Stocks set records all through August. The rally is setting investors up for a rough fall. The valuation of SpaceX (SPCX) matters because pushing it to higher levels makes the market more open to the Anthropic IPO, which may come soon. The numbers do not make sense. The market is starting to see this as Chinese models appear out of nowhere and turn out just as valuable as the high-priced US models. This looks like getting ahead of ourselves, similar to the tech bubble.
What a rough fall looks like: popular high-priced stocks take a real plunge. Too much money is crowded into the same trades. Retail is the late money, and they end up the bag holders.
The IPO Playbook
There is a known playbook with big IPOs. The stock gets run up, which makes it exciting. Then it sells into the index, or gets added to the index once all the institutions have gotten in early. With SpaceX and Tesla (TSLA), as soon as they were added to the index, the stocks fell 8 to 10% or more, because the institutional money was already there. Expect the same playbook with Anthropic and possibly OpenAI, if it reaches an IPO. Chip stocks and the other hyperscalers have had big runs. These crowded trades will unwind in the fall.
SpaceX's Numbers
SpaceX was selling off this morning after releasing Grok 4.6, its rival to OpenAI's ChatGPT and Anthropic's Claude. The stock is well up off its low of 104 but well below its high of 225.
With someone as rich as Elon Musk around it and such a small float, it takes only a tiny share of his wealth to buy call options and move the stock. For every option bought, the seller of that option has to buy 100 shares. That is an easy way to push a stock up without spending much, and he has plenty to spend. This likely happened with Tesla in the past and is happening in large part with SpaceX. Wall Street is probably doing supportive buying too, to make the market receptive for another big IPO payday. On the day of the SpaceX IPO, Wall Street bankers made $100 million.
To justify a $1.9 trillion valuation, the company has to be the most profitable company in the history of the world times two. The implied profits are twice as high as Apple's (AAPL) profits today. At the same time, revenues would have to be one and a half times higher than Amazon (AMZN), the company with the largest revenues in the world. The stock is priced to be twice as good as the greatest companies in the world, on both the top and bottom line. More than enough good news is priced in.
Avoid the Stock
The market magic trying to push the stock up to make room for more super-profitable IPOs is a head fake, and the valuation is far too high. SpaceX is a great company but still makes no money. Even if it does make money, will it make as much as the stock price already implies? That is the real question, and the answer is no.
Lockups keep rolling over. New unlocks happen every 2 to 3 weeks. December 8th is when the 180-day lockup period concludes, and on June 12th Elon Musk's entire stake unlocks.
Other AI stocks carry similarly high valuations, though not quite as extreme. SpaceX is an outlier in how high its expectations for future cash flows run. All these stocks are priced as if each will be the sole winner with near-monopolistic margins and profits. That is not going to happen. In every major revolution - electricity, steam power, the internet - there was a huge gold rush, lots of stocks got great valuations, and only a few remained standing. AI will be the same. They are not all going to be winners.
Red Flags
There are eight red flags on SpaceX. One: the company is unwilling to attest to the quality of its internal accounting controls. Every other major public company has its auditors confirm those controls are good. SpaceX could do it and chooses not to.
Another: the investors giving SpaceX money are really giving the private equity investors a bailout. Those PE investors made tons of money all the way up, pushed the value higher, and now jump off, passing the risk to equity and public investors. Public investors get no voting rights. It is a "tails you lose, heads I win" deal, and the risk-reward does not make sense. Plenty of profitable, underpriced stocks can give comparable long-term returns without this crazy risk.
OpenAI and Anthropic
OpenAI, which may be pushed to next year, and Anthropic are both valued around $900 billion, under a trillion, and both are overinflated.
The open-source models, especially from China, show there is no value in the model itself. There has never been value in models by themselves; they are only as good as their inputs. The same is true of the human brain: the smartest person in the world does no good without knowledge or experience. No model can perform alchemy on data. If the underlying data is not accurate and not put into a good taxonomy and ontology, the model has little value.
Model values are all converging because they are all based on the internet, and the internet is not a corpus of truth. It is a corpus of many opinions and bad data, with some truth that is hard to find. That is why a Chinese company can, in a couple of years, produce something about as good as Grok, OpenAI, or Claude. This shows that the AI-model layer of the stack is more of a commodity than a source of profits.
Pure-play AI model companies like OpenAI and Anthropic will carry very low valuations in the not-too-distant future. Their hope is that a bigger company, maybe Apple (AAPL) or another large firm, buys them and gives them a home with distribution, because the standalone business is not good.
