SKM · SK Telecom Co., Ltd. · Stock research

Current Thesis

The Korea AI-infrastructure narrative has gotten louder while SKM's ADR has gotten cheaper, and that divergence is the whole read. Since the 2026-06-02 close of $46.00, the ADR has made a near-unbroken series of lower highs into $30.62 on 2026-07-17 (−3.59% on the day), about −33% in six weeks and −35% off the $47.18 52-week high. The July news flow was, on paper, the best of the cycle: a 15GW AI data-center plan announced 2026-07-05, a CEO-led AI DC Integrated Task Force that lifted the Seoul line +8.47% to ₩92,200, and a JPMorgan upgrade to Overweight on ~2026-07-16 with the price target raised to ₩110,000 from ₩80,000. The ADR closed lower the session after the upgrade. When a stock cannot hold a bid on an analyst upgrade stacked on top of a $91.5B buildout announcement, the marginal buyer has already been served.

What changed underneath is the character of the story. Through May the pitch was a margin-accretive AIDC line bolted onto a cash-generative carrier AIDC revenue +89.3% YoY to ₩131.4B in the Q1 print (2026-05-09). Since 2026-07-05 the pitch is ₩140T of Yeongnam capex, a 6-K on 2026-07-10 confirming that a ~₩1T capital increase and KKR-backed equity financing are under review but undecided, and final-stage talks to sell up to 49% of the Ulsan asset (KKR 29%, IMM, Stonebridge). A dividend telecom funding a hyperscaler-sized buildout is a different security than the one that re-rated in June. Equity holders are being asked to underwrite the option, and they are voting with the tape.

Bullish and bearish views on SK Telecom Co., Ltd.

The model's bull view on SK Telecom Co., Ltd. (SKM), in brief: AIDC revenue is compounding in reported numbers: +89.3% YoY to ₩131.4B in Q1 2026 (reported 2026-05-09), driven by GPU-as-a-Service demand and utilization ramp at newly live sites. The bear view: Good news stopped working: the 15GW announcement (2026-07-05) produced a $31.67 ADR close on 07-06 and $31.05 on 07-07. Both cases follow in full.

Bull Case

• AIDC revenue is compounding in reported numbers: +89.3% YoY to ₩131.4B in Q1 2026 (reported 2026-05-09), driven by GPU-as-a-Service demand and utilization ramp at newly live sites.
• The buildout is dated and staged, not aspirational: 2026-07-05 plan targets 5GW domestic capacity opening in phases from 2029 a >2GW southeastern (Gyeongsang) cluster plus 1GW in Jeolla inside a 15GW long-run ambition, with "AI Factory" operations slated to begin 2027 on the Nvidia DSX platform.
• Sell-side is finally moving: JPMorgan (Stanley Yang) went Neutral → Overweight ~2026-07-16, PT ₩110,000 from ₩80,000. HSBC had only reached Reduce. The upgrade cycle is early, not exhausted.
• Sovereign policy tailwind: the plan is explicitly linked to Korea's "AI G3" strategy, which usually means land, grid interconnect, and permitting move faster than a private developer could manage alone.
• Third-party capital validates the asset: KKR and SK launched a $1.3B Korean renewables platform on 2026-07-01, and a KKR-led consortium is negotiating for up to 49% of Ulsan outside money pricing the data-center estate above the parent's implied valuation.
• Franchise repair is real: +210,000 net handset adds in Q1 2026 against ~550,000 lost YoY post-USIM-breach, with the dividend reinstated at ₩830/quarter after the Q3/Q4 2025 suspension. Local yield screens ~3.95% on ₩3,540.
• Cheap on forward numbers: 14–15x forward earnings versus ~55x trailing, with the gap closing as breach-related costs roll off.

Bear Case

• Good news stopped working: the 15GW announcement (2026-07-05) produced a $31.67 ADR close on 07-06 and $31.05 on 07-07. The JPMorgan upgrade produced $30.62 on 07-17. Two mainstream-headline catalysts, zero net price. That is distribution.
• The ADR is the weaker line: the Seoul listing closed ₩89,700 on 2026-07-16, +5.53% on the session; the ADR closed lower the same day and again the next. Won depreciation plus dollar-holder selling means the US expression is not capturing the local rally.
• Funding structure is the live risk: the 2026-07-10 filing puts a ~₩1T capital increase explicitly on the table and says the market gets an update "within a month." Selling 49% of the crown-jewel Ulsan asset while raising equity is how a levered buildout gets financed, and both dilute the holder who bought the AI story.
• The base business is shrinking: FY2025 revenue −4.69% to ₩17.10T, earnings −68.41% to ₩388.61B. AIDC growth compounds off ₩131.4B per quarter against a ₩17T top line it cannot carry consolidated growth for years.
• The non-infrastructure AI line is going backwards: AI B2B/B2C revenue fell −10.3% YoY to ₩45.0B in Q1 2026 on cloud weakness. The "A." assistant/services layer is not a grower.
• Capital returns compete with capex: ₩140T ($91.5B) of announced regional investment sits in direct tension with a dividend only reinstated two quarters ago. Income holders who came back in May have a reason to leave again.
• Structurally a low-beta carrier: a 0.4–0.6-beta regulated telecom re-rates once on a story and then goes inert. The June episode +18.85% on 2026-06-01, $46.00 close on 06-02, full round-trip to $32.16 by 06-30 is that mechanic in miniature.

Setup & Price Structure

Broken and still descending. The ADR lost the $36–$38 May breakout shelf in mid-June and has not retested it. July traced a textbook lower-high sequence: $31.05 (07-07) → bounce to $33.23 (07-10) → $30.62 (07-17), with the bounce failing well beneath the prior swing. Price sits below the 20-EMA and the 50-DMA, both declining. The 52-week range is $19.66–$47.18; the entire June spike is now air above, and the nearest meaningful demand is the pre-rally low-$20s. Analyst consensus on the ADR is Hold at a $34.55 target price is below the average PT, which in a downtrend is a warning that estimates are stale rather than an upside signal.

The only constructive read is a possible washout base: $30 is a round number, the forward multiple is mid-teens, and the local line is holding a rising trend the ADR is not. But there is no higher low yet, no volume climax, and no reclaim of $33.23. A base is a thing you observe, not a thing you anticipate. Fresh entries here are catching a knife with a $91.5B capex plan and an unresolved equity raise on the other side of the blade.

Catalyst Calendar (next 30 days)

• ~2026-08-05 / 2026-08-06 (est.) Q2 2026 results. The binary. Watch AIDC revenue growth (Q1 base: +89.3% YoY, ₩131.4B), the AI B2B/B2C line (Q1: −10.3% YoY), net handset adds, and any capex-guide revision tied to the 15GW plan.
• by ~2026-08-10 SK Telecom committed on 2026-07-10 to update the market "within a month" on AI data-center financing structure. This resolves the ~₩1T capital-increase question and the KKR-led consortium's Ulsan stake.
• Ongoing, undated Ulsan 49% stake sale close (KKR 29% / IMM / Stonebridge). A headline valuation on the data-center estate is the single most likely positive surprise.
• Ongoing dividend declaration cadence alongside the Q2 print; confirmation or non-confirmation of the ₩830/quarter run-rate is the tell on whether capex is crowding out capital returns.
• No Nvidia, PDUFA, or index event inside the window.

What Would Change Our Mind

The bullish flip requires price confirmation, not more announcements. Specifically: a weekly close back above $33.50 — that holds the 07-10 swing high, ideally on the financing update landing as a stake sale at a premium valuation rather than a dilutive primary raise. Second, AIDC revenue holding ≥60–80% YoY growth at the ~2026-08-05 print with a capex plan the balance sheet can carry without equity. Third, the ADR beginning to outperform EWY and the Korean AI cohort rather than lagging it if this merely tracks the country ETF, the country ETF is the cleaner instrument with none of the funding risk.

Conversely, a weekly close below $29.50 breaks the July shelf and opens the gap toward the pre-June-rally low-$20s, with the $19.66 52-week low as the structural floor. A confirmed primary equity raise announced before the Q2 print would accelerate that path.

Correlation Notes

• Tracks EWY (MSCI South Korea) and the won closely; for pure Korea-AI beta, EWY carries the theme with no single-name funding risk.
• Indirect memory-cycle linkage via SK Square's ~20% SK Hynix holding a group-level tail, not a line on the SKM P&L.
• Correlated to the Nvidia/AI-datacenter complex (NVDA, VRT, and the Korean AI-power names) on narrative days, but with a 0.4–0.6 beta that damps upside far more than it damps downside once a story unwinds.
• Rate-sensitive as a dividend carrier: a hawkish global rate impulse hits the yield-support bid at the same time it raises the cost of a ₩140T buildout the two risks are not independent.
• SK Broadband is now wholly owned (99.24% → 100%), so fixed-line and data-center cash flows consolidate fully; no minority leakage, but also no separate listing to mark the asset.

Notes

Anthropic's $100M strategic investment (2024-08-05) remains a durable structural tie-in and is the reason SKT shows up in Western AI-partnership screens more than its revenue mix warrants.

Where It Sits Now

Legacy Pivot, and the pivot is real but the market has moved from pricing the opportunity to pricing the bill. The theme is ACCELERATING; this expression of it is not. Stand aside until a higher low prints.

Beginner-Trap Check

Trading below every recent moving average with an upgrade that failed on arrival, an unresolved dilution question, and earnings inside three weeks. This is the exact profile where cost-basis anchoring and "it's cheap now" reasoning destroy capital. There is no averaging-down case here because there is no established base to average into.