[X-File] The Lottery Bought with Your Phone Bill

Four-line summary

1. A company that was hacked last year and had its CEO fired went from 50,000 to 140,000 KRW and is now back in the 80,000 range.

1. It's not because telecom is doing well. The 132.1 billion KRW quietly invested in 2023 became 4 trillion KRW.

1. One-fifth of SK Telecom's market cap is not from telecom, but from a stake in another company.

1. Your phone bill travels around Silicon Valley and returns as a customer for SKT's data centers.

1. From 50,000 to 140,000, what happened in between?

Last year, SK Telecom was a living hell.

USIM information was leaked, and people were lining up in front of dealerships. Photos of people waiting since dawn to change their USIMs were on the news for days. The CEO was dragged to a National Assembly hearing, bowed his head, and eventually stepped down after leading the company for over four years. The stock price was in the mid-50,000 range. There were even rumors that they wouldn't be able to pay dividends.

But now it's 88,000 KRW. In the meantime, it went up to 139,500 KRW and came back down. Starting from the 50,000 range, hitting 140,000, and now at 88,000—it fluctuated this much in just one year.

Was it because subscribers suddenly increased? No. Adding a few hundred thousand more people doesn't add 6 trillion KRW to the market cap. Was it because the rate plans sold well? Not that either. The telecom market has been a game of musical chairs for years.

The real reason lies in August 2023.

At that time, SKT quietly put $100 million, about 132.1 billion KRW, into a new AI company in the US. The company's name is Anthropic. They are the ones who make Claude. Everyone is familiar with Claude now, right?

Back then, no one cared. A telecom company putting money into a Silicon Valley startup was something they always did.

That 132.1 billion KRW has now become 4 trillion KRW. Roughly a 30x return. They're damn good at stocks.

But this post isn't about that 30x return. It's about what's hidden behind that 30x.

1. While 132.1 billion became 4 trillion

Let's look at one number first.

On May 28, 2026, Anthropic was valued at $965 billion in a new investment round. That's over 1,400 trillion KRW. You can get a sense of the scale if you think of it as being in the same league as Samsung Electronics' market cap. Three months prior, it was $380 billion. It jumped two and a half times in three months. In this round, it surpassed OpenAI for the first time.

Here's something funny.

The stake SKT bought in 2023 was 2%. Now it's around 0.3%.

Why did the stake decrease? Not because they sold it. It's because Google, Amazon, and Microsoft came in afterward, and new shares were constantly issued. When new shares are issued, the existing shareholders' portion shrinks proportionally. This is called 'dilution' in difficult terms, but simply put, it's like four people sharing a pizza, and then twenty more people show up. My slice got smaller.

But the pizza grew to the size of the Gocheok Sky Dome in the meantime.

The stake percentage was cut to one-seventh, but the value of the stake grew 30 times. This means the company's value jumped at a much crazier speed than the rate of dilution.

So, the calculation shows that the 0.3% SKT currently holds is worth roughly 4 trillion KRW.

Let's clarify one thing here. This 4 trillion is not a number disclosed by the company. It's a calculated value derived by securities firms multiplying Anthropic's recent valuation by SKT's stake percentage. What's actually written on the company's books as of the end of last year was around 1.4 trillion KRW. So, 4 trillion is an estimate of "how much we might get if we sell now," not confirmed cash. Don't get confused by this, okay?

1. A venture capital firm is hidden behind the telecom sign

SKT's market cap is around 18 trillion KRW. 4 trillion of that is the Anthropic stake.

This means one-fifth is not telecom, but shares in another company.

If you subtract 4 trillion, 14 trillion remains. KT's market cap is in the 13 trillion range.

So, the articles from a few months ago saying "SKT's market cap is larger than KT and LG Uplus combined" weren't about telecom competitiveness. If you just measure the telecom core, SKT and KT are in a similar weight class. What won wasn't SKT, but the Anthropic stake in SKT's wallet.

And this didn't happen by accident.

SKT set up an investment entity called Atlas Investment in the Cayman Islands. They set up another shell company called Astra AI Infrastructure in Delaware, USA. Through these, they put money into Perplexity and Penguin Solutions. Through an entity called SK Telecom America, they entered Together AI, SambaNova, and Twelve Labs. Domestically, they invested in Scatter Lab, MakinaRocks, and Konan Technology.

Cayman Islands, Delaware. You know what those names mean. They dug multiple channels where taxes and regulations are easy.

Based on public disclosures, SKT's AI-related investment assets were around 2 trillion KRW last year, and 68% of that is the single bet on Anthropic. One bet is more than twice as large as all the others combined.

In summary, SKT has been a company doing venture capital under a telecom sign for several years. No one saw it that way, until one bet exploded 30 times.

And that's how venture capital works. A game where you put in ten, nine die, and one survives. The issue is that the market realized late that SKT is a company playing that game, and is now adding that premium to the stock price.

1. The money has started to draw a circle

This is the reason for writing this post. Everything before was just bait.

Let's lay it out chronologically. Follow along slowly.

August 2023. SKT puts $100 million into Anthropic. 2% stake. Up to here, it's a normal investment.

May 2026. Anthropic raises $65 billion in new funding, and SKT enters this round again. They could have sold some and left since they already made 30x, but they didn't. Instead, they put in more. And in the same round, Samsung Electronics, SK Hynix, and Micron enter together as 'strategic infrastructure partners.'

Three companies that make memory entered as shareholders of one AI company simultaneously. This is not common.

And two months later. July 24, 2026, San Francisco.

SK Telecom and Anthropic sign a document. The content is that Anthropic will participate in the gigawatt-scale data center project SKT is building in Korea. How much they will use and in what way will be negotiated going forward.

Now, let's look at the structure.

I kept saying the company SKT put money into became worth 4 trillion, right? And that company now comes in as a customer renting SKT's data center. Samsung and Hynix, who put money in the same round, make and sell the memory that Anthropic will buy with that money.

Money draws a circle. Every time it goes around, the ledgers of all companies on that circle get thicker.

Doesn't this look familiar? It's the structure they've been fighting over for months in the US. Nvidia invests in OpenAI, OpenAI buys Nvidia chips with that money, so Nvidia's revenue increases, and so Nvidia's stock price rises. One side calls this an ecosystem, and the other calls it a game of hot potato.

A Korean version has quietly emerged. But no one talks about this until the end. Everyone stops at "SKT hit the jackpot with Anthropic."

Don't misunderstand. I'm not saying this structure is a scam. I think this might be part of credit creation.

If there is real demand, this is just smart teaming. They tie their interests together and grow together, and this is actually becoming the standard in the AI infrastructure world. It's much easier to win contracts if you make the customer a shareholder in advance.

But what if the demand is inflated? Then it just becomes a mutual back-scratching to raise values. I buy your company, you buy my goods, so both revenues look like they're increasing, and so both stock prices rise.

The problem is that at this point, there's no way to distinguish between the two.

You only know if they are real customers or just scratching each other's backs when numbers are stamped on a contract. But so far, what's been signed is just "let's do well together." The circle is only proven real when numbers come out showing how many gigawatts will be used for how many years at what price.

That day will be the real turning point for this stock.

1. So, are they making money with data centers?

I should throw some cold water here.

SKT's AI data center revenue in Q2 of this year is 136.2 billion KRW. It grew 92.5% compared to a year ago. Almost double. The growth rate is truly excellent.

But SKT's total revenue in the same quarter is 4.3591 trillion KRW.

Do the math. AI data centers account for a bit over 3%. Even if you scrape together A-dot, AI cloud, and AI customer centers, it's 4.5%.

The other 95% is the money that leaves your account every month via automatic transfer.

The blueprint announced by the SK Group is 15 gigawatts. You might not get a sense of this number. A nuclear power plant produces roughly a bit over 1 gigawatt. So, the plan is to suck up the electricity equivalent to a dozen nuclear plants into data centers. In terms of money, we're talking hundreds of trillions, not just trillions.

So how much money is SKT actually putting in here this year?

It's 330 billion KRW.

They created a new company dedicated to data centers called SK Hyper, and the total amount they decided to put in is 750 billion, which will be paid out until 2030. This year's share is 330 billion.

They drew a hundred-trillion-won picture but only took out 330 billion this year. This means the company has only dipped its toes in.

This itself isn't something to criticize. In fact, the company is being quite cautious. They are following the order of securing customers first and then building. Naver chose the opposite way—fixing the money first and then digging. We'll have to wait about three years to see which side was right. What do you think of Naver's judgment?

What I'm trying to say is this:

The expectations reflected in the stock price are far ahead of the money the company is actually betting. The company has only dipped its toes, but the stock price is already calculating as if they've jumped all the way in.

And one more thing. How much SKT actually takes in this structure is still an unsolved problem. They chose a method of finely splitting the risk and sharing it with multiple partners. As the risk decreases, the take also decreases. That's why some point out that even if the business is a huge hit, SKT's share might be thinner than expected.

1. The real bottleneck is electricity, not GPUs

When talking about data centers, everyone starts with how many GPUs they have, but what's really strangling them now is electricity.

Let's look at one number. The vacancy rate for data centers in the Seoul metropolitan area in the first half of this year is 1.1%. It was 6.9% in the second half of last year. In half a year, they've effectively become full. Capacity increased by 10%, but vacancies plummeted. It means the speed of people wanting to enter is much faster than the speed of building.

Why can't they build? Not because there's no land. It's because they can't bring in electricity.

To bring electricity from power plants to data centers, you need transmission lines, but they aren't being laid. Out of 54 transmission network projects included in the national power plan, 20 are delayed due to resident opposition and permit delays. People don't want transmission towers in their neighborhoods. The sentiment is understandable, and that's why it's even harder to solve.

So, what the industry is saying is this: it used to be a battle over location and tenants, but now the timing of securing electricity determines the business schedule.

In SKT's 15-gigawatt plan, this part is the emptiest. Where to connect to the power grid, who to sign a long-term contract with, and who will bear the cost are all under review. The largest cost item in a data center is electricity, and the blueprint came out first while that item is still a blank space.

But there's a twist here.

If electricity is the bottleneck, the value of assets that already have electricity connected rises.

The data center being built in Ulsan is like that. It draws electricity from two power generation companies, Ulsan GPS and SK Multi Utility. Both are SK affiliates. So, SK has both the company that makes electricity and the company that uses electricity in one family.

When selling stakes in this data center, some overseas investors reportedly wanted to buy stakes in the power generation companies as well, not just the data center. They feel uneasy if the company providing electricity to the data center they are investing in is owned by someone else. It's like someone else holding the leash.

This might be SKT's real moat. It's a much more tangible strength than anything related to AI. But no one talks about this. It's not interesting.

For reference, that stake sale ended up with KKR, a US private equity fund, as the preferred negotiator, and domestic managers splitting the remaining half. Originally, they considered handing it all to KKR, but they felt burdened by handing over an AI data center, which is treated as a national strategic asset, entirely to a foreign fund.

This is a very significant point. It's a signal that AI data centers are no longer just real estate but have become political assets. It also means that what happens in this field going forward won't be driven purely by economic logic. You get what I'm saying, right?

1. Why is a judge acting as a telecom CEO?

In October last year, a new representative came to SKT. Jung Jae-heon.

His background is a bit unusual. Born in 1968, graduated from Masan High School, went to Seoul National University Law School, passed the bar exam, and was a judge. He did it for 20 years. When he was a presiding judge at the Seoul Central District Court, he even handled the appeal for the Choi Soon-sil tablet PC case.

He is the first lawyer-turned-CEO in SKT's history.

He joined SKT as the head of legal affairs in 2020, and when SK Square was created in 2021, he went as a founding member and oversaw the investment side. He was effectively the Chief Financial Officer. Then, when the USIM incident broke out last year and the previous CEO stepped down, he took the seat.

Up to this point, the picture is obvious. Appointing a former judge to a company that caused trouble to settle lawsuits and regulations and quietly restore trust. A risk-management type CEO.

But the first message this man threw after taking office was that he would transition from telecom to an AI data center operator.

And now, he even serves as the private-sector co-chair of a national AI data center project involving both the government and the private sector. He even split the company organization into two: telecom and AI.

He was put there to clean up, but he's drawing the biggest picture.

How should we read this? It can be seen in two ways.

Looking at it positively, the key is that he was the person who oversaw investments at SK Square. He is someone who knows how this field works. Being a former lawyer, his sensitivity to contracts and regulatory risks is actually a weapon in a situation where they are deeply involved in government projects like now.

Looking at it cynically, a person who has never run a telecom business for 20 years is the CEO of a telecom company. That business that generates 95% of SKT's current revenue. AI data centers are still at 3%. They've essentially put a person optimized for a 3% new business as the head of a 95% business.

Of course, one could argue that telecom is a field where growth has stopped anyway, so it just needs management. But if that management were easy, the USIMs wouldn't have been hacked last year.

1. Of all times, the board is shaking now

The timing is truly cruel.

In July this year, the KOSPI fell 22% in one month. SKT was hit even harder. It pushed down from the 130,000 range to the 80,000 range.

There are mechanisms to forcibly stop trading if the market falls too quickly, and that happened several times this year alone. In late July, it happened two days in a row.

Why did it become like this? The trigger was the collapse of leverage products concentrated in semiconductor stocks, but the real reason lies beneath.

The question the market is asking has changed.

Until last year, it was "Are this quarter's results good?" Now it's "Can this AI investment continue at its current speed in the future?"

The fact that SK Hynix's stock price tanked despite a significant increase in profits because it didn't meet expectations is proof. It didn't fall because performance was bad; it fell even though it was good. It means the market has stopped looking at performance and started looking at sustainability.

This question aims exactly at the heart of SKT's data center story.

One more thing overlaps.

Anthropic privately filed a draft of its IPO documents with US securities authorities on June 1. There are rumors it will go public as early as October, and the underwriters are Morgan Stanley and Goldman Sachs. For reference, OpenAI postponed its IPO schedule to next year.

What this means is that a chunk worth more than 20% of SKT's market cap will be valued at market price every day starting from October.

Right now, because it's unlisted, they can comfortably add it as "roughly 4 trillion." But once it's listed, that 4 trillion becomes a number that fluctuates in real-time. If it explodes upward, it's a jackpot, but if it slips, that 4 trillion is the first thing to be cut from SKT's value.

And SKT doesn't seem to have any intention of selling. They had a chance to sell and leave in May, but they put in more instead.

You also have to look at interest rates. The Bank of Korea raised the base rate from 2.5% to 2.75% on July 16. It's the first hike in 3 years and 6 months, and all seven members of the Monetary Policy Committee voted in favor. The fact that inflation rose to the 3% range for two consecutive months was a big factor.

Telecom stocks are originally bought for dividends. People buy them because dividends are better than bank interest, but if interest rates rise, that calculation is shaken. Until last year, securities firms raised target prices with the logic that "telecom stocks will be revalued if interest rates go down," but half of that logic is already broken.

Oh, and the USIM hacking lawsuit. This isn't over yet.

More than 15,000 people have filed damage lawsuits, and they are ongoing. The company even filed a cancellation lawsuit, refusing to accept the record-scale fine imposed by the Personal Information Protection Commission. They rejected a mediation plan to compensate 300,000 KRW, and they also rejected a 100,000 KRW one.

Right now, the costs aren't being recorded. But once a judgment comes out, everything that was postponed will be processed at once. The problem is when it will explode, not if it will explode.

1. The Gates of Heaven and the Gates of Hell

The Gates of Heaven

Anthropic goes public in October and its value jumps further. The 4 trillion in SKT's wallet becomes 6 trillion, then 8 trillion. Since the stake value is recorded as a visible number every day, the market starts to value it more comfortably.

At the same time, Anthropic signs a real contract for SKT's data center. The moment numbers showing how many gigawatts will be used for how many years appear in a public disclosure, the data center becomes revenue, not a blueprint. From then on, the yardstick for an infrastructure company is applied, not a telecom stock. Since it's a business that receives rent regularly every year, they give it a generous value.

The electricity problem could also turn into an opportunity for SK. When others can't build because they can't get electricity, a house that has a power generation company as an affiliate can build. The more severe the bottleneck, the more the one who breaks the bottleneck takes it all.

The core telecom business isn't bad either. Subscribers have turned back to net growth, and dividends have been maintained at 830 KRW per share for two consecutive quarters. The company has promised to return more than half of its adjusted net income to shareholders until 2026. This means there is a supporting force from below.

Securities firms' target prices are set between 110,000 and 140,000 KRW. Since it's 88,000 KRW now, it looks like there's quite a bit of room.

But this is only from a few large domestic firms; the average target price including overseas firms is 99,000 KRW. 12% above the current stock price. Some called for a high of 150,000 KRW, and some called for 55,000 KRW.

For the same company, one person says 150,000 and another says 55,000. It's a three-fold difference. This is the real state of SKT right now. No one can determine the value of this company.

Let me add one more word here. Korean securities firms rarely write 'sell' reports. There are about four sell opinions in the total tally, but most of those are from overseas. So, it's natural that target prices are all above, and you shouldn't use that itself as a basis. Just be aware that domestic and overseas firms see the same company so differently.

The Gates of Hell

First scenario. Anthropic's IPO is delayed, or it falls after going public.

4 trillion becomes 2 trillion and evaporates from SKT's market cap. What remains is a telecom company whose growth has stopped, and it returns to the same value as KT. If you think about where the 6 trillion that rose in the meantime came from, it's obvious where it will go.

Second. The IPO goes well, but no data center contract comes out.

There's no progress for months from the "let's do well together" document. When they calculate the cost of procuring power, it's much more expensive than expected, so margins become thin. 15 gigawatts shrink to 5 gigawatts, and 5 gigawatts shrink back to 1 gigawatt. It's not the first time we've seen a blueprint cut in half in this field.

Third. This might be the most painful.

In this current circular structure, SKT is the weakest link.

Samsung and Hynix still have places to sell memory even if Anthropic shakes. Whether it's Nvidia or Google, there's a line of people who will buy. But the data center SKT is building is in Korea. How many customers are there who will enter a Korean data center in gigawatt units? You can count them on one hand. If one anchor customer leaves, there's no suitable replacement.

This means that if the circle is broken, this is the place where water pools first.

Fourth. Interest rates rise once more. Dividend attractiveness is cut again, and money that came in for dividends leaves.

Fifth. The hacking lawsuit judgment comes out just then. Postponed costs are recorded all at once, adding another negative factor to an already shaking board. But I don't think this is that big.

And these five might not come separately. If the AI field cools down, Anthropic's value falls, data center contracts are delayed, and interest rates can't be lowered because of inflation—and in the midst of that, the judgment comes out. Bad news usually comes in clusters.

Currently, the PER of this stock is 49. The average for the same industry is 28. However, this number is an illusion caused by profits being destroyed by the hack last year, so you shouldn't believe it as is. If you calculate with the next year's profit, it comes down to about 16. This is an ambiguous spot for a telecom stock—neither cheap nor expensive.

An ambiguous spot means that the answer doesn't come from numbers alone. The person who called for 150,000 and the person who called for 55,000 are looking at the same financial statements. Ultimately, it's a battle of stories.

As always, all of this is Jhonberman's personal speculation, and you have to take care of your own account, right?

P.S. If they're going to make 4 trillion by putting in 132.1 billion, why do they even do telecom? Because they collected that 132.1 billion from your phone bills.