/News

Two engines didn't ignite. That's what happened, and that's what made the headline on a Friday in July. Starship's Flight 13 test was scrubbed, SpaceX said it would swap the engines and try again in a few days, and the stock, which was already sliding, dropped below $125, about ten dollars under the price of its offering the month before. Down 4.7% by mid-session. A Motley Fool piece on the subject made the sensible point that a delay of a few days does the company no harm, and that if you liked the stock the day before, there's no reason to like it less the day after. I agree. And that's exactly why this interests me, because if the scrub doesn't justify the drop, something else does, and almost nobody bothers to say what. It wasn't the rocket. It was the price. Or rather, the distance between the price and anything you can measure with a ruler. The following Monday the stock slid again, to under $123, and the same series of articles noted that even there each share cost about 84 times trailing twelve-month revenue, and an infinite multiple of the profits that don't exist yet. Eighty-four times. When someone tells me a company is worth that, my first reaction isn't to argue the number. It's to ask which story has to be true for the number to make sense, and how many of those stories have to be true at the same time. The timeline, because short memory is the best friend of anyone selling stock. SpaceX debuted on the Nasdaq on June 12, 2026, with the offering priced at $135, described in the coverage as the largest IPO in history. On day one the price hovered around $161. Four days later, on June 16, it touched $225.64, which is still the high. That same day the $60 billion all-stock purchase of Cursor, the AI coding company, was announced. Options started trading on the Wednesday, and within two sessions the stock had lost close to a fifth of its value. Two weeks after the debut, in a week when the Nasdaq Composite fell around 4.6%, the stock dropped 20.2% in five sessions. So by the time the engines failed in July, the stock had already lived a full cycle of euphoria and disappointment. The rocket was the last drop. Chosen because it was the drop within reach. Something I learned late, and by being wrong a lot of times: a chart only seems to tell a story because we add the caption afterwards. I remember spending an entire afternoon convinced that a traffic spike on one of my sites came from an article I'd published the day before. Only at the end of the day, looking at Search Console with its two or three days of lag, did I realize the spike came from an old page someone had dropped into a forum. The new article had nothing to do with it. I'd picked the nicest explanation. The analysts who explain SpaceX's daily wiggles are doing the same thing, with more money around and a better tie. Now, what is actually being valued there? Today's SpaceX isn't the rocket company a lot of people still picture. In February 2026 it absorbed Musk's xAI and began reporting in three segments: space, connectivity and AI. Space is the Falcons, Starship, the launch business. Connectivity is Starlink, with roughly 10,500 active satellites as of June. AI is Grok and the social network X, which came with xAI and brought the losses along. According to the filing, revenue in the first six months of 2026 was $12.5 billion against $8.1 billion in the same period of 2025, which is real growth. But in 2025 the company, already carrying xAI, lost $4.9 billion. I struggle to write about this without sounding sour, and maybe I am sour. But what I see is an excellent, proven business in Starlink, plus a launch business that dominates its market with almost no real competition, stapled to an AI business that burns cash at the speed AI businesses burn cash. Whoever buys the stock at this valuation is paying the premium for the first part and betting that the third part will pay for itself someday. And by announcing Cursor on the very day of the peak, SpaceX seemed to declare that the third part was still going to grow. Which brings me to the point I find most underrated in everything I read: the second-order effects of paying for an acquisition in stock. When a company buys another with its own paper, it's implicitly saying its paper is worth more than the market will tolerate. It works well when the price rises, because every share issued is worth more. When it falls, the same deal costs more shares for the same nominal value, or the seller ends up holding paper that was worth $225 on a Tuesday and $178 by Friday. I haven't read the exact terms of how the exchange price was fixed in the Cursor deal, so I don't know how much of the dilution is fixed and how much floats, and I distrust anyone who claims to know without having read the document. What I know is that the market read it as immediate dilution for people who bought on the open market, and that the reading was fast enough to be one of the causes of the June drop. Worse is the third floor of the reasoning, which nobody puts in headlines. A stock that rises a lot gives the company currency to buy anything that comes along. The company buys. The stock falls. The currency is worth less, which makes the next acquisition more expensive and the previous one harder to justify. It's a loop that feeds itself in both directions, and the people trading SPCX on rocket headlines aren't even looking at this mechanic. Let's talk about the lock-up. It dominates the August commentary and deserves a long paragraph, because it's a place where everyone has a firm opinion and almost nobody is entirely right. On August 6, two days after the first quarterly results, up to 911.5 million shares became eligible for sale, around $116 billion at the price of the time. Gary Black, who carries a lot of weight in that corner of the market, said retail investors buying the dip made no sense when the float was about to double on Thursday. The logic is solid: more potential supply, same demand. And then the stock went up. Instead of sinking on unlock day, it bounced roughly 35% off its August lows and ended up trading above the offering price. Why? The unlock concerned shares that could be sold, not shares that would be sold, and a good part of the fear was already in the price. There was also the Nasdaq-100 inclusion on July 7, which forces index funds with hundreds of billions in assets to buy. That makes for a slightly humbling lesson: mechanical demand, the passive kind, can matter more than the supply everyone was busy counting. None of that changes my mind, though. Musk's roughly 6.4 billion shares stay locked until June 12, 2027, and a significant part of what's coming in October and November, further employee tranches, isn't small. I get suspicious of rebounds that happen right before more supply arrives. I still can't tell if that's caution or bias. The second-quarter results were the episode that irritated me most. On August 4 the company published its first report as a listed company. Revenue beat expectations and the net loss was $541 million, against about a billion in the same quarter a year earlier. The stock rose more than 9% in the regular session. After hours it fell 7.3%, to $116.22. A beat, and the stock drops. The usual explanation is that it had run up beforehand, and sure, but it's also a good example of what happens when the market stops buying the narrative and starts wanting numbers. A revenue beat only impresses when the margin for error on the multiple is large. At 84 times sales, revenue can grow 37% in a year and the stock would still need to grow 37% every year for a decade to normalize the multiple. No quarterly result fixes that. Part of me thinks I'm being too hard. A company that cut its loss roughly in half and has Starlink generating recurring revenue isn't a scam. It isn't. It's a fantastic company at a price that demands perfection. Both things are true at once, and anyone who keeps only one of them is choosing what suits them. Analysts, by the way. The average price target, in the numbers I saw in August, was about $235, with the average rating at "Buy". But the range ran from $117 to $800. Eight hundred. When the range of professional forecasts spans a factor of nearly seven, the average stops being an estimate and becomes a decorative number. DZ Bank, initiating coverage, put a "Sell" on it with a $100 target and talked about crash risk because of massive future capital needs. Phillip Securities, also at "Sell", pointed to customer concentration and contracts that can be cancelled on just 90 days' notice. I find that argument more interesting than people seem to, because it contradicts the image of an impassable moat. A big customer with an easy exit isn't a fortress, it's a lease. A digression, and I apologize. At home I have an old board, from an era when the world was less tidy, that I bought for a few euros at an electronics market and never managed to get working the way I wanted. I spent a weekend trying to find out why the bus wouldn't respond, and in the end the answer was a resistor I had swapped myself for the wrong one. I think of this every time someone says the problem "has to be" something sophisticated. It rarely is. It's usually the wrong resistor. Back to it. Maybe I'm treating the Flight 13 scrub with too much contempt. I'll contradict myself a little. What a scrub shows, beyond the delay, is that the Starship schedule is part of the product. How much of the valuation rests on the promise that Starship will cut launch cost per kilo to a fraction of today's, and put into orbit the AI infrastructure the company needs to train its models? In June the company unveiled the Gigasat factory, around eleven million square feet, with a goal of one gigawatt a year of space-based compute by late 2027. If that's the plan, then every week of Starship delay is a week of delay on a valuation that depends on everything running on time. A rational market doesn't need to react violently. But a market paying 84 times sales is, by definition, also paying for the absence of delays. So yes, the price fell because of the rocket. Just not for the reason the headline suggests. There's also the question of options and short selling, which one day I'll have to study properly. Since June 17, anyone who wants to bet against SPCX has a practical instrument for it, and from then on the price had a second voice, a louder one. In September I read that the short volume ratio had hit 59.5, treated as a sign of growing bets against the stock. I'm very suspicious of that indicator. It measures selling volume, not open positions, and there are market makers short-selling for reasons that have nothing to do with an opinion. But because the number appears in every summary, it ends up working as a self-fulfilling prophecy, and that's already a fact about the market, even if the data is weak. One psychological detail that seems very human to me. The $135 offering price looked like a floor, and it wasn't. In a way it's almost comical: a price with no direct relation to the company's value, chosen by a committee of banks, turned into a symbol of protection by people who weren't on the committee. When the stock lost that level in July, the next low came near $110, a drop of almost 51% from the June 16 high. Anyone who bought at the debut, or the week after, spent weeks underwater, and that type of shareholder, the retail investor who came in because of the noise, is the one who sells fastest on the first recovery. There's another third-order effect. The next rally runs into the supply of people who only want to get out at breakeven. And then it turned. In September the stock was around $142, and it closed the month at $150.86. On October 1 it touched $148.07 and from there climbed hard, 7.63% in a single day on October 5, closing at $171.09 on close to 135 million shares traded. Morgan Stanley reiterated its bullish stance. On Friday, October 9, the close was $162.57. I don't have a solid thesis on why, and anyone who has one with great certainty is probably drawing the target after firing the arrow. What I can say is that in a month the stock went from lows to the highest levels since June, and the only visible thing that changed was the mood. Mood isn't a valuation model. It's a variable, but it isn't a model. I noticed there's a group of commentators, the ones writing "AI predictions" for the end of October, giving targets from $147 to $162, all very tidy. It's laughable. SPCX has had implied volatility that touched around 115%, and people are asking a language model for a number three weeks out. If I were the author of that text I'd be ashamed, and I'm saying so here, where I happen to be the only one writing and nobody can contradict me. There's also another tranche of shares being unlocked later this month, and as far as I know SpaceX hadn't announced, as of October 11, the date of its third-quarter results, with estimates pointing to November. I've always thought the most dangerous moment for a stock that has climbed almost 18% in six days is the day when more supply arrives and less news does. But I don't know. It could easily go the other way, with the supply absorbed as in August, and me left explaining why I was wrong. What stays with me, at this distance, is something smaller than the valuation. I've spent years looking at traffic, user and revenue charts for sites of my own, asking myself what separates a curve that rises because the product is good from a curve that rises because someone shared a link. You almost always find out months later. With SPCX it's the same, except there are billions of dollars behind every doubt. A rocket that fails to light two engines gives you about as much information as a spike from a forum post. Almost none. What's left is to see what the market does when the next data point actually matters.