What happened
The Korea Times reports that SK Group Chairman Chey Tae-won called for business models that can generate enough AI revenue to sustain further investment. Speaking at Ulsan Forum 2026, Chey warned that the sector could face an investment bubble if capital spending does not produce sufficient returns. He also identified speed, scale and safety as pillars of SK's AI strategy and said the group's AI data center in Ulsan is planned to reach nearly 900 megawatts.
According to The Korea Times, Chey Tae-won made the comments Friday at Ulsan Forum 2026, SK Group's annual gathering in Ulsan, South Korea. He said AI needs better monetization models and should eventually reach a structure in which profits finance additional investment.
The report says Chey warned of a possible AI investment bubble if the large amounts of capital and resources flowing into the sector fail to generate adequate returns. He described speed, scale and safety as key pillars of SK's AI strategy.
Chey also said industrial AI requires large volumes of data that individual companies or regions cannot secure alone, arguing that a nationwide effort is necessary to achieve economies of scale.
The Korea Times reports that SK's AI data center under construction in Ulsan, jointly developed by SK Group and AWS, has expanded to nearly 900 megawatts. The project is estimated at 7 trillion won, or $5.2 billion, and is scheduled to begin operations in the second half of 2027. The report did not name future partners or provide a detailed commercialization plan.
Source details: koreatimes.co.kr ↗
Why it matters
The remarks highlight a central constraint on the AI buildout: large investments in computing, data centers, energy and industrial deployment need durable revenue rather than continuing solely on expectations of future demand. SK's comments are significant because the company is tying AI commercialization to manufacturing scale, nationwide data access and infrastructure expansion. The report does not establish whether SK's plans will be profitable, whether the stated capacity is fully funded, or whether the project will meet its schedule.
The comments put financial sustainability alongside technical performance and safety as a condition for continued AI expansion. That matters practically because data centers and industrial AI systems require substantial infrastructure before customers can generate measurable returns.
The Ulsan project illustrates the scale of the infrastructure question. A facility approaching 900 megawatts would require major decisions about power, construction, customers, networking and operations, but the report does not independently verify the capacity or the project's financial assumptions.
SK's emphasis on manufacturing data suggests that industrial AI adoption may depend on cooperation across companies and public institutions rather than isolated deployments. The report provides no evidence yet of specific production results, customer commitments or realized efficiency gains.
All project details and statements here are attributed to The Korea Times. No independent test of the facility, confirmation from AWS, or public financial documentation is provided in the source.
What to watch next
The Korea Times says SK plans to announce additional partnerships with global technology companies as the Ulsan project progresses. The reported facility, being developed with AWS at an estimated cost of 7 trillion won, is scheduled to begin operations in the second half of 2027. Key unknowns include the identity and commitments of future partners, the services and customers that will support the business model, the project's energy arrangements, and whether the nearly 900-megawatt capacity represents an immediate operating target or a longer-term buildout.
The next concrete signal will be whether SK identifies additional technology partners and specifies their financial, infrastructure or customer roles.
The project is reportedly scheduled for operations in the second half of 2027, but the source does not provide milestones, procurement status, power sourcing details or evidence that the schedule is secured.
Observers should look for disclosed customers, pricing, utilization targets and revenue expectations. Those details would show whether the proposed data center has a self-sustaining business model rather than only a large capacity target.
It is also unknown how SK will obtain and govern the industrial data Chey described as necessary for scale, or how safety requirements will be implemented in manufacturing deployments.