SK Telecom is carving its AI infrastructure business out of SK Broadband and bringing in KRW 3.08 trillion (about $2.2 billion) from KKR and the IMM Investment-Stonebridge consortium to accelerate the buildout. The new company, SK Horizon, will operate and expand AI data centers and submarine-cable infrastructure, initially targeting 318 MW of capacity. The restructuring comes as AI computing demand pushes data-center operators to compete for scarce power, capital and connectivity.
SK Telecom Spins Out AI Data Centers as Infrastructure Race Intensifies
The AI infrastructure business is becoming too capital-intensive and strategically important for many telecom operators to manage as a conventional business line.
SK Telecom (SKT) is responding by creating a dedicated AI data-center infrastructure company, SK Horizon, and bringing external investors into the business.
The South Korean telecommunications giant said August 27 that it will split its wholly owned subsidiary SK Broadband into the surviving SK Broadband and newly established SK Horizon. The transaction will separate the data-center and submarine-cable operations from SK Broadband’s fixed-line, media and enterprise businesses.
KKR and the IMM Investment-Stonebridge consortium have agreed to invest a combined KRW 3.08 trillion in SK Horizon. Once all phases of the transaction are completed, SKT will retain a 51% controlling stake, while KKR will hold 29% and the IMM consortium 20%.
The structure gives SKT a dedicated infrastructure vehicle with outside capital to fund expansion while preserving strategic control.
That model reflects a broader shift in the data-center industry. AI infrastructure requires enormous amounts of power, specialized computing equipment, cooling capacity and network connectivity. Financing those assets increasingly looks more like infrastructure investing than conventional telecom capex.
SK Horizon Will Start With 318 MW
SK Horizon is expected to oversee a portfolio totaling 318 MW of AI data-center capacity.
The portfolio includes eight operating facilities across South Korea, including locations in Seocho, Ilsan, Bundang, Gasan, Centum, Yangju and Pangyo, along with new facilities under construction in Ulsan and Guro. The company will also expand its submarine-cable infrastructure, which SKT describes as critical to serving globally connected AI businesses.
The distinction between data centers and network infrastructure is becoming less important as AI workloads become more distributed.
Large AI models require enormous volumes of data to move between computing clusters, cloud platforms and users. High-capacity terrestrial and subsea networks therefore become part of the AI infrastructure stack rather than a separate telecom concern.
SK Horizon’s combination of data centers and submarine cables gives SKT a way to control more of that stack.
The company will operate alongside SK Hyper, another SKT AI infrastructure business established in July. SKT plans to oversee the group’s overall AI data-center strategy and work with global technology companies, while SK Horizon focuses on operating and expanding infrastructure. SK Hyper is tasked with developing new large-scale projects, with a target of opening 5 GW in phases by 2029 and ultimately reaching 15 GW by 2035.
Those numbers illustrate the scale of SKT’s ambition.
AI Data Centers Have Become a Power-and-Capital Problem
The timing of the restructuring is not accidental.
AI data centers consume substantially more power than conventional facilities because of GPU-heavy workloads, high-density computing and increasingly demanding cooling systems.
Gartner estimates worldwide data-center electricity consumption will reach 565 TWh in 2026, up 26% from 2025. AI-optimized servers are expected to account for 31% of data-center power consumption this year, with their consumption surpassing conventional servers in 2027.
The bottleneck is increasingly not simply finding customers for computing capacity. It is securing electricity, land, grid connections and financing.
Gartner’s 2026 research identifies power availability as a constraint on AI capacity expansion, while CBRE says South Korea’s data-center market is experiencing increasing demand alongside supply constraints. Seoul metropolitan data-center supply is expected to exceed 1,450 MW by 2028, with power availability becoming an increasingly important factor in site selection.
That makes SK Horizon’s capital structure significant.
KKR is one of the world’s largest infrastructure investors, while IMM Investment and Stonebridge Capital bring South Korean investment expertise. Their participation gives SKT access to additional capital without surrendering control of the new infrastructure company.
KKR and IMM Signal Institutional Confidence
For KKR, the transaction fits into a global investment strategy around digital infrastructure and power assets.
The firm said it has more than $100 billion in infrastructure assets under management and more than $70 billion invested across digital and power assets. Its investment in SK Horizon is being made primarily through its Asia-Pacific infrastructure strategy.
The IMM-led consortium brings domestic institutional capital into the structure.
That combination is notable because AI infrastructure increasingly requires both global capital and local market knowledge. Data centers are physically tied to electricity grids, regulatory frameworks, land and connectivity infrastructure, making local operating expertise difficult to replace with purely financial capital.
SKT’s 51% stake also means the telecom company remains responsible for the strategic direction of the platform.
SKT Is Building an AI Infrastructure Stack
The restructuring is part of a larger attempt by SK Telecom to reposition itself around AI infrastructure.
The company will retain responsibility for the group’s overall AIDC strategy and partnerships with global technology companies. SK Horizon becomes the operating and infrastructure-expansion arm, while SK Hyper focuses on developing new gigawatt-scale projects.
That resembles the way large infrastructure ecosystems are increasingly being organized: one layer handles strategy and technology partnerships, another owns and operates physical assets, and another develops the next generation of capacity.
It also creates clearer investment boundaries.
AI infrastructure projects can require billions in capital before generating revenue. Separating those assets into a specialized company can make financing, partnerships and portfolio-level investment decisions easier to manage.
For SKT, the goal is not simply to build more server space. It is to position South Korea as a regional AI infrastructure hub.
That ambition faces serious competition.
Amazon Web Services, Microsoft Azure and Google Cloud continue to expand global cloud infrastructure, while companies such as NVIDIA supply the accelerators driving AI workloads. Telecom operators, infrastructure funds and specialist data-center developers are increasingly competing for the same power and land resources.
SK Horizon’s differentiation will therefore depend on more than its MW target. Its access to power, network connectivity, customers, AI computing partnerships and ability to deliver facilities on schedule will determine whether the platform can compete with larger global infrastructure ecosystems.
Submarine Cables Add a Strategic Dimension
The inclusion of submarine cables makes the strategy particularly interesting.
Subsea cables carry the overwhelming majority of intercontinental internet traffic, making them critical infrastructure for cloud computing, telecommunications and increasingly AI workloads.
For an AI infrastructure operator, connectivity can influence where workloads are hosted and how efficiently data moves between regions.
South Korea’s position as a major technology and semiconductor economy gives SKT a strategic base from which to pursue that infrastructure opportunity.
But it also puts SK Horizon within a much broader geopolitical discussion around digital sovereignty.
Governments increasingly want critical computing, connectivity and data infrastructure that is resilient to supply disruptions and geopolitical risks. South Korea’s AI infrastructure strategy therefore has implications beyond commercial data-center capacity.
What It Means for Enterprise AI Buyers
For enterprises, SKT’s restructuring is another sign that AI infrastructure is becoming a specialized market with its own capital and operating models.
Companies adopting AI will increasingly need to evaluate more than cloud pricing and GPU availability. Power reliability, geographic redundancy, network latency, data sovereignty and long-term capacity commitments are becoming strategic considerations.
The same dynamics are visible globally. Gartner expects data-center power demand to continue rising sharply through the end of the decade, while investors are increasingly treating data centers as infrastructure assets rather than conventional technology real estate.
SK Horizon’s creation does not guarantee that SKT will become a dominant AI data-center provider. But it shows how the infrastructure layer of the AI economy is evolving.
The companies building AI models may attract the headlines. Increasingly, the companies that secure the electricity, buildings, chips and networks underneath those models will determine how quickly the AI economy can scale.
Market Landscape
The AI data-center market is increasingly divided among several interconnected groups:
- Cloud hyperscalers: AWS, Microsoft Azure and Google Cloud continue to build and lease large-scale AI capacity.
- AI infrastructure specialists: Data-center operators and developers are expanding high-density facilities optimized for GPU workloads.
- Telecom operators: Companies such as SK Telecom are leveraging existing networks, data centers and connectivity assets to enter AI infrastructure.
- Infrastructure investors: KKR, Blackstone, Brookfield and regional investment firms are supplying capital for increasingly expensive data-center projects.
- AI hardware providers: NVIDIA and other semiconductor companies are driving demand for high-density compute infrastructure.
South Korea is becoming an increasingly important market within this competition. CBRE identifies constrained supply and secured power as major considerations for Korean data-center investment, while Gartner says power availability is becoming a global constraint on AI capacity expansion.
The strategic advantage may increasingly go to operators that can combine power, land, connectivity, capital and AI-compute expertise rather than simply offering physical data-center space.
Top Insights
- SK Telecom is spinning out SK Broadband’s data-center and submarine-cable operations into SK Horizon, creating a dedicated AI infrastructure company.
- KKR and the IMM Investment-Stonebridge consortium will invest KRW 3.08 trillion, giving SK Horizon capital to accelerate its AI data-center expansion.
- SK Horizon targets 318 MW across operating and developing facilities, while SK Hyper separately targets 5 GW by 2029 and 15 GW by 2035.
- Gartner forecasts global data-center electricity consumption will reach 565 TWh in 2026, underscoring the power constraints facing AI infrastructure developers.
- The restructuring gives SKT a three-part AI infrastructure model spanning strategy, data-center operations and large-scale project development.
Power Tomorrow’s Intelligence — Build It with TechEdgeAI
