SK Telecom’s operating profit rose 1,041.3% in the third quarter from a year earlier, a multiple that says less about the three months to September than about the quarter it is being compared with.
The South Korean carrier reported operating profit of 552.3 billion won, or about $411.4 million, for July through September. Revenue rose 13.1% to 4.49 trillion won. Net income was 411.8 billion won, against a loss of 166.7 billion won a year earlier.
The percentage is a base effect. A gain of that size implies operating profit of roughly 48 billion won in the same quarter of last year, a level low enough that almost any recovery would produce a four-digit comparison. The absolute figure is the one to read: 552.3 billion won on 4.49 trillion won of revenue works out to an operating margin near 12%, a respectable number for a carrier and a better one than the headline multiple conveys.
The company attributed the quarter to steady growth in its core mobile service business and to new businesses built around artificial intelligence. Sales at its AI data centers rose 6.2% from a year earlier, while other AI-related businesses grew 27.3%. The company did not break out the revenue those units contribute, which leaves the reader to judge whether the growth rates describe a business of consequence or a small base expanding quickly.
That ambiguity matters because the mobile market in Korea is saturated. Three carriers serve a population of about 52 million, handsets are sold on multi-year installment plans and regulators have pushed down the cost of plans. Revenue growth in a market like that has to come from something other than selling more connections, and SK Telecom has spent several years arguing that the something is AI.
The strategy has three visible parts. The company sells data center capacity to enterprises and to model developers, which is the business growing 6.2%. It sells AI services to consumers and companies, the category growing 27.3%. And it runs the network those services depend on, which gives it a cost advantage in distribution and a disadvantage in capital intensity.
Data centers are the part with the clearest economics and the longest lead time. A carrier already owns land, fiber and the electrical capacity that a server hall needs, and it employs the engineers who keep a network running. What it does not own is the chips. A GPU cluster has to be bought years before a customer’s workload arrives, and the depreciation schedule starts on delivery. That is why carriers entering the business tend to report rising revenue and flat profit in the first years, then a step change once the capacity is fully used.
The Korean market has an additional feature that helps and hurts. The country has a dense concentration of semiconductor manufacturing, electronics and gaming companies within a short distance of Seoul, which gives a domestic data center a customer base that does not need to be recruited from abroad. It also means competition from the same companies, several of which build their own capacity.
Net income tells a different story from operating profit. The swing from a loss of 166.7 billion won to a profit of 411.8 billion won is larger than the operating improvement alone would explain, which suggests the year-earlier quarter carried items that did not repeat. Companies in the sector often book impairments, spectrum costs or restructuring charges in a single quarter, and a comparison across a year is only useful once those are stripped out. The company’s release does not itemize them, and analysts will ask on the call.
For investors, the question is whether AI revenue can grow fast enough to change the shape of the consolidated numbers. At 6.2% growth, the data center business is expanding more slowly than the other AI lines, which is unusual for a category that is supposed to be capacity constrained, and it may reflect the timing of when new halls come into service. The 27.3% growth in other AI businesses is the faster number but the smaller base.
The stock has traded for years on the assumption that a Korean carrier is a utility with a dividend, and any re-rating depends on evidence that the AI businesses can earn returns above the cost of the capital they consume. One quarter of a four-digit profit increase is not that evidence. A year of margins near 12% with AI revenue compounding at these rates would be.
There is a version of this story in which the carrier becomes an infrastructure company. Land, power, fiber and engineering talent are the scarce inputs in AI computing, and SK Telecom has all four in a country that is building models and chips of its own. There is another version in which the AI businesses remain a modest adjunct to a mobile operation that grows with inflation, and the profit multiple is remembered as an artifact of a weak comparison.
The fourth quarter will not settle it either. What would settle it is disclosure: revenue and operating profit for the data center unit on its own, capex attributed to AI capacity, and a utilization figure. Until the company publishes those, the elevenfold increase stands as an impressive number attached to a business whose new engine is still smaller than the old one.
