SK Telecom’s revenue from AI data centers fell 7.4% from the previous quarter to 126.1 billion won in the third quarter of 2026, according to the investor briefing the company published on 4 November. SKT says the decline was “due to a gap in GPUaaS revenue during the transition to new long-term contracts”.
In the same quarter, SK Hyper, the AI data center unit SKT launched in July, recorded 148.6 billion won of data center development capex, which the briefing lists separately from the 461.4 billion won of capex at SKT and SK Broadband. By our arithmetic, that is 1.18 times the quarter’s AI data center revenue.
The key insight: By our arithmetic, SK Hyper’s 148.6 billion won of data center development capex in the third quarter was 1.18 times SK Telecom’s AI data center revenue of 126.1 billion won, which fell 7.4% from the second quarter, per the company’s briefing.
Consolidated revenue was 4.4973 trillion won, up 13.1% from a year earlier and 3.2% from the second quarter. Operating income was 552.3 billion won, against 48.4 billion won a year earlier and 566.0 billion won in the second quarter. Net income was 411.8 billion won, and SKT declared a quarterly dividend of 830 won per share.
The release says revenuegrowth was “driven by growth in the company’s wireless and fixed-line telecommunications businesses”. Mobile service revenue rose 21.6% year on year to 2,569.7 billion won, which the briefing says reflects “the base effect of last year’s Customer Appreciation Package” and a shift to customer lifetime value-centric marketing. Fixed-line revenue grew 7.4%.
The same totals appear in SKT’s preliminary fair disclosure on Korea’s DART system, filed on 4 November in units of 100 million won. The filing says the figures are unaudited K-IFRS results that may change in the external audit.
SK Telecom’s two AI revenue lines by quarter, KRW billion, from its investor briefings for Q1, Q2 and Q3 2026 (earlier quarters from their comparison columns; unaudited). AI DC revenue fell 7.4% quarter on quarter to 126.1 billion won. The 1.18x ratio of SK Hyper’s data center development capex to AI DC revenue is our arithmetic.
The AI Lines
SKT reports two AI revenue lines. AI DC revenue, at 126.1 billion won, was up 6.2% from a year earlier, which the briefing attributes to “higher DC utilization rates and submarine cable traffic growth”. AI B2B/B2C revenue rose 27.3% year on year and 14.2% from the second quarter to 70.0 billion won, “thanks to orders for Cloud business”.
By our arithmetic, AI DC revenue was 2.8% of consolidated revenue, and the two AI lines together, 196.1 billion won, were 4.4%. GSMA Intelligence put AI at about 2.5% of telecom operators’ revenue in its third-quarter tracker, as we reported on 2 November; the two measures are defined differently.
Across the seven quarters in SKT’s last three briefings, AI DC revenue went from 69.4 billion won in the first quarter of 2025 to 136.2 billion won in the second quarter of 2026. The third quarter is the first quarter-on-quarter decline in that series. In its first-quarter briefing, SKT credited AI DC growth partly to “revenue contribution from GPUaaS”.
The Build-Out
The release says SKT set up SK Hyper in July 2026 to lead AI data center business development and SK Horizon in August 2026 to run and expand the infrastructure. It says SK Horizon is developing new data centers with a capacity of approximately 180 MW in Ulsan and Seoul.
The briefing lists 137 MW across eight data centers today and 318+ MW from 2030. The release says SK Hyper “is in discussions with potential customers on business terms, with a focus on Ulsan”, and that once agreements are finalized the company plans to accelerate development on sites it has secured.
“Built on our two pillars of telecommunications and AI, we were able to achieve both recovery and growth at the same time,” said Park Jong-seok, SKT’s chief financial officer, in the release.
Business Pill · Backlog
An evergreen Business Engineer explainer on backlog: revenue that is signed but not yet earned. It teaches the general idea only and does not discuss SK Telecom.
How We Counted
All figures are from SKT’s English investor briefings for the first, second and third quarters of 2026, its 4 November press release and its DART filing. Earlier quarters come from the comparison columns of the later briefings. Amounts are in won as SKT reports them; we have not converted them.
The 1.18 ratio divides SK Hyper’s 148.6 billion won of development capex by 126.1 billion won of AI DC revenue. The shares divide each AI line by 4,497.3 billion won of consolidated revenue.
The Structural Read
As we read the briefing, SKT’s AI data center revenue has two parts that moved differently this quarter: the data centers and submarine cable, which the briefing says grew on utilization and traffic, and GPUaaS, whose revenue had a gap while customers moved to new long-term contracts.
As we read the release, construction and customer contracts are at different stages: SK Horizon is developing about 180 MW in Ulsan and Seoul, while SK Hyper says it is in discussions with potential customers on business terms, with a focus on Ulsan.
SK Telecom, Q3 2026 investor briefing
“Down QoQ due to a gap in GPUaaS revenue during the transition to new long-term contracts”
Three Implications
FOR GPU CLOUD BUYERS SKT’s briefing says its GPUaaS revenue had a gap during a move to new long-term contracts; it does not give their length or value.
FOR TELECOM INVESTORS The release attributes the quarter’s revenuegrowth to wireless and fixed-line; the briefing ties the 21.6% rise in mobile service revenue partly to the base effect of last year’s Customer Appreciation Package.
FOR DATA CENTER DEVELOPERS The briefing lists 137 MW across eight data centers today and 318+ MW from 2030, with new capacity in Ulsan and the Seoul metropolitan area.
The Business Engineer Lens
This story maps onto the Business Engineer framework The AI Supercycle.
The framework describes its fifth layer, Compute Capacity, as “Hyperscaler GPU clusters, neoclouds, custom-silicon estates.”
As we read it, SKT sits in that layer through its data centers and GPUaaS, and this quarter’s GPUaaS gap shows revenue there following contract timing. The framework is the owner’s argument and does not discuss SK Telecom.
The briefing does not name the GPUaaS customers, say how long or how large the new long-term contracts are, or say when the revenue gap closes. It does not report profit for the AI lines.
The 148.6 billion won is for data centers under development, not the ones that earned the quarter’s revenue, so the ratio compares one quarter’s build spending with one quarter’s sales; it is not a return. SKT’s earnings call was scheduled for 4 p.m. Seoul time on 4 November, after this piece was written.
This Week In The AI Supercycle
Anthropic’s Reality Check, The Capital Squeeze, & The New World Order
This week’s issue: for two years the AI build-out was priced as a growth story. This week it started being priced as a credit story.
SKT’s AI data center revenue fell 7.4% in the quarter, which the company puts down to GPUaaS customers moving to new long-term contracts. Its new data center unit recorded 148.6 billion won of build capex, 1.18 times that revenue by our arithmetic, while the release credits the quarter’s growth to wireless and fixed-line.
95,000+ executives read Business Engineer for the AI strategy frameworks cited by ChatGPT, Claude, and Perplexity.
A note on sourcing. The figures are from SK Telecom’s Q3 2026 press release, its investor briefings for the first three quarters of 2026 and its DART filing of 4 November; they are preliminary and unaudited. The ratios are our arithmetic.
Gennaro Cuofano is a CRO and tech executive who has worked in AI since late 2015, starting with bringing NLP, natural-language generation, voice and chatbot products to the marketing industry. His background is in law and finance: he holds a Master's degree in Law and an International MBA with an emphasis on corporate finance (LUISS Business School and the University of San Diego, 2012), and worked as a financial analyst at a real-estate investment firm in San Diego and as an assistant controller. His work focuses on business model strategy, business engineering and, more broadly, structural analysis: how companies make money, read from their own filings. He created FourWeekMBA and leads research there and at The Business Engineer, his newsletter on AI and business strategy, with over 95,000 subscribers and more than 1,000 published analyses. His writing has also appeared on Entrepreneur (for example, How to Digitize a Family Business in Europe) and HackerNoon. Find him on LinkedIn and Substack. See how we source.
