SKM Options Bet as an Anthropic Curve Exposure

Asked my Claude to summarise for me in case you might want to see a summary of my reasoning. Might be useful. I do my DD together with Claude , so this summary is an incomplete glimpse into my Due Dilligence. It is not a perfect presentation , but I wouldn't have written so much myself.

# SK Telecom (SKM) - Due Diligence v2

An indirect Anthropic position with a telecom business attached
Working analytical draft · v2 May 14, 2026 (incorporates the FY2025 annual report) · For personal use

-----

## §1 - Summary

SK Telecom is South Korea's largest wireless carrier. It holds 3,703,141 Anthropic shares, acquired for ~$101.6M in August 2023. As of the FY2025 audit, those shares were marked at ~$1.0B (KRW 1,376,232M) - a ~10× step-up over cost - at an implied Anthropic valuation of ~$350-380B (Series G era). The stake itself, in absolute terms, is large. What's contested is what percentage of Anthropic those 3.7M shares represent - see §3.

The trade, stated honestly: at ~$39.5, SKM has roughly doubled off its 52-week low (~$19.66) and trades near its 52-week high (~$40.49). US sell-side average price targets sit below the current price; Morningstar flags it as trading well above fair value. The post-spin telecom-only trading range (2022-2024) implies a core floor ~35-40% below today's price. This is not a value investment with embedded free optionality - it is a directional bet on Anthropic's valuation continuing to rise and being marked/monetized, with a telecom business underneath that cushions, but does not prevent, a large drawdown.

Most material new information (since v1):

- The FY2025 annual report was obtained and read. Core business numbers are now audited rather than estimated - and they're worse for FY2025 than the v1 DD conveyed. Q1 2026 is a genuine bounce but off a deep hole.
- The Anthropic ownership reading that v1 used (~0.55%) is now one of two live readings, not the consensus answer. The filing literally states 0.3% end-of-year. Whether that reflects real dilution or a denominator-presentation change is the open question (§3).
- Anthropic talks of a $30-50B raise at $900B+ pre-money, board decision expected May 2026, not yet signed. Bullish for the mark; potentially delays the IPO, which is a direct risk to any fixed-expiry option position (§10).
- Anthropic gross margin resolved: roughly −94% (2024) → ~50% (2025), with Anthropic itself projecting ~77% by 2028. Materially better than the bear-case "40% forever" framing.

### Headline metrics (May 14, 2026)

### Anthropic snapshot

-----

## §2 - The business, with audited FY2025 numbers

SK Telecom is a mature, dominant mobile network operator. Three segments: Cellular (~73% of revenue), Fixed-Line / SK Broadband (~25%), and Other (~2%, including AI services and T-commerce). ~42.7% mobile market share; three-player Korean oligopoly. Note: the SK Hynix stake went to SK Square in the 2021 spin-off - SKM has zero direct Hynix exposure.

### FY2025 vs FY2024, consolidated (management basis)

### Subscriber & ARPU data (from the filing)

- MNO subscribers: 30,853k (FY25) vs 31,786k (FY24) - −933k. The breach cost almost a million subscribers.
- 5G: 17,492k, 80% penetration, 45.5% 5G market share.
- Churn: 0.9% (FY25) vs 0.8% (FY24).
- Billing ARPU: KRW 27,845 (FY25) vs KRW 29,355 (FY24) = −5.1%.
- Quarterly ARPU 2025: Q1 29,202 → Q2 29,204 → Q3 24,125 → Q4 28,848. Q3 was the bill-credit trough; Q4 recovering but not back to pre-breach. Note: the "ARPU rising on a smaller higher-intent base" turnaround pattern (e.g. Bumble) does not apply here - SKM's ARPU fell.

### Breach cost, quantified from the filing

- Revenue deduction (customer-appreciation packages, penalty waivers): KRW 454,143M (~$329M).
- Operating expenses for breach follow-up: KRW 211,998M (~$154M).
- FY2025 P&L impact: ~KRW 666B (~$483M). This excludes the ~KRW 700B / 5-year future security capex commitment.

### Capex was cut 19% (SKT-separate basis)

KRW 1,244B (FY25) vs KRW 1,544B (FY24), with network investment dropping from KRW 1,259B to KRW 734B. This flatters near-term FCF; if sustained, it's underinvestment.

### Q1 2026 - the recovery print (reported May 7, 2026)

Operating income back above KRW 500B for the first time since the breach; +210k handset net adds; AI data center revenue +89% YoY; dividend resumed (Q1 dividend declared). HSBC upgraded to Hold (from Reduce). This is a real positive signal - but it is one quarter, off a deep base, and the FY2025 dividend per share was still less than half of FY2024's. "Resumed" ≠ "restored."

### The OCI vs P&L bridge (this matters)

- Operating business (P&L): net profit down 73% - a breach-scarred year.
- Equity (OCI): total comprehensive income attributable to owners actually rose to KRW 1,938B (from KRW 1,409B FY24), because the Anthropic markup flowed through OCI: the OCI reserve jumped from KRW 647B to KRW 2,026B (+1.38T). Retained earnings barely moved.
- The Anthropic gain papered over a bad year - but only in the equity section, not in earnings, and not in cash. Don't confuse the two.

-----

## §3 - The Anthropic stake: what's settled, what isn't

### What is unambiguously established by the filing

From the FY2025 annual report (page 528, equity investments schedule):

The column structure was cross-checked against two other holdings on the same page whose facts are independently verifiable (Hana Financial's implied per-share price from the column reading came out to ~94,100 KRW vs. an actual late-2025 market price of ~90,000-95,000 KRW). So the layout interpretation is solid.

What is settled: SK owns exactly 3,703,141 shares of Anthropic. The auditor marked them at ~$1.0B. SK sold nothing. The beginning-of-year mark (~$140M at 0.7%) is consistent with a Series-D-era valuation (~$18-20B implied). The ending-of-year mark (~$1.0B at 0.3%) is consistent with a Series-G-era valuation (~$330-380B implied).

### What is NOT settled: which denominator the audit uses

The 0.7% → 0.3% drop with share count unchanged would, taken at face value, require Anthropic's total share/unit base to have grown ~2.3× during 2025. Anthropic raised Series E, F, and G plus converts, so the denominator did grow significantly. But a 2.3× expansion in twelve months - across one set of priced rounds - is a lot, and there is a coherent alternative explanation worth taking seriously.

Reading A - "0.3% is the real economic stake."

Reading B - "0.3% is a fully-diluted accounting figure; economic stake is ~0.5%."

Which is right? The filing's audit is internally consistent at 0.3% - but that consistency does not actually distinguish Reading A from Reading B, because under B the auditor would use the same fully-diluted denominator for both the % and the implied per-share price computation, producing the same ~$1.0B book value either way. My earlier framing that "internal consistency proves 0.3%" was wrong; it doesn't.

What would settle it: the Anthropic S-1 cap-table disclosure (post-filing), or a future SKT report that explicitly states the share-count denominator basis.

### Stake-value table under both readings

The cleanest method anchors to the audited book and scales linearly to future Anthropic valuations. Reading B assumes the "real" economic stake at IPO is ~70% larger than the audited fully-diluted percentage, i.e. ~0.5% vs 0.3% - a midpoint of the Substack's range. (At an actual IPO the difference depends on what fraction of the option pool has been granted; estimates here are illustrative.)

The ~2× spread between the two readings is the single biggest source of uncertainty in the trade.

-----

## §4 - Sum-of-the-parts valuation

Working assumptions: core telecom ~$10B (mid; see §5); other listed/venture stakes ~$0.9B (dominated by SKT's ~$590M listed Hana Financial Group position, which is real, liquid, and on the balance sheet - see §4a); ~386M ADS.

Interpretation: under Reading A, at ~$39.5 the market is pricing well above the central case (it is essentially paying for ~$1T+ Anthropic and the high end of core). Under Reading B, the current price is roughly fair-to-cheap at a $1T Anthropic mark. The trade's value framing flips depending on which reading you believe. This is exactly why §3 matters.

-----

## §5 - The core telecom "floor"

As a pure-play telecom (post-2021 SK Square spin-off), SKM traded around an $8-10B market cap through 2022-2024. That is the empirical revealed-preference value of the core before the Anthropic narrative took hold.

- Core floor (empirical, 2022-2024): ~$8-10B → ~$21-26 per ADS.
- Current price: ~$39.5.
- The floor is ~35-40% below current price.

This is the structural reason this is not a "value with embedded optionality" trade. If the Anthropic narrative deflates - a down round, an IPO delay that drags on, an AI sentiment correction - there is real room to fall.

Mitigants: dividend is back (though cut); AIDC growing 89% YoY in Q1; AI data center segment may earn a higher multiple over time. But these don't change the basic point - the margin of safety that existed when SKM was in the $20s has largely been spent.

### The core-multiple disagreement (why the bullish write-ups get to "free Anthropic")

The bullish framing (e.g. Nosey Parker) values the core at 4.5× Korean-peer-median EV/EBITDA → ~$16.2B, with a DCF "sense-check" around $20B. At a $16B+ core, the entire current ~$15B market cap is explained by the telecom alone, so the Anthropic stake plus venture book appear "free or better than free."

This document uses the empirical 2022-2024 trading range (~$8-10B). Both are legitimate; the gap between them (~$6-8B) is larger than the disputed value of the Anthropic stake under Reading A. Which you believe depends on whether you think 2022-2024 was underpricing the core (bull view) or correctly pricing a no-growth quasi-utility with a then-impending data-breach event (bear view). The honest framing: the "free Anthropic" conclusion is an output of choosing the higher multiple, not arithmetic. Pick the multiple deliberately.

-----

## §6 - The bull thesis

1. Anthropic re-rating is real and ongoing. $380B (Feb 2026) → $1T secondary (April) → $900B+ pre-money primary round in talks (May). If the round closes at $900B+ pre-money, that is a primary mark and forces an upward remark of SKT's carrying value over time.
2. Gross margin trajectory is now more credible. ~−94% (2024) → ~50% (2025), with Anthropic projecting ~77% by 2028. That's the most material positive update since v1, and it strengthens the durability of a high private valuation.
3. The IPO, if it happens, is a hard catalyst. "As early as October 2026" per Bloomberg. A public listing forces price discovery, drives the FVOCI mark, and draws attention to backdoor plays.
4. SKM is the cleanest liquid Anthropic proxy. Amazon/Google/Microsoft hold larger absolute stakes but are far too big for it to matter. At ~$15B market cap, SKT is the only major listed name where the stake meaningfully moves the equity.
5. Retail discovery may be early. 13F data through late 2025/Q1 2026 mostly shows generalists and quants, not dedicated thematic-AI funds. The narrative has hit newsletters and articles, not mainstream retail saturation. If an S-1 lands, a second wave of buying is plausible.
6. Core recovery is tracking. Q1 2026 confirmed the breach recovery is real; dividend resumption (even off a cut base) and AIDC growth give the core its own modest path.
7. The Korean "holdco discount" regime is being actively dismantled. Commercial Code Amendment effective July 22, 2025 extended directors' fiduciary duty to shareholders; the Value-Up programme is the first serious legislative push on this; the KOSPI is up ~80% YTD 2026. This is a real structural tailwind to multiple, regardless of the Anthropic outcome.
8. Reading B (if right) cuts the "stake is half what you thought" objection in half. Under Reading B at $1T Anthropic, the stake is ~$5B - restoring much of the original DD's math.

Foreign-ownership mechanic: Korean law caps foreign ownership of telecoms at 49%. If foreign demand presses against that cap on the KOSPI line, incremental foreign buying gets pushed into the ADR (SKM) - a structural tailwind specifically for the US-listed instrument.

AIDC nuance: the +89% YoY growth headline is real but the operational footprint (Gasan, Pangyo) is at its capacity ceiling; new build won't contribute until 2028 due to KEPCO power-connection lead times (12+ months). So AIDC supports a somewhat higher core multiple over time, but is not a 2026-2027 earnings driver.

-----

## §7 - Catalysts timeline

-----

## §8 - Bear case and risk factors

-----

## §9 - Scenario analysis

Built at your stated weights (>75% probability that Anthropic ≥ $1T), with both ownership readings shown side by side. Core telecom, Anthropic valuation, and venture book all co-move with the scenario.

Probability-weighted expected ADR:

- Reading A: ~$40 (essentially flat to current; +1%)
- Reading B: ~$50 (+27% to current)
- Blended 50/50: ~$45 (+14%)

The gap between the two readings (~$10/share in expected value) is larger than most other variables in the model. If you knew which reading was right, your expected return on the stock would move by ~25 percentage points. You don't, yet.

Probability of finishing below current price (~$39.5):

- Reading A: ~55%
- Reading B: ~30%
- Blended: ~42%

-----

## §10 - The Jan 2027 $50 call

Probability-weighted intrinsic at expiry:

- Reading A: ~$1.35
- Reading B: ~$5.70
- Blended 50/50: ~$3.50

Against a placeholder ~$3 premium:

- Reading A: ~85% chance worthless; needs Anthropic ~$1.8T+; negative EV.
- Reading B: ~55% chance worthless; needs Anthropic ~$1.4T+; modestly positive EV.
- Blended: roughly break-even on EV - but with high variance and significant correlated dependence on the ownership question resolving favorably.

The structural problem the option faces, in either reading:

- The May 12 news that Anthropic may raise $30-50B privately is double-edged. Bullish for the stake's value; potentially delays the IPO beyond the Jan 2027 expiry. Multiple outlets explicitly noted this. The hard catalyst could move outside your window.
- SKM options are thin. Real premium and IV need to be checked live; a wide spread can quietly erase the edge.

Alternatives that hedge the IPO-timing risk:

- A longer-dated expiry (June 2027 / Jan 2028) - more premium, but buffers the IPO slip risk.
- A call spread (buy $45 / sell $55) - caps upside, roughly halves premium, better risk-adjusted if you think $50-55 is plausible but $60+ is unlikely.
- The stock itself - no expiry, ~3% dividend, the telecom floor as partial downside cushion. If your conviction is in the direction rather than a specific catalyst window, the stock is a materially cleaner instrument.

-----

## §11 - Open questions / verification checklist

In rough priority:

1. Resolving §3 - the denominator question. The cleanest resolution would be (a) the Anthropic S-1 cap-table disclosure, when filed, or (b) a future SKT filing that explicitly states whether the % uses basic or fully-diluted accounting. Until one of those, the ~2× spread between Reading A and Reading B persists. Worth checking the H1 2026 SKT semi-annual report (filed ~August 2026) for any update to the % shown.
2. Confirm whether the ~$900B+ Anthropic round closes - and what it implies for IPO timing. Board decision was expected this month.
3. Watch SKT Q2 2026 earnings (Aug 2026) for confirmation that the breach recovery and dividend are holding. A weak Q2 unwinds the Q1 thesis.
4. For an options trade: pull current premium, IV, and open interest on the Jan 2027 $50 and a longer-dated equivalent (Jan 2028 if listed). Compare.
5. Check foreign-ownership ratio on the KOSPI line (Naver Finance / KRX). If foreigners are near the 49% cap, the ADR has a structural tailwind.
6. Check the FY2025 20-F filing (English version, filed ~April 29 2026, on SEC EDGAR) for any English-language disclosure that resolves §3 or further clarifies the carrying value.

-----

## §12 - Bottom line

SKM is a real company with a real, cash-generative (if currently bruised) telecom business, that owns a real, audited ~$1B slice of Anthropic. The Anthropic stake genuinely scales: at $2-4T Anthropic, the stake is worth $5-22B (range across readings), and the stock works on either reading. That path exists.

But three things are now true that the original DD did not adequately price:

1. The ownership percentage that determines half the math is itself contested. At $1T Anthropic, the stake is worth either ~$2.7B (Reading A) or ~$5B (Reading B) - a ~2× spread that the audit by itself cannot resolve. Until Anthropic IPOs or SKT clarifies, this uncertainty is the trade.
2. The audited FY2025 core business is meaningfully weaker than the v1 DD conveyed. Operating income −41%, EBITDA −15.5%, net income −73%, ARPU −5.1%, ~933k subscribers lost, dividend per share cut 53%, capex cut 19%. The Q1 2026 bounce is real but small relative to the hole.
3. At ~$39.5, the market is already pricing somewhere between a $1T Anthropic outcome (Reading B) and a $1.5T+ outcome (Reading A). You are not buying the stake cheap on either reading.

For the Jan 2027 $50 call specifically: the expected value depends meaningfully on which reading is correct and on the IPO happening within the window. Under Reading B with IPO confirmed, it's modestly positive EV. Under Reading A or with IPO delayed, it's negative EV. The most honest characterization is a roughly break-even bet with significant downside risk and concentrated dependence on two correlated unknowns - not the strongly positive EV the v1 DD suggested.

The stock is the cleaner instrument than the option for any version of this thesis where you believe in the Anthropic direction but are uncertain on timing. Size accordingly - against money you've already decided you can lose.

-----

Compiled May 14, 2026 from: SK Telecom FY2025 Annual Report (filed via DART, March 18 2026, 532pp); Bloomberg, TechCrunch, CNBC, and Sacra on Anthropic's funding; Forge Global secondary-market data; SEC filings (6-K, prior 20-F); current market data; and a third-party retail write-up (Nosey Parker, Substack, May 13 2026) whose denominator hypothesis is presented as Reading B in §3. The Anthropic carrying value and share count are filing-verified. Future scenarios are illustrative and depend on inputs (Anthropic's future valuation, IPO timing, denominator basis, capital allocation decisions) that no document can verify in advance. Not investment advice. Standing conflict of interest: this analysis was produced with an AI made by Anthropic, the company whose valuation drives the thesis.