Loading analysis...
SK Telecom Re-rating via Anthropic Valuation
Show original post
SK Telecom $SKM Handing Out Free Stakes in $4T Anthropic
Ticker: $SKM
Positions (proof: https://imgur.com/HdXnafh):
| Contract | Position | Market value (USD) |
|---|---|---|
| SKM Jun18'26 $40 Call | 770 | 235,487.36 |
| SKM Sep18'26 $45 Call | 330 | 139,133.21 |
| SKM Sep18'26 $50 Call | 430 | 131,230.69 |
| SKM Jun18'26 $45 Call | 450 | 71,539.08 |
| SKM May15'26 $35 Call | 100 | 44,495.99 |
| SKM May15'26 $40 Call | 250 | 22,301.91 |
| SKM Sep18'26 $55 Call | 100 | 21,041.23 |
| SKM Jun18'26 $55 Call | 400 | 18,815.80 |
| SKM Jun18'26 $50 Call | 200 | 17,144.55 |
| Total | 3,030 contracts | $701,189.82 |
TLDR
- Anthropic blatantly worth gazillions. I'm saying $4T. This is well ahead of market consensus and could be down to the crack I've been smoking.
- Found some live data points that suggest it is even more dominant than thought.
- $SKM at EV $22.4B, $15B Mcap and ADR $39.08 is holding 0.54% of Anthro on the DL.
- $SKM invested in other AI lovelies from 2022. Some of them glitter too.
- SOTP in $SKM has 18% upside on the core alone.
- Oh forgot Anthropic. Throw them in and you get total equity $38.2B and ADR $99.57. +155% upside.
- I am long via the May, June and September call stack above. Pick your own strikes.
The $4 trillion Anthropic
Revenue explosion
Anthropic's reported ARR by month:
| Month | ARR ($B) | Source |
|---|---|---|
| Dec 2025 | 9 | Anthropic |
| Jan 2026 | 14 | Reuters / The Information |
| Feb 2026 | 19 | The Information |
| Mar 2026 | 30 | Anthropic press release 7 April |
| Apr 2026 | 45 | Financial Times, early May ("almost 45", five sources) |
YTD growth rate is 12,400% annualised. Holy smokes.
From the $45B April base:
| MoM rate | Dec 2026 ARR ($B) | Context |
|---|---|---|
| 50% flat | 1,070 | More than Amazon's revenue today. Punchy. |
| 25% flat | 265 | Past Meta's $215B. Microsoft is at $320B. |
| 10% flat | 71 | Still remarkable. |
| Tapering 45% MoM down to 10% at YE | 304 | My pick |
Comparisons
Only comp for growth at this scale is NVIDIA, which peaked at 206% YoY in late 2023 at a $73B annualised quarterly run rate. 16x Rev. Today $5T on $200B revenue at 60% growth. EV / Revenue 24x.
Snowflake got 130x on $500M revenue at IPO on 120% growth. Palantir is on 42x on $8B today with 71% growth.
12,400% is more than that. I say 90x Revenue for Anthropic. $4T please.
Cross-check: NVIDIA peaked at ~16x forward revenue when growing at 200%. Snowflake IPO'd at 130x at 120%. Anthropic at 12,400% should clear 90x in its sleep.
Morons may say global enterprise IT spend, per finger-in-the-air merchants Gartner, is $1.6T. Claude isn't going to take 20% of that by year-end.
Lucky that is not the TAM. Knowledge work is what is up for grabs. Keyboard warriors. Which means all you losers.
Some of you may start whimpering that other models are just as good, or that the free ones work just as well. Nah. Do you want the best or not? Not you oxygen-kidnappers of course, I mean real businesses, because that's where Anthropic's revenue is coming from.
This powerful market research is likely already too complex for you to follow, but I have gone the extra mile...
OpenRouter: a rubbernecker's dream
OpenRouter lets you switch between hundreds of models willy nilly. It is for the connoisseurs and the tinkerers. Good guide to what is hot and what is not. They happen to give real time data.
Last week:
| Company | Tokens (T) | Tokens share | Revenue ($M) | Dollar share | Avg price ($/M) |
|---|---|---|---|---|---|
| Anthropic | 3.87 | 18.2% | 33.2 | 63.7% | 8.59 |
| OpenAI | 1.85 | 8.7% | 8.1 | 15.5% | 4.37 |
| 3.43 | 16.1% | 3.9 | 7.5% | 1.14 | |
| MoonshotAI | 2.00 | 9.4% | 2.9 | 5.6% | 1.47 |
| Z.ai | 0.84 | 3.9% | 1.2 | 2.3% | 1.40 |
| DeepSeek | 2.99 | 14.0% | 1.0 | 1.9% | 0.33 |
| MiniMax | 1.03 | 4.9% | 0.6 | 1.1% | 0.54 |
| Qwen | 0.77 | 3.6% | 0.4 | 0.8% | 0.52 |
| xAI | 0.78 | 3.7% | 0.3 | 0.5% | 0.37 |
| Tencent | 1.10 | 5.2% | 0.1 | 0.3% | 0.12 |
| Others | 2.60 | 12.3% | 0.4 | 0.8% | 0.15 |
| Total | 21.26 | 100% | 52.1 | 100% | 2.45 |
Good God.
Anthropic taking 64% of the whole market. Charging double OpenAI and 8x Google. Moonshot can take a bow. Everyone else is nowhere.
The top models?
| Rank | Model | Weekly tokens (B) | Weekly revenue ($M) |
|---|---|---|---|
| 1 | Claude Opus 4.7 | 1,350 | 14.85 |
| 2 | Claude Sonnet 4.6 | 1,540 | 10.16 |
| 3 | Claude Opus 4.6 | 589 | 6.48 |
| 4 | MoonshotAI Kimi K2.6 | 1,700 | 2.68 |
| 5 | OpenAI GPT-5.5 | 424 | 2.12 |
Like China in the Ping Pong at the Olympics. Gold, Silver and Bronze.
Mythos model still too powerful to be released of course. Will soon be blunted enough to totally dominate.
$4T might be a bit punchy to swallow? Don't worry, you can get a stake in Anthropic for free.
Hiding in Korea
$SKM is the Verizon of Korea. Except they got a stiffy for AI ahead of the pack. ADR listed in NY at $39.08.
The Boring Core
$3.6B EBITDA on 4.5x peer multiple. $16.2B.
The Data Centres
I stripped out the Data Centres business above, because that is the new rock and roll. 137 MW of compute. One round the corner in Ulsan, Korea sold 49% to KKR for $41M per MW. That one was a bit tastier. I take half of that multiple for $SKM.
The DC have $107M EBITDA. Comps at Equinix, Digital Realty and NextDC are around 20x.
Blended mark: $2.3B.
The AI Stiffy
They made a tonne of AI investments, and some of them have turned into gold:
| Investment | Listed | Book ($M) | My valuation ($M) | $ uplift |
|---|---|---|---|---|
| Rebellions | Imminent | 131 | 785 | +654 |
| Hana Financial | yes | 332 | 742 | +410 |
| IonQ | yes | 228 | 484 | +256 |
| Penguin Solutions | yes | 128 | 307 | +179 |
| Persona Identities | no | 3 | 102 | +99 |
| Perplexity AI | no | 10 | 62 | +52 |
| Subtotal of the six | 832 | 2,482 | +1,650 | |
| Others | mixed | 1,290 | 1,429 | +139 |
| Grand total | 2,122 | 3,911 | +1,789 |
Top 4 marked to market. Next 2 mainly based on recent funding rounds. For the others re-valued (or not) based on the same. Another $3.9B.
Sum Up
| Component | Value ($B) |
|---|---|
| Core (4.5x EV/EBITDA) | 16.2 |
| Data Centres | 2.3 |
| Investments ex Anthropic | 3.9 |
| Enterprise Value | 22.4 |
| Less: net debt | (4.8) |
| Ex Anthropic equity value | 17.7 |
| Current market cap | 15.0 |
$17.7B / current shares = $46.11 per ADR. Gets us to 18% above today's price. And that is before Anthropic.
Bonus Giveaway
$SKM put $100M into Anthropic in August 2023 before it was cool, at a $5B valuation. They called it a "2%" stake at the time.
A few clowns still parrot 0.30% today, because they took the number from a loosely worded SKM disclosure in English and never checked. Other clowns are saying 2% as they have no idea how numbers and dilution work. Korean companies are not known for being forward in sharing the details of their company and have hardly mentioned this stake since.
However, they have disclosed their ownership to those who read the 442-page Korean-language 2025 business report that just came out last month. On pages 441 and 442, they tell you they have 3,703,141 shares in Anthropic.
Working through dilutions, unvested equities, unissued shares, DLOMs, fully diluted calculations, estimated dilution from recent fundraising (Series F, Series G, employee tender, Amazon and Google strategic), it seems to me like they now own 0.54%. 5% DLOM applied below.
Korean Meritz Securities and Korea Investment & Securities both cite 0.58% post-Series-G (pre April 2026 strategic equity). Within rounding of me.
| Anthropic value ($B) | Anthropic stake ($B) | Total equity ($B) | Implied ADR ($) | Upside vs spot $39.08 |
|---|---|---|---|---|
| 380 (Series G priced) | 2.0 | 19.6 | 51.13 | +31% |
| 1,000 | 5.1 | 22.8 | 59.43 | +52% |
| 2,000 | 10.3 | 27.9 | 72.81 | +86% |
| 4,000 (central) | 20.5 | 38.2 | 99.57 | +155% |
| 5,000 (strong bull) | 25.7 | 43.3 | 112.95 | +189% |
| 10,000 | 51.3 | 69.0 | 179.85 | +360% |
Catalysts (the next 6 to 9 months)
- May ARR print (mid-June). $50 to 60B confirms trajectory. $65B+ validates the central as conservative. Sub-$50B and the thesis cools.
- Mythos broader release. Currently restricted to a small set of partners (JPM, Goldman, Treasury, Fed). Broader rollout repositions Anthropic from frontier lab to systemically important infrastructure.
- Series H pricing. Bloomberg and TechCrunch have the active round in the $850 to 900B range, imminent. Above $900B closes the argument that trillion-plus is speculative.
- IPO. Goldman, JPM and Morgan Stanley named as underwriters. S-1 rumoured Q3. Target window Q4 2026, October earliest. On IPO the K-IFRS Level 3 illiquidity discount collapses to zero and the SKM stake becomes a mark-to-market liquid public security overnight. That is the one event that resolves the thesis.
Risks
A few that are actually real, in descending order of how much they keep me up at night:
- My $4T valuation is 4x market consensus. You should account for my crack habit.
- Anthropic IPO slips into 2027. The whole leveraged trade leans on price discovery within 6 to 9 months. Slippage hurts the front-month options the most.
- Holdco discount sticks. Korean holdco structures have historically traded at a 30 to 50% NASPERS-style discount. The Commercial Code Amendment (July 2025) and the Value-Up programme are dismantling that, but the discount may not fully unwind.
- Ownership opacity. SKM does not publish the ownership stake percentage. My 0.54% could be off by 20% in either direction. Anthropic does not let shareholders share proprietary data, so don't expect clarity until S-1.
- Model commoditisation. OpenRouter shows Claude winning on dollar share, not that competition went away. GPT-5.5 or Gemini 4 lands a haymaker, the multiple compresses fast.
- Korean governance. SKM is not run by a lunatic (cf. NASPERS), but the transmission of Anthropic IPO proceeds to minority shareholders is not guaranteed.
Not financial advice. I cannot explain what a semiconductor is. DYOR.
ETA: Deep Dive I shared with some in the comments here. There is the model underlying it all linked there too.
The OP frames NYSE ADR-listed SK Telecom (SKM) as an indirect exposure to Anthropic. The core argument is not the Korean telecom business itself, but rather that the market has not adequately priced SKM’s Anthropic stake. The post discloses that the OP holds 3,030 short- to medium-dated SKM options, covering expirations in May, June, and September 2026, with strike prices from $35 to $55. The notional positioning is highly concentrated, with a market value of about $700,000. The OP’s first layer of argument is that Anthropic’s revenue growth is extremely steep: he lists ARR rising from $9 billion in December 2025, to $30 billion in March 2026, and then to “nearly $45 billion” in April, and extrapolates year-end 2026 ARR using a high monthly growth rate. He then compares Anthropic with high-growth software/AI assets such as NVIDIA, Snowflake, and Palantir on revenue multiples, subjectively assigning a 90x revenue multiple and an approximately $4 trillion valuation. The second layer of evidence is OpenRouter alternative data, claiming that Anthropic accounts for only about 18% of token volume but captures about 64% of dollar revenue, and that Claude Opus/Sonnet occupy the top three highest-revenue model spots. This is used to argue that Claude has pricing power in high-paying enterprise/API use cases. However, commenters also point out that OpenRouter is more like “traffic at one intersection,” with a sample biased toward developers and model enthusiasts, and cannot directly represent the whole market. The OP also acknowledges it is noisy and can only suggest that enthusiasm for new models and price competition have not yet obviously eroded results. The third layer is SKM’s valuation: the OP values the traditional Korean telecom core business at $1.62 billion using approximately $360 million of EBITDA and 4.5x EV/EBITDA; values the data-center business at $230 million using 137MW of capacity and a midpoint around 20x peer EBITDA; estimates other AI/financial investments such as Rebellions, Hana Financial, IonQ, Penguin Solutions, Persona, and Perplexity at a combined $390 million; and, after deducting $4.8 billion of net debt, derives $17.7 billion of equity value excluding Anthropic, corresponding to about $46.11 per ADR, or 18% upside versus the $39.08 current price cited in the post. The attached image is exactly this waterfall chart: core telecom 16.2, AI data centers 2.3, non-Anthropic external investments 3.9, net debt -4.8, resulting in 17.7 of equity value excluding Anthropic; then adding about $20.5 billion of stake value under a $4 trillion Anthropic valuation, giving total equity value of $38.2 billion, an implied ADR value of $99.57, and 155% upside. The most important and also most fragile links are the ownership percentage and Anthropic valuation. The OP says SKM invested $100 million in Anthropic in August 2023, then about 2%, and that the latest Korean annual report discloses 3,703,141 shares; after adjusting for financing rounds, employee equity, strategic investments, and a 5% liquidity discount, he estimates the current stake at about 0.54%, and says Korean brokers have a 0.58% figure. But public English/brokerage materials also show a 0.3% figure, suggesting that “common-share ownership percentage” and “fully diluted percentage” may be getting mixed. Catalysts include future monthly ARR, broader Mythos release, a possible new financing at $850 billion to $900 billion, and a rumored Q4 2026 IPO. Risks include the fact that $4 trillion is far above the latest official $380 billion financing valuation, an IPO delay would hurt short-dated options, the Korean holding-company discount and minority-shareholder value transmission are uncertain, and AI model price wars/open-source substitutes could compress multiples.
The comment sample is small but contains substantive disagreement. Supporters argue that SKM is a rare indirect exposure to Anthropic, and someone also adds that foreign ownership limits on Korean telecom companies may create an ADR premium. The main rebuttals focus on two points: OpenRouter’s sample cannot represent the whole market, and Anthropic ownership may be closer to 0.3% rather than 0.54%. The OP acknowledges that OpenRouter is only biased alternative data, and explains that 0.7%, 0.3%, and 0.54% come from different dilution methodologies. Another highly upvoted comment simplifies the post as a “high-risk Anthropic proxy.”
Loading Truth Social...
Loading live updates...