Why it matters
The strategic read
Moves Korea from a gigawatt AI-cloud partnership headline into a national-scale buildout roadmap. SKT is framing telecom networks, HBM memory, nuclear and LNG-backed power, data-center construction, operations and project finance as one sovereign AI infrastructure stack. The important signal is not a single signed tenant lease yet; it is that SKT is openly planning 5 GW of staged capacity from 2029 and a 15 GW long-term target, making Korea another candidate hub for hyperscaler and frontier-model compute outside the U.S.
Source-supported terms
What was disclosed
- 01
SK Telecom announced on July 5, 2026 that it will pursue construction of AI data centers with a scale of up to 15 GW.
- 02
SKT said the goal is to make Korea an AI infrastructure hub in Asia and link the buildout to the Korean government's AI G3 strategy and balanced regional development tasks.
- 03
SKT said it is reviewing core infrastructure elements including power, siting and operations.
- 04
The release says constructing a typical 1 GW-class AI data center may require project cost of approximately KRW 70T because of high-performance AI computing infrastructure and rising memory prices.
- 05
SKT said such costs are expected to be financed through company investment, strategic partner investment, long-term customer contracts and project financing.
- 06
SKT plans to start with the AI data center currently under construction in Ulsan.
- 07
SKT plans to build more than 2 GW across the southeastern Gyeongsang region and an additional 1 GW in the southwestern Jeolla region.
- 08
The domestic plan targets 5 GW of AI data-center capacity opened in stages starting in 2029.
- 09
The Korean SKT newsroom says SKT will expand sequentially toward 15 GW by 2035 while considering demand and investment conditions.
- 10
SKT said it will lead design, construction and operation as the AI infrastructure architect while SK Group affiliates contribute semiconductor, energy-solution and data-center capabilities.
- 11
SKT cited Korea's HBM position, nuclear and LNG power supply conditions, and GW-class semiconductor-fab operating experience as reasons Korea can attract global big-tech AI infrastructure demand.
- 12
The release points to SKT's existing AWS-linked Ulsan AI data center and its NVIDIA AI Factory / DSX collaboration as related groundwork.
Disclosure boundary
What is known—and what is not
Primary SK Telecom Korean newsroom announcement plus English PRNewswire release. Public materials disclose the named sponsor, 15 GW long-term target, 5 GW domestic staged-opening plan from 2029, Ulsan starting point, more-than-2-GW Gyeongsang cluster, 1 GW Jeolla plan, financing categories, KRW 70T estimated cost for a typical 1 GW-class AI data center, SKT's design/build/operate role and SK Group full-stack AI infrastructure framing. Public materials are explicitly forward-looking and do not disclose final investment decision, signed anchor tenants, project-finance commitments, power-service agreements, utility tariffs, site-by-site permits, EPC awards, construction packages, GPU procurement, cloud-service contracts, customer remedies or binding delivery milestones for the full 15 GW plan.