The scale of the project keeps growing. SK Group’s artificial intelligence data center under construction in Ulsan, South Korea, has been expanded to nearly 900 megawatts of capacity, Chey Tae-won, the conglomerate’s chairman, said on September 13. The figure is larger than what had been planned and underscores how quickly the project is scaling.
The Ulsan site is a joint effort between SK Group and Amazon Web Services, the cloud unit of Amazon. The two companies announced the project as a flagship of South Korea’s push into AI infrastructure, with an expected investment of 7 trillion won, or about $5.2 billion. Operations are planned for the second half of 2027.
Chey said the group will announce more partnerships with global technology companies as the project moves forward, without naming them. The comment signals that SK expects the data center to draw a broader set of tenants and partners in addition to Amazon, each bringing workloads and capital of their own.
The nearly 900 megawatts would make the Ulsan facility one of the largest AI data centers in Asia. For comparison, many of the hyperscale data centers being built around the world today run in the tens to low hundreds of megawatts. A site approaching a gigawatt represents a different order of industrial commitment.
South Korea has been aggressive about courting AI infrastructure. The government has eased permitting, offered incentives and framed data centers as a strategic industry in a country whose economy depends on semiconductors and exports. The Ulsan project fits that policy, pairing a domestic conglomerate with a leading American cloud provider.
The investment also reflects a shift in how data centers are financed. Rather than building capacity and waiting for tenants, developers are now signing anchor customers first, often in partnership with the cloud providers that will occupy the space. SK’s arrangement with AWS follows that model, with the two sides sharing the risk and the returns.
Power is the binding constraint, and South Korea has it in limited supply. Data centers at this scale require reliable, high-voltage electricity, and the country’s grid has been stretched by industrial demand. The Ulsan project is expected to draw substantial power, and its size is partly a function of what the local grid and new generation can support.
The 900 megawatt figure matters because capacity, once built, is hard to move. A data center is a bet on demand that will materialize years after the investment is made. SK’s decision to expand before the first building is finished suggests the company has commitments, or at least strong signals, that justify the larger footprint.
Analysts said the expansion reflects the same dynamic visible across the industry: the largest AI operators are securing capacity years in advance, and developers are racing to meet that demand. The result is a cycle in which announced capacity keeps rising, even as questions grow about whether it will all be used.
The timeline adds to the pressure. A 2027 opening means the project must be finished while the AI market is still deciding how much compute it needs. If demand softens, SK and AWS will be left holding capacity they contracted for years earlier. If demand holds, the early expansion will look prescient.
Chey’s broader message was confidence. By expanding the project and promising more partnerships, he signaled that SK sees AI infrastructure as a durable business rather than a cyclical one. The company has been reshaping its portfolio around energy and technology, and the data center is central to that strategy.
The partnership with AWS gives the project a tenant of last resort. Amazon has committed to large volumes of cloud capacity and is expanding its footprint across Asia, where data sovereignty rules increasingly require in-country infrastructure. A facility in South Korea backed by a local partner helps AWS meet that requirement.
The question is who else fills the space. SK has not said how much of the capacity AWS will take or which other customers might sign on. The partnerships Chey hinted at could include chipmakers, cloud providers or sovereign funds, each of which would bring different terms and a different mix of demand.
South Korea’s competitive position is part of the appeal. The country sits near the semiconductor supply chain, has skilled engineering labor and offers proximity to Asian markets. Those advantages have made it a target for data-center investment, and the Ulsan project is the largest single expression of that so far.
What remains to be seen is whether the financing holds. Projects of this scale require debt and equity arranged years ahead, and the recent volatility in AI-related stocks has made capital more expensive. People familiar with the project said the funding is in place for the current phase, with later phases to be financed as demand materializes.
The expansion, in short, is a statement of intent. SK Group is building for a future in which AI capacity is measured in gigawatts rather than megawatts, and it is committing capital now to be positioned when that future arrives.
