its energy grid and investment priorities, as skepticism rises over the feasibility and timing of massive tech infrastructure bets.
South Korea's major mobile carriers, including SK Telecom and KT, are strategically reallocating capital expenditure from traditional 5G network infrastructure toward expansive AI data center development, with SK Hyper targeting up to 15 gigawatts of capacity by 2035. This shift is driven by soaring demand for generative AI and high-performance computing, evidenced by SK Telecom’s 92.5% year-on-year surge in AI data center revenue, even as mobile service revenues decline. Nonetheless, investments in 5G standalone mode and future 6G technologies continue alongside this pivot, indicating a rebalancing rather than a reduction in overall CAPEX.
The explosive growth of AI data centers and semiconductor clusters is poised to reshape South Korea’s national energy landscape, with an estimated 260 TWh of additional annual power demand—nearly half the country's 2025 generation capacity—projected by 2030. This unprecedented surge necessitates a paradigm shift from the traditional 'industrial siting first, grid expansion later' model to prioritizing power supply capacity as a strategic national asset, as emphasized by Professor Lee Jong-su. However, renewable energy expansion alone, even with an ambitious 100 GW target by 2030, will fall short, requiring diverse carbon-free sources and urgent grid infrastructure upgrades to overcome spatial mismatches and project delays.
Despite robust government backing exemplified by a trillion-dollar semiconductor and robotics investment plan aiming to decentralize industry from Seoul, market skepticism looms large over South Korea’s tech infrastructure ambitions. Investors are retreating from high-growth tech sectors into defensive industries amid concerns over the sustainability and timing of these massive projects, with some viewing the enthusiasm as a 'collective delusion.' This cautious capital environment could complicate SK Telecom’s aggressive AI data center expansion, underscoring the tension between visionary state-led initiatives and market realities.
Current South Korean tax credit policies for AI data centers are misaligned with industry realities, as they favor direct equipment ownership and exclusive self-use, conflicting with the prevalent rental and shared service models of large-scale AI facilities. Industry leaders like SK Telecom's Lee Young-tak and Kakao's Kim Se-woong advocate for parity with semiconductor tax incentives, proposing to raise the credit rate from 15% to 20% and expand eligibility to include buildings and cooling infrastructure. Additionally, power constraints in the Seoul metropolitan area are fueling calls for integrated relocation policies that enable entire AI services to move regionally without sacrificing responsiveness, highlighting the need for comprehensive, service-level incentives.
