Korea's third great infrastructure achievement after the expressway and the internet
At a moment when the world's appetite for artificial intelligence is outrunning the infrastructure built to serve it, SK Telecom has announced an ambition to reshape that imbalance — pledging to construct 15 gigawatts of AI data center capacity across South Korea, with the first phase arriving in 2029. The company frames this not merely as a commercial venture but as a civilizational project, placing it alongside the nation's expressways and broadband networks in the lineage of transformative infrastructure. Korea's mastery of high-bandwidth memory, its stable power grid, and its semiconductor heritage position it as a credible answer to a global shortage that analysts project will reach 15 gigawatts in the United States alone by decade's end. Whether ambition becomes architecture depends on the partners, tenants, and political will that SK Telecom can now draw to its side.
- Global demand for AI computing is growing at nearly 20 percent annually while supply falls dangerously behind, creating a window that SK Telecom is racing to fill before rival regional hubs claim it.
- The financial stakes are immense — each gigawatt of capacity carries a price tag near 70 trillion Korean won, forcing the company to stitch together capital from partners, long-term contracts, and project financing before a single server rack is installed.
- Korea's unique combination of world-leading memory chip expertise, nuclear-anchored grid stability, and SK Group's own semiconductor fabs gives the project a credibility that pure real-estate plays elsewhere cannot match.
- SK Telecom is threading the buildout through South Korea's national 'AI G3' strategy and a regional development agenda, deliberately routing data centers away from Seoul to spread economic weight across the country.
- The entire enterprise now hinges on securing anchor tenants among the world's largest technology firms — without them, the financing structure and the geopolitical bet underneath it begin to unravel.
In early July, SK Telecom announced it would build 15 gigawatts of AI data center capacity across South Korea — a project the company places in the same historical register as the Gyeongbu Expressway and the nation's high-speed internet rollout. The first 5 gigawatts, distributed across the southeastern and southwestern regions, are scheduled to come online in stages from 2029 onward.
The economics are formidable. A single gigawatt-class facility can cost roughly 70 trillion Korean won, driven by the price of high-performance chips and memory. SK Telecom plans to finance the effort through its own capital, strategic partnerships, long-term customer contracts, and project financing — a structure that signals an expectation of major international tenants from the outset. The timing is deliberate: McKinsey projects global data center demand growing 19 to 22 percent annually through the decade, with the United States facing a 15-gigawatt shortfall by 2030. Technology companies are already scouting locations beyond American borders, and Korea has emerged as a serious candidate.
Several structural advantages underpin the pitch. Korea leads the world in high-bandwidth memory, the specialized chips at the heart of AI systems. Its grid is stable, anchored by nuclear plants and LNG terminals. And SK Telecom's parent conglomerate, SK Group, already operates semiconductor fabs at gigawatt scale — giving the company decades of relevant industrial experience. President and CEO Jung Jai-hun has framed the project as a natural extension of existing capabilities, building toward an 'AI Factory' concept he outlined at an industry summit last November.
The buildout is also woven into South Korea's 'AI G3' national strategy, which aims to place the country alongside the United States and China among the world's top AI powers. The deliberate placement of facilities outside Seoul reflects a parallel government goal of balanced regional development.
SK Group is mobilizing its full infrastructure stack — semiconductors, energy, construction, and operations — with SK Telecom serving as architect and operator. The company has already collaborated with major global technology firms on smaller projects, laying early groundwork. But the scale of the ambition is unambiguous: 15 gigawatts is a generational bet, and its success depends entirely on whether SK Telecom can now secure the anchor tenants, the financing, and the sustained political will to carry it through.
SK Telecom announced in early July that it will build out 15 gigawatts of AI data center capacity across South Korea, a project the company frames as the nation's third great infrastructure achievement, after the Gyeongbu Expressway and the rollout of high-speed internet. The first phase—5 gigawatts spread across the southeastern and southwestern regions—is scheduled to come online in stages beginning in 2029. The company is positioning Korea as a computing hub for the world's largest technology firms at a moment when demand for AI training and inference capacity is outpacing supply globally.
The economics are staggering. A single gigawatt-class AI data center can cost approximately 70 trillion Korean won to construct, given the expense of high-performance chips and memory. SK Telecom plans to finance the project through a combination of its own capital, investment from strategic partners, long-term customer contracts, and project financing—a structure that suggests the company expects to attract major international tenants from day one. The timing is deliberate. McKinsey forecasts that global data center demand will grow between 19 and 22 percent annually through the end of the decade, while supply lags behind. The United States alone faces a projected shortfall of about 15 gigawatts by 2030. Global technology companies are already scouting locations outside America to build out their computing infrastructure, and Korea has emerged as a serious contender.
Several factors make Korea attractive. The country has built world-leading expertise in high-bandwidth memory, the specialized chips that power AI systems. Its power grid is stable, anchored by nuclear plants and liquefied natural gas terminals. And SK Telecom itself operates semiconductor fabrication plants at gigawatt scale, meaning the company has decades of experience managing the kind of industrial infrastructure that AI data centers demand. Jung Jai-hun, SK Telecom's president and chief executive, framed the project as a natural extension of the company's existing capabilities. At an industry summit last November, he outlined a roadmap to position SK Telecom as the nation's leading AI operator, with plans to launch an "AI Factory"—a next-generation data center concept—in 2027 and scale it upward from there.
The project is not purely commercial. SK Telecom has aligned the buildout with the South Korean government's "AI G3" strategy, an ambition to position the country among the world's three leading AI powers alongside the United States and China. The company is also threading the project through the government's broader goal of balanced regional development, deliberately placing data centers in the southeast and southwest rather than concentrating them in Seoul or the capital region. This regional distribution is meant to spread economic benefit and attract global investment beyond the traditional tech hubs.
SK Group, the sprawling conglomerate that owns SK Telecom, is mobilizing its full stack of AI infrastructure capabilities for the effort. The group holds the necessary pieces: semiconductors, energy solutions, and data center construction and operations expertise. SK Telecom will serve as the architect and operator, designing and running the facilities while coordinating with group affiliates that supply power, chips, and other critical inputs. The company has already been working with major global technology firms on smaller data center projects, laying groundwork for the larger buildout.
The scale of the ambition is worth pausing on. Fifteen gigawatts is not a modest expansion. It represents a bet that Korea can capture a meaningful share of the world's AI computing demand, that global technology companies will trust their most critical workloads to facilities in South Korea, and that the country's government and private sector can execute a multi-year, multi-billion-dollar infrastructure project without major disruption. Jung Jai-hun acknowledged the scope in his statement, saying the company would work closely with government, industry, and local communities to make Korea "Asia's core AI infrastructure hub." The first phase comes online in 2029. Everything that follows depends on whether the company can secure the anchor tenants, the financing, and the political will to see it through.
Citações Notáveis
This AI data center project is aimed at preemptively preparing the computing infrastructure that the global AI ecosystem needs. We will work closely with the government, industry, and local communities to help Korea grow into Asia's core AI infrastructure hub.— Jung Jai-hun, President and CEO of SK Telecom
A Conversa do Hearth Outra perspectiva sobre a história
Why does SK Telecom think Korea can win this? There are data centers everywhere now.
Korea has something specific: it makes the memory chips that power AI systems better than almost anyone else. And the power grid is stable. But the real advantage is that SK Telecom already runs semiconductor fabs at this scale. They know how to manage gigawatt-level infrastructure.
So this isn't really new technology. It's about having the right pieces already in place.
Exactly. The technology is proven. What's rare is having a company that can coordinate semiconductors, power, and operations all at once, in a country with stable governance and energy supply.
The government is involved. How much of this is actually about business versus national strategy?
Both. SK Telecom frames it as Korea's third great infrastructure project, after the expressway and the internet. That's not accidental language. They're saying this matters to the nation, not just to shareholders.
And the timeline—2029 is three years away. That's fast for something this big.
It is. But they're not starting from zero. The Ulsan facility is already under construction. They're scaling something that exists, not inventing it.
What happens if they can't find the customers? Fifteen gigawatts is a lot of capacity to fill.
That's the real risk. They're betting that global tech companies will want to put their AI workloads in Korea. If that doesn't happen, the economics fall apart. But they're already talking to major firms, so they're not going in blind.