SK Telecom’s Profit Jumps 67% as AI Data Center Revenue Surges

SK Telecom reported a 67.3% jump in quarterly operating profit, giving Google News a striking number tied to surging AI data center revenue.

The Korean carrier earned KRW 566.0 billion in operating income during the second quarter of 2026. Revenue reached KRW 4.3591 trillion, while net income totaled KRW 466.0 billion.

Its AI data center business supplied the clearest growth signal. Quarterly revenue from that operation rose 92.5% from a year earlier to KRW 136.2 billion.

Yet the headline does not mean AI data centers produced the entire profit increase. SK Telecom says the comparison also benefited from one-time expenses that depressed its second-quarter results in 2025.

That distinction matters. The company is recovering from an expensive cybersecurity crisis while preparing an AI infrastructure expansion measured in gigawatts, not server racks.

Investors therefore face two stories inside one earnings release. One is a measurable data center business that is already expanding. The other is an ambitious construction plan whose financing, customers, power supply, and returns remain unsettled.

SK Telecom’s immediate opponent is not another carrier. It is the gap between the economics of its operating data centers and the capital demands of its proposed AI infrastructure empire.

What SK Telecom’s Google News Headline Leaves Out

The 67.3% profit increase is real, but it combines business recovery, cost control, and a favorable comparison with AI data center growth.

According to SK Telecom’s Q2 earnings, consolidated revenue increased only 0.5% from the same period in 2025. Operating income, by contrast, rose from KRW 338.3 billion to KRW 566.0 billion.

The difference between those growth rates should immediately temper a simple revenue-driven interpretation. A company whose sales increased 0.5% did not generate a 67.3% profit gain through top-line expansion alone.

SK Telecom identifies efficient, profitability-focused management as one contributor. Operating income also increased 5.3% from the preceding quarter, suggesting that the recovery was not exclusively a year-over-year accounting effect.

Still, the prior-year comparison carries unusual baggage. During the second quarter of 2025, SK Telecom absorbed one-time costs associated with a major customer data breach.

Those expenses included subscriber identity module replacements and compensation for retail stores. The company also suspended new customer sign-ups for part of that period.

The unusually weak base makes the 2026 growth percentage look more dramatic. It does not make the reported profit invalid, but it changes what that percentage reveals about the business.

A year earlier, operating income had fallen 37.1% to KRW 338.3 billion. Net income had dropped 76.2% to KRW 83.2 billion, according to breach reporting published after those results.

SK Telecom’s latest operating profit is therefore best viewed as a restoration of earnings power. It is not evidence that a new AI operation created two-thirds more profit within one year.

The latest result nevertheless exceeded the expectations available before the announcement. Mirae Asset Securities had estimated second-quarter operating profit of KRW 557.5 billion, about 3.5% above the market consensus it cited.

Actual operating income came in at KRW 566.0 billion. That placed the result above both the brokerage estimate and the earlier consensus benchmark.

The forecast also correctly identified higher data center utilization as an earnings contributor. It expected additional revenue from the Pangyo facility and improved performance at SK Broadband.

The profit recovery is not occurring only inside the AI division. SK Telecom said its mobile business continued adding subscribers as it worked to restore customer trust.

The company also declared a quarterly dividend of KRW 830 per share for the second consecutive quarter. The dividend signals greater confidence in near-term cash generation, although it says little about long-term data center returns.

Google News readers should consequently separate three distinct developments:

• SK Telecom’s core telecom profitability recovered.

• Last year’s one-time breach costs created a favorable comparison.

• AI data center revenue grew much faster than total company revenue.

The third development is the most strategically important. The first two explain much of the eye-catching profit percentage.

That combination creates a better question than whether AI “caused” a 67% profit jump. The relevant question is whether the existing AI data center operation can scale without sacrificing the restored profitability celebrated in the headline.

AI Data Center Revenue Is Becoming Material

SK Telecom’s AI data center business remains small beside the group, but its growth has become too large to dismiss as a presentation metric.

AI data center revenue reached KRW 136.2 billion in the second quarter. That represented 92.5% growth from the same period one year earlier.

The segment accounted for slightly more than 3% of consolidated quarterly revenue. It is not yet large enough to determine the entire company’s performance.

However, its growth rate sharply exceeded the group’s 0.5% revenue increase. The divergence makes data centers an increasingly important source of incremental growth.

The pattern also extends beyond a single quarter. SK Telecom reported first-quarter AI data center revenue of KRW 131.4 billion, up 89.3% year over year.

That means the business generated KRW 267.6 billion during the first half of 2026. Second-quarter revenue also increased from the first quarter, though at a much slower sequential rate.

The quarterly pattern matters because year-over-year percentages can exaggerate growth from a small base. Consecutive revenue gains provide stronger evidence that the operation is attracting sustained demand.

The company attributes the latest increase to expanded operations. Earlier disclosures connected growth to higher utilization at facilities including Gasan, as well as GPU-as-a-Service revenue.

GPU-as-a-Service gives customers access to graphics processors through a cloud model. It lets businesses rent computing capacity without buying and operating every accelerator themselves.

That service can generate more value than conventional colocation, where customers primarily lease space, power, cooling, and network connectivity. It also exposes the provider to greater hardware, software, utilization, and customer-concentration risks.

SK Telecom wants to move further along this value chain. The company describes an AI cloud built for model training, inference, and agent-based applications, rather than a facility that merely houses customer equipment.

Training creates or updates an AI model from data. Inference is the process of running that trained model to produce an answer, prediction, image, or action.

Both workloads require dense computing systems, but their economics differ. Training jobs can consume large clusters for limited periods, while inference demand depends on recurring application usage.

The distinction will become important as SK Telecom discloses more detail. Revenue from a heavily utilized existing data center does not automatically predict returns from a newly constructed gigawatt-scale facility.

SK Telecom has not published enough segment detail to calculate a standalone AI data center operating margin. Readers therefore cannot determine how much of the group’s KRW 566.0 billion operating income came directly from this business.

The company’s reporting also groups several AI activities under a broader strategy. Investors should avoid treating all AI-related revenue as interchangeable.

Data center leasing, GPU access, enterprise software, model services, and consumer AI products have different capital requirements. They also produce different margins and renewal patterns.

The current data shows that customers are paying SK Telecom for AI infrastructure. It does not yet show that every proposed expansion will reproduce the economics of its current facilities.

That is the central tension behind the latest result. SK Telecom has crossed from AI infrastructure planning into measurable revenue, but its planned scale is advancing much faster than its disclosed operating record.

The Real Test Is Scaling Revenue Into Gigawatts

SK Telecom is trying to convert a KRW 136.2 billion quarterly business into national-scale infrastructure, creating a financing challenge far larger than its current earnings base.

In July, the company established SK Hyper, a dedicated business development unit for AI data centers. SK Hyper is expected to secure land and power while attracting large global customers.

The separation gives the expansion a focused organization. It also makes the distance between current operations and future plans easier to see.

SK Telecom initially plans to bring a combined 5 gigawatts of capacity online in stages beginning in 2029. Its longer-term ambition reaches 15 gigawatts.

A gigawatt measures electrical power, not computing performance. For an AI data center, it indicates the enormous energy infrastructure needed to support accelerators, networking equipment, storage, and cooling.

SK Telecom says a typical one-gigawatt AI data center might require approximately KRW 70 trillion to construct. Its 15GW plan therefore cannot rely only on ordinary corporate capital spending.

The company expects to combine its own investment with strategic partners, project financing, and long-term customer contracts. Those funding channels shift some risk, but none eliminates it.

Project financing usually depends on a project’s expected cash flows and contracted demand. That makes anchor tenants, power agreements, construction schedules, and borrowing terms central to the investment case.

Long-term customer commitments can improve financing conditions. They can also constrain pricing or require SK Telecom to meet strict delivery, availability, and performance obligations.

Mirae Asset estimated that new capacity would begin contributing more meaningfully to consolidated revenue from 2028. The brokerage maintained a positive view but said the funding structure, business model, and demand outlook needed greater clarity.

That qualification deserves as much attention as the optimistic forecast. The existing earnings release does not disclose signed demand covering the proposed 5-gigawatt first phase.

SK Telecom has identified South Korea’s industrial base as an advantage. The country has memory-chip expertise, advanced manufacturing, established networks, and experience operating energy-intensive semiconductor facilities.

Yet access to technology does not solve every infrastructure constraint. Data centers require suitable land, dependable grid connections, equipment delivery, water or alternative cooling systems, and community support.

The company’s expansion schedule extends over several years because those inputs cannot appear instantly. Even well-funded projects face permitting delays, transmission bottlenecks, and changing chip architectures.

A facility designed around one generation of accelerators must also remain economical as newer systems improve performance per watt. Faster chips can support more workloads, but they can change power density and cooling requirements.

SK Telecom’s current revenue provides evidence of market demand. It does not validate the full 15-gigawatt target.

The company is effectively using a profitable telecom foundation to establish credibility in infrastructure. Its challenge is to prevent the proposed infrastructure buildout from weakening that foundation.

This is why the 67.3% operating profit increase matters beyond one earnings season. Restored cash generation gives SK Telecom more capacity to negotiate partnerships and fund early development.

It also raises the standard for future performance. Investors who celebrate higher telecom margins will expect management to protect those margins while committing to capital-heavy projects.

The expansion can strengthen the company if customers commit early and utilization rises on schedule. It can strain returns if construction runs ahead of contracted demand.

Nvidia Strengthens the Plan but Does Not Remove the Risk

Nvidia gives SK Telecom a credible technical route into AI cloud services, while leaving commercial demand and infrastructure execution in SK Telecom’s hands.

SK Telecom and Nvidia announced plans in June to build a gigawatt-scale AI cloud in Korea. The first AI factory under that collaboration is scheduled to begin operating in 2027.

An AI factory is Nvidia’s term for infrastructure designed to produce AI outputs at scale. It combines accelerated computing, networking, software, data center systems, and operational tools.

The companies plan to use Nvidia’s DSX architecture. The reference design connects computing systems with facility planning and software intended to manage performance, resilience, and multiple customers.

Their Nvidia collaboration gives SK Telecom access to an established accelerator platform and a recognizable ecosystem for developers and enterprise buyers.

It also addresses part of the technical coordination problem. Building an AI service requires more than installing GPUs in an ordinary server room.

Operators must manage power density, cooling, networking, storage, workload scheduling, security, failure recovery, and software compatibility. A coordinated architecture can reduce deployment friction.

SK Telecom brings telecommunications networks, customer relationships, enterprise operations, and existing data center experience. Other SK Group companies add capabilities in memory, energy, construction, and industrial operations.

This combination creates a plausible full-stack strategy. It connects components, facilities, connectivity, and customer-facing services within one corporate group.

However, a technology partnership is not the same as contracted utilization. Nvidia can help define the systems, but it cannot guarantee that customers will rent enough capacity at profitable rates.

The partnership also ties SK Telecom’s deployment economics to Nvidia’s product road map. New accelerator generations can increase useful computing output, yet they may require expensive upgrades.

Large cloud providers face the same refresh-cycle problem. Hardware that attracts customers today can lose relative appeal when a more efficient platform enters the market.

SK Telecom must also compete for equipment and customer attention. Global cloud platforms already offer extensive AI services, developer tools, and international regions.

The Korean carrier’s strongest differentiation is likely to come from local infrastructure, regulated workloads, industrial relationships, and sovereign AI requirements.

Sovereign AI refers to systems developed or operated under a country’s control over data, infrastructure, and policy. Governments and regulated businesses may prefer local capacity for sensitive workloads.

That demand can support domestic providers even when global clouds offer greater scale. It can also depend heavily on government procurement, regulation, and national technology policy.

SK Telecom says its AI cloud will support enterprise, physical, agent-based, and sovereign AI applications. Physical AI connects models with machines such as robots, vehicles, or manufacturing equipment.

Those use cases fit South Korea’s industrial strengths. They also need more than raw computing capacity.

A factory customer may require secure data pipelines, specialized models, low-latency connections, operational integration, and technical support. SK Telecom must show it can deliver that complete service profitably.

Nvidia’s involvement reduces questions about the basic computing architecture. It does not settle the larger issue of who pays for each phase and how quickly that capacity fills.

The partnership should therefore be treated as an execution enabler. It is not proof that SK Telecom’s long-term capacity target will earn an acceptable return.

Korean Telecom Rivals Face the Same AI Economics

SK Telecom’s earnings put pressure on KT and LG Uplus, but every Korean carrier must balance slower mobile growth against capital-intensive AI ambitions.

South Korea’s telecom market gives its largest carriers dependable recurring revenue. It also offers limited room for rapid growth from conventional mobile subscriptions.

Average revenue per user has matured, while network investments and customer acquisition still consume capital. That pushes carriers toward enterprise services, cloud computing, and AI infrastructure.

KT and LG Uplus are pursuing related opportunities. Their individual strategies differ, but all three carriers want data centers and enterprise AI to become new growth engines.

Before the results, analysts expected mixed second-quarter performance across the sector. Telecom forecasts anticipated that one-time costs and different expense structures would separate the carriers’ earnings.

SK Telecom ultimately reported stronger operating income than that consensus forecast. Its AI data center growth also gave the market a measurable indicator that infrastructure demand is reaching carrier financial statements.

This places pressure on rivals to disclose comparable evidence. Announcing a data center target is easier than showing recurring revenue, utilization, and operating profit.

Still, SK Telecom’s 67.3% gain should not become a direct performance benchmark for KT or LG Uplus. The figure includes a favorable comparison with the breach-affected quarter of 2025.

Different base effects can produce radically different annual percentages. KT, for example, faced an unfavorable comparison in forecasts because an earlier property-development gain had lifted its prior-year result.

The more useful comparison concerns business quality:

• How much data center capacity is operational?

• How much revenue comes from AI workloads?

• Are customers renting GPUs, leasing facilities, or buying managed services?

• How much capacity is backed by long-term contracts?

• What capital must the carrier contribute?

• What return does management expect after energy and hardware costs?

SK Telecom currently has an advantage in the first two questions. It has disclosed operating revenue and rapid year-over-year growth.

Its planned expansion also benefits from relationships across the SK Group. SK hynix’s memory expertise and the group’s energy and industrial capabilities strengthen the infrastructure narrative.

That group structure can improve coordination, but it introduces another risk. Transactions among affiliates can make economic value harder to evaluate unless costs, contracts, and ownership remain transparent.

The competitive battle may also extend beyond Korean carriers. Global cloud providers can sell AI computing directly to Korean companies or participate as anchor tenants in local facilities.

Specialist data center operators bring their own experience in development, financing, and customer acquisition. Utilities and construction groups control other critical parts of the project.

SK Telecom must therefore compete while collaborating. A global cloud company might rent capacity from SK Hyper in one project and challenge its AI cloud services elsewhere.

This mixed relationship makes simple carrier rankings less useful. The winner will be the operator that secures power, financing, equipment, and customers in the right order.

SK Telecom’s current data center revenue shows that it has started that process. Its next competitive test is whether SK Hyper can repeat it at a much larger scale.

The 67% Jump Still Carries a Cybersecurity Shadow

SK Telecom has restored profitability faster than it has erased the strategic consequences of its 2025 data breach.

The favorable comparison behind the latest result came from a serious operational failure. In April 2025, SK Telecom disclosed a cyberattack involving subscriber identity information.

The company replaced SIM cards across its customer base, halted new subscriptions temporarily, compensated affected retailers, and introduced customer support measures.

Those actions reduced second-quarter profit in 2025. Later customer benefits and security spending extended the financial impact.

That history explains why a direct comparison between the two quarters needs care. It also creates a continuing test for SK Telecom’s AI strategy.

AI infrastructure customers place sensitive data and critical workloads inside a provider’s systems. Trust, isolation, access control, incident response, and operational resilience are part of the product.

A telecom breach does not prove that a separate AI data center will be insecure. It does show why enterprise buyers will examine SK Telecom’s controls closely.

The planned facilities could host workloads related to manufacturing, robotics, communications, and public-sector services. Interruptions or unauthorized access could carry consequences beyond an ordinary consumer application.

SK Telecom has emphasized customer trust and security since the incident. The latest earnings release says mobile subscriber additions continued, suggesting that commercial recovery is underway.

However, subscriber recovery is not equivalent to independent security validation. Customers often remain with a carrier because switching is inconvenient or competing offers are similar.

Enterprise infrastructure contracts involve deeper technical reviews. Buyers may request audit reports, data-location guarantees, access policies, incident procedures, and contractual remedies.

SK Telecom’s challenge is to turn its breach response into verifiable operational discipline. That requires more than reassurance in an earnings statement.

The company also needs to avoid using AI growth as a distraction from accountability. Its future infrastructure role will increase the amount and importance of data under its control.

This is the skeptical angle that should follow every upbeat Google News headline about the profit rebound. Last year’s security costs now make current growth look stronger, but the underlying trust obligation has grown.

The breach also illustrates how one operational event can alter capital allocation. Funds intended for growth can be redirected toward remediation, compensation, legal exposure, and infrastructure replacement.

Gigawatt-scale projects create additional concentration risk. A failure affecting a major facility can disrupt many customers simultaneously.

Investors should watch for detailed security disclosures alongside construction updates. Independent certifications, tenant requirements, resilience design, and incident reporting practices will be as important as server counts.

The strongest version of SK Telecom’s strategy combines profitable infrastructure with better operational controls. The weakest version uses recovered earnings to expand before governance catches up.

The second-quarter numbers cannot distinguish between those paths. They simply show that the company has regained financial momentum.

What to Watch After the Google News Surge

Three signals will reveal whether SK Telecom is building a durable AI infrastructure business or extending an attractive headline into an expensive promise.

The first signal is the next sequence of AI data center revenue. Investors should compare sequential growth, not only year-over-year percentages.

Second-quarter AIDC revenue rose to KRW 136.2 billion from KRW 131.4 billion in the first quarter. That is continued growth, but it is far less dramatic than the annual percentage.

A sustained quarterly increase would show that new demand is arriving after the easiest utilization gains. Flat revenue would weaken the argument that current momentum supports a much larger buildout.

Margins matter just as much. SK Telecom should eventually disclose enough information to separate profitable utilization from revenue purchased through heavy hardware investment.

The second signal is contracted demand for the 2027 and 2029 capacity. Named anchor customers, binding capacity commitments, or clearly described long-term agreements would strengthen the financing case.

General statements about demand from global technology companies are not equivalent to contracts. Project lenders and investors need evidence that customers will use the facilities after construction.

The mix of customers will also matter. A single large tenant can accelerate financing but increase concentration and bargaining risk.

Several committed customers can diversify revenue, though they make system design and scheduling more complicated. Either structure can work if pricing covers capital and operating costs.

The third signal is the financing and power structure for SK Hyper. Readers should look for disclosed ownership, partner contributions, grid agreements, and the amount of capital SK Telecom must supply.

A project financed mostly through outside partners and long-term contracts would reduce pressure on the telecom balance sheet. Heavy direct funding without secured demand would increase risk.

Power milestones deserve particular attention. Land can be acquired before enough electricity is available, leaving a project unable to operate at its planned scale.

The scheduled 2027 AI factory offers an early technical checkpoint. Delivering it on time would support confidence in the broader plan, though it would not validate all 15 gigawatts.

The first phased 5-gigawatt target begins later, in 2029. That creates several years during which technology, financing conditions, and AI demand can change.

SK Telecom’s latest earnings deserve attention because they contain genuine progress. AI data center revenue nearly doubled from a year earlier, while consolidated profitability recovered.

The result does not justify treating the entire 67.3% increase as an AI dividend. A weak prior-year base and improved cost management remain essential parts of the explanation.

For readers following the company through Google News, the next task is simple: look beyond the annual percentage. Track quarterly AIDC revenue, signed customer demand, and SK Hyper’s funding commitments.

Those three indicators will show whether SK Telecom can convert operating momentum into infrastructure returns. Until they arrive, the earnings release is evidence of a promising business, not proof of a 15-gigawatt success.