삼성전자와 SK그룹이 현장 준비 진척도가 미미하다고 밝히면서, 오픈에이의 스타게이트 프로젝트가 한국에서 실행 지연을 겪고 있다.
Nearly a year after the announcement of “Stargate Korea,” OpenAI, Samsung, and SK have yet to finalize site selection, power supply arrangements, or funding structures. The initiative, part of a broader partnership announced last October, aims to build two dedicated AI data centers—one led by Samsung SDS and the other by SK Telecom—but construction and financial frameworks remain unresolved. This domestic stagnation coincides with similar setbacks across OpenAI’s global Stargate program, where several international projects face downsizing or outright cancellation.
The effort stems from letters of intent tied to a larger agreement among the three companies, under which Samsung Electronics and SK hynix will supply memory chips to power OpenAI’s Stargate facilities. The program envisions a network of global data centers designed to deliver the computing capacity required to train and operate OpenAI’s artificial intelligence models. During a press briefing on Oct. 1, 2025, following a meeting between OpenAI CEO Sam Altman and President Lee Jae Myung—attended by Samsung Electronics Executive Chairman Lee Jae-yong and SK Chairman Chey Tae-won—then-presidential chief of staff Kim Yong-beom stated: “Altman said he wants to return the benefits of this important partnership to Korea, which is why the companies are investing in AI data centers — one each in South Jeolla and in Pohang.” He added that Altman “framed the continued infrastructure investment as a way of returning part of the benefit Korea has given them.”
Despite the high-profile announcements, corporate responses indicate significant uncertainty. An OpenAI spokesperson declined to comment on the investment structure, the company’s direct involvement, or construction progress, stating only that the domestic project remains “still under discussion.” Samsung SDS, the IT services division managing the data center build-out alongside OpenAI, noted that “nothing has been confirmed” regarding the project and emphasized that final decisions rest with OpenAI. Similarly, SK Telecom reported no advancement since the October memorandum of understanding, clarifying that OpenAI’s requirements have not yet been finalized.
Industry sources point to logistical hurdles as a primary bottleneck. One anonymous executive noted that securing adequate power and water supplies remains the most critical requirement for modern data centers. “Acquiring power and water supply are the two key conditions to be met for data centers that are being raised most often these days,” the source said. “Those take up a huge amount of time, and it is understood that companies are still scouting for a proper site. There’s nothing concrete with OpenAI — at most, each company is approaching the project on its own, independently of the other.”
These developments stand in contrast to earlier domestic reports published last October, which claimed Samsung and SK were actively constructing facilities in Pohang and the southwestern region, respectively, with initial capacities slated at 20 megawatts each. The current stalemate also contradicts remarks made by Minister of Science and ICT Bae Kyung-hoon in February. When questioned about the project’s status during a committee session, Bae stated: “I understand it is currently underway,” adding that “a targeted timeline for groundbreaking [was] set in March.”
Officially, however, OpenAI has not formally designated Korea as part of its global Stargate rollout. According to the company’s October 2025 report, “AI in South Korea: OpenAI’s Economic Blueprint,” OpenAI uses language such as “explore” and “assess” to describe potential expansions with Samsung SDS and SK Telecom. While domestic media and SK have characterized the effort as “Stargate Korea”—with SK pledging to jointly build “an OpenAI-dedicated AI data center in the southwestern region of Korea, realizing Stargate Korea”—the overall trajectory remains unclear. Globally, Korea’s slow pace mirrors broader challenges within the Stargate initiative. Originally pitched as a $500 billion U.S. national project targeting 10 gigawatts of capacity by 2029, OpenAI expanded the concept to the United Arab Emirates, Norway, Britain, Argentina, and India over the past year. However, facilities in Norway and Britain were paused earlier this year due to unfavorable energy pricing and regulatory concerns. In Norway, OpenAI and Nscale failed to finalize an offtake agreement, leading Microsoft to assume control of the project.
The push for AI infrastructure is largely driven by competition among U.S. hyperscalers, including Amazon, Microsoft, Google, and Meta, with OpenAI operating under similar pressures. Securing training compute capacity has become a top priority across the sector, yet OpenAI’s aggressive expansion timeline faces growing scrutiny over financial feasibility. Internally, the term “Stargate” has reportedly lost its strategic significance after being popularized by OpenAI’s policy and public relations teams as a catchall label for various infrastructure deployments, according to the Financial Times.
Meanwhile, South Korea continues to advance its own ambitious AI infrastructure roadmap. The government plans to invest more than 1 quadrillion won ($739.67 billion) toward data center expansion, targeting a total capacity of 18.4 gigawatts by 2035. The initiative aims to decentralize development away from the greater capital region, with SK, GS, and Naver leading the effort as part of three so-called megaprojects proposed by the Lee Jae Myung administration as a cornerstone of national economic growth.
BY LEE JAE-LIM [lee.jaelim@joongang.co.kr]