A telco cash cow chasing AI proof
- The core business is still mobile and fixed-line telecom, with 2024 cellular services at 74.2% of revenue.
- The AI push is real, with AIX revenue of KRW 193 billion last year and AIDC revenue of KRW 397.4 billion.
- The 2025 data leak hurt hard, but Q1 2026 showed about 210,000 net handset subscriber adds and dividends resumed.
- The next big test is whether the A. AI agent can turn free users into paid users in H1 2026.
- The bear case is simple: subsidy wars and AI spending could eat the cash that funds the whole plan.
Stable again, but not proven
SK Telecom has two stories at once. One is a mature telecom business that still pays the bills. The other is an AI buildout that could change how investors value the company if it proves profitable.
The bull case has improved since the 2025 cybersecurity shock. Management said MNO, its mobile network business, added about 210,000 handset subscribers in Q1 2026. It also restarted quarterly dividends. That is a strong sign that customer trust is coming back.
AI is the upside path. SK Telecom is building AI data centers, selling GPU-as-a-Service, which means renting out powerful AI chips, and pushing enterprise AI through AIX. Its A. personal AI agent is set to test paid subscriptions in H1 2026, while A.X K1, a model with more than 500 billion parameters, anchors the business AI pitch.
The hard part is profit. AI data centers need heavy spending before they earn back cash. The A. agent has users, but paid demand is still unproven. Meanwhile, the old telecom business can still be hurt by subsidy fights after the handset subsidy law was abolished.
Subscriptions fund the AI bet
Most money comes from monthly telecom bills. Customers pay for 5G mobile plans, broadband internet, IPTV, cable TV, and business communications. This gives SK Telecom recurring cash flow, but growth is slow because Korea is a mature telecom market.
The AI business is the second engine. AIDC, or AI data centers, sells computing capacity for AI workloads. GPU-as-a-Service rents AI chips to customers that need training or inference power. AIX sells AI and digital tools to businesses, using SK Telecom's own models and partner models.
For consumers, the company wants to move from free or ad-like AI services to subscriptions. The A. agent passed 10 million cumulative subscribers in July 2025, but the real business test is whether users will pay for features such as A.Note and other A. services.
This model can work if telecom cash stays steady while AI starts to scale. It can break if marketing costs rise, if churn returns, or if AI assets sit underused after SK Telecom spends the money to build them.
Phones, fiber, agents, and chips
5G mobile service
This is the main profit base. It earns monthly plan fees, usage fees, device sales revenue, and interconnection fees.
SK Broadband and Pay TV
This includes broadband internet, IPTV, cable TV, fixed-line phone service, and business communications. It is slower growth, but it adds scale and steady customer relationships.
AI Data Centers and GPU-as-a-Service
SK Telecom is selling AI compute through data centers and rented GPUs. Its Ulsan AI data center with AWS is aimed for 2027 operations, and it has an MoU for a dedicated OpenAI data center in Korea's Southwest region.
AIX enterprise AI
AIX is the business AI line. Management said AIX revenue was KRW 193 billion last year and is expected to grow 30%.
A. personal AI agent
A. is the consumer AI agent. It passed 10 million cumulative subscribers in July 2025, and management is targeting a paid model in H1 2026.
A.X K1 and Aster
A.X K1 is a Korean hyperscale AI model with more than 500 billion parameters. Aster is an agentic AI service that can help test demand outside the core telecom bundle.
Still mostly cellular
The mix below uses 2024 revenue from SK Telecom's 2024 Form 20-F. AI revenue is still reported inside the broader telecom lines, so these shares show the filing segments, not a clean AI versus telecom split.
What could break
Subsidy war after MDDIA repeal
High impact · Medium oddsThe handset subsidy law was abolished on July 22, 2025. That can push carriers to spend more to win or keep valuable phone customers. If SK Telecom has to match heavy subsidies, marketing costs could rise and margins could fall.
AI capex outruns AI revenue
High impact · Medium oddsAI data centers need expensive chips, power, land, and cooling. SK Telecom has strong partners, including AWS and OpenAI, but utilization and margins are not yet proven at scale. If demand is slower than planned, the company may own costly assets with weak returns.
Paid A. launch disappoints
Medium impact · Medium oddsA. has a large user base, but free users do not always become paying users. The company is targeting paid monetization in H1 2026. Weak conversion would lower confidence in the consumer AI pillar.
Cybersecurity trust fades again
High impact · Low oddsThe Q1 2025 data leak caused a new subscriber suspension, USIM replacement planning, churn, and large one-off costs. Q1 2026 data looked better, with about 210,000 net handset adds and dividend payments resuming. A fresh incident would hurt trust and could bring more regulatory costs.
Core telecom keeps shrinking per user
Medium impact · Medium oddsSK Telecom's ARPU fell to Won 29,355 in 2024 from Won 29,874 in 2023. MVNO growth also pulls average revenue lower because these customers use the network at lower revenue per user. If 5G upgrades cannot offset that pressure, the cash cow gets weaker.
In one breath
Is SK Telecom mainly a telecom company or an AI company?
It is still mainly a telecom company by revenue. The AI plan is important because it could add growth, but the core cash flow still comes from mobile and fixed-line subscriptions.
What is the biggest near-term catalyst for SKM?
The key near-term catalyst is the paid launch for the A. AI agent in H1 2026. Q2 2026 earnings are also important because they should show whether Q1's handset subscriber recovery is holding.
Why did the 2025 cybersecurity incident matter so much?
It hit the customer base directly. Management disclosed a suspension of new subscriber sign-ups, planning for USIM replacements across a 24 million customer base, a loss of 750,000 handset subscribers, and KRW 250 billion of one-off dealer compensation in Q2 2025.
What should investors watch in the AI data center business?
Revenue growth alone is not enough. Investors should look for utilization, customer commitments, power costs, GPU costs, and margins, because those decide whether AI data centers become a good business or just a large spending project.