Share with your CTO
SK Telecom is betting Korea can become Asia’s anchor for AI compute capacity, announcing plans to build up to 15 gigawatts of AI data center capacity by 2035, starting with 5GW online by 2029. The first facility, a hyperscale build in Ulsan developed alongside AWS, targets operation in the second half of 2027. A single 1GW-class AI data center now runs roughly 70 trillion won in project costs, so SKT is structuring this through project financing and long-term customer contracts rather than pure balance-sheet spending. SK Group affiliates covering semiconductors, energy, and construction are being folded into the architecture.
What this means for your business
If your AI infrastructure planning still treats the United States and Western Europe as the only serious compute markets, SKT’s move is a direct challenge to that assumption. McKinsey projects a 15GW supply shortfall in the US alone by 2030, and that gap is exactly what SKT is positioning to absorb from the Asian side. CTOs at multinationals who have deferred thinking about Asia-Pacific inference capacity, disaster recovery, or data residency because no credible hyperscale alternative existed there are running out of time to treat it as a future problem.
The financing structure deserves more attention than the headline gigawatt number. SKT is explicitly planning to fund a significant portion through long-term customer contracts, which means the company needs anchor tenants before 2027 shovels hit the ground at scale. That flips the usual dynamic: SKT isn’t just building supply and waiting for demand, it’s offering early partners preferential capacity in a market that will be structurally undersupplied. Any CTO whose organization is currently negotiating hyperscaler capacity in the Asia-Pacific region is negotiating against a future where SKT-anchored alternatives exist, and that changes the leverage in those conversations today, not in 2029.
The deeper signal here is what SKT’s “AI infrastructure architect” framing actually means. By integrating SK Group’s semiconductor, energy, and construction capabilities under a single design and operations umbrella, SKT is constructing a vertically integrated stack that hyperscalers like AWS typically control internally. The AWS partnership on the Ulsan facility is interesting precisely because it suggests even AWS sees value in SKT’s local integration rather than pure organic build-out. If that model proves out, the leading indicator to watch isn’t the gigawatt count but whether a second major hyperscaler signs on before 2027, which would confirm demand is real and shift the competitive question from “will this get built” to “who gets locked out.”
Based on reporting from SKT to Build 15GW AI Data Center, Targeting Asia’s AI Infrastructure Hub, originally published 2026-07-04 21:00:00.
