META's story has changed: AI spending made free cash flow negative, it is no longer capital-light, and dilution risk is rising while reward is not. He sees it as overvalued, trimmed 4.5% to 3.5% of his portfolio, will gradually take profit above 4% rather than sell all, and may be wrong if AI succeeds.

When I bought Meta in 2022, it was because of the high free cash flow that the company was generating, and the market was focusing on things that it should not have been focusing on.

But today, when we look at the last quarter, this is the first quarter ever since the company went public that the free cash flow is negative.

The story of the company has changed. They are making bigger and bigger investments in AI, which, of course, will cost them more and more money. And now they will potentially even dilute owners, dilute shareholders.

And I have to ask myself the question, as an owner of Meta, whether I should be selling or not. I have sold Meta in the past because the first time I bought it was in 2016, and it was a mistake to sell.

It would have been better if I’d just held Meta. When the price got cheaper, I’d buy more and just keep holding Meta, because in 2019, the stock price crashed. In 2022, it crashed again.

These were great times to buy. I bought, but I always took profits.

So let's look at these earnings. What are the things that are great in these earnings, and what are the things I don't like? There's only one thing which is great. Revenues are growing very rapidly.

This is because of the main business, and AI is contributing to this. AI is really making the ads business better.

And now there is Muse. Muse AI, which has grown to 10 million customers, 10 million downloads in the US and Canada, which are high-income countries. This is potentially going into other countries.

But will Meta be making that much money on this product? That's the question.

They say that they are going to monetize it through transaction fees. This is for most people who are not going to pay for Meta AI, just like most people don't pay for ChatGPT and other AI tools.

And for people who pay, of course, they're going to pay $16 a month or $80 a month or much more than that, based on how much they're using.

But just like we learned earlier this week that Anthropic lost over $80 billion for the first half of the year, this is the same thing that is going to happen to OpenAI and even Meta, because these AI tools cost a lot of money for them to provide.

And as long as they are free or at such a low price, these companies are always going to be losing money.

Meta has an advantage in terms of revenues because they have all these apps—Facebook, Instagram, WhatsApp—where they can quickly propagate Muse AI, reach billions of people quickly.

But if these people really successfully use Meta AI, Muse AI, without generating the income—not just revenues, income—for Meta, then it is a money-losing business.

Meta said that the compute cost will go down next year. It is going down. It will keep going down. But people are going to find new ways to use these products, and to remain competitive, they have to keep updating the semiconductors they have invested in, get better products, and it costs money.

Meta, when I first invested in it, it was because it was a capital-light business, which is no longer the case. And that's why they have taken debt in order to increase their capital expenditures.

Now they are suggesting that they will potentially have stock dilution. They are creating new shares. The new shares are not causing share dilution by themselves—the Class C shares.

It's just that they have zero voting power, and this will eventually facilitate share dilution, in order not to lower the voting rights of Mark Zuckerberg, so that he can maintain control of the company.

So he will maintain control of the company, spend as much money as he wants, and the owners who don't have that much power in the company are going to be diluted. This is unfortunate, and in a way, it has already happened.

You look at the metaverse, Reality Labs. They spent so much money there, and now they say that they are not going to be losing $20 billion a year. It's going to be $15 billion a year because the investment is moving into AI.

So we have already lost over $100 billion, and now we are being told, as owners, that we will lose less on that, but also we'll be spending that money somewhere else. It's not the money that is coming back to us.

Because dilution is not the same as taking debt. If a company takes $10 billion of debt, you expect, okay, maybe it is going to make, let's say, $10 billion just to break even for owners.

In a way, it's not that bad because once the debt is repaid, we can move on. But it is not the same with dilution because it is forever. You always have to give part of your profits to these new owners.

Even if they don't have any voting rights, they are entitled to the profits. That's the issue with dilution.

So, what am I going to do now? Actually, I already trimmed some of my Meta last week. It used to be 4.5% of our portfolio. Now, it's 3.5%. It's a small position. I'm not so worried about it.

Will I sell everything today? No, not at all. Because the risk is increasing, the way I see it. But I am not able to quantify that risk yet. And maybe it is going to be successful.

Maybe I'm going to be wrong on that. It will be a mistake to sell.

So I will gradually take profit if it goes, uh, let's say, beyond 4% of my portfolio. But for today, I'm not selling anything. I will trim if it gets more expensive because it is overvalued.

But at the same time, I don't want to sell everything just right now.

It's just that this is something very hard to know, if it will work or not. Unlike the metaverse, where maybe it was not going to work, and, uh, yeah, we lose money. Mark Zuckerberg was thinking 200 years in the future.

It was too early. AI, this is something happening now. Meta, in a way, has to win. But are they doing the right thing? Nobody can say that for sure.

But Meta is still fighting to create the best model, and AI agents by themselves, they cost more money, more compute, more power compared to just a large language model like ChatGPT.

And there are costs that we are not really seeing yet. It's off-balance-sheet. They have liabilities that are off-balance-sheet. They have, let's say, in 2030—they say to such companies that we are going to build data centers, but it's not going to be built right now.

It's in 2030. So they cannot put that as an asset or a liability on the balance sheet. But they have a commitment that in 2030, they have to spend that much money.

And we are not even going to talk about the fact that they paid a huge settlement. Yes, some of you may say that it's not a cash outflow. It's just, uh, on the income statement.

But they still have to pay every year, even though on the income statement, it is a one-time expense.

So this is a business that is becoming more and more risky. It's not becoming that risky, like, let's say, Fannie Mae and Freddie Mac, when Warren Buffett realized that he had to sell everything.

It's not that risky that today I'm looking at the company and I have to sell everything. But the risk is increasing, and the reward is not—the potential reward is not—increasing that much.

But for the time being, I'm going to trim if it gets more expensive than that, take profit gradually. There is no rush to sell everything.

Maybe—I hope I'm wrong because it's a wonderful company. It used to be. I hope they win in AI, and then, uh, yeah, 3% of my portfolio. It's not that bad.

So how this is related to Meta is that if Meta really enters the hotel booking business through Meta AI, they will also have to handle these issues.

So the fines they are paying today, that's very little compared to what may happen if these things are not considered. That's another risk with Meta today.