Over the years, I have learned never to doubt Elon Musk. I spent a long stretch of my career covering tech, and for most of it, the smart take on Musk was that he was a showman one bad quarter away from ruin. Short sellers bet on it, and a good chunk of the tech press wrote it with real enthusiasm. Time after time, Musk made them look silly.
So I don't write this lightly. But I think there's finally real reason to doubt him.
The Cybertruck isn't even the first casualty this year.
Before we get to the doubts, however, it's worth remembering just how many times Musk took a bet that looked insane and cashed it.
A long track record of proving people wrong
Start with electric cars. In 2008, Tesla was a niche company selling Lotus-based Roadsters to rich Californians, and that year it was on the verge of bankruptcy. Musk poured in his own fortune to keep it alive. The Model S followed in 2012, and in 2023 the Model Y was the best-selling vehicle in the world, beating the Toyota RAV4. The rest of the auto industry has spent years playing catch-up to a company it once treated as a joke.
Then there's rocketry. SpaceX's Falcon 1 failed on its first three launches. "If the 4th launch had failed, SpaceX would not exist," Musk said later. It reached orbit in September 2008. In 2015, SpaceX landed a Falcon 9 booster upright, which plenty of people in the aerospace world thought couldn't be done economically. Today, SpaceX dominates the launch business, and Starlink had 13.6 million subscribers at the end of September.
And then Twitter, the one everyone was sure would sink him. Musk paid $44 billion for it in 2022, then cut roughly 80% of the staff. Advertisers fled. The site was supposed to collapse under its own weight. It didn't. Musk folded X into xAI in 2025, SpaceX absorbed xAI in a merger announced this February, and in June SpaceX went public in the largest IPO in history, raising $75 billion and finishing its first day worth more than $2 trillion. That dying website Musk supposedly overpaid for is now a piece of the most valuable company he has ever built.
That kind of résumé buys a lot of patience, but that patience isn’t endless.
The Cybertruck gets the axe
On Friday, Nov. 6, Musk announced that Tesla would be "winding down production of the Cybertruck at Gigafactory Texas around the middle of next year." The stated reason: to "make room for increased Cybercab production," Tesla's steering-wheel-free robotaxi.
The next day, Musk claimed on X that Cybertruck orders were "far more than we can make before dismantling the line" and said that Tesla would add an extra shift to fill them.
So Tesla is killing a truck that, according to its CEO, it can't build fast enough.
The sales figures don't back that up. Tesla doesn't break out Cybertruck sales, but Cox Automotive estimates it sold about 11,586 in the U.S. through the first nine months of 2026, according to Drive Tesla. Tesla's own shareholder deck lists more than 125,000 trucks per year of installed capacity at Giga Texas, and the original plan called for 250,000. Electrek estimates the truck is selling at roughly 12% of that plan. Kelley Blue Book figures Cybertruck sales fell 48% from 2024 to 2025. And according to S&P Global Mobility data cited by Electrek, Musk's own companies, mainly SpaceX, bought about 19% of the Cybertrucks registered in the U.S. in the last quarter of 2025.
This is the same truck Musk said had demand "off the charts," the one that supposedly racked up more than a million reservations. Those were refundable $100 reservations, which are cheap. A factory running at a fraction of capacity is not.
Musk's explanation kept shifting over the weekend, too. First it was about making room, which Electrek pointed out makes little sense in a factory with more than 10 million square feet. Then, on Saturday, came the most honest answer he has given: "The biggest thing most people fail to understand is that there are a very limited number of smart, dedicated people! They can only do so many things."
That's a remarkable thing for the CEO of Tesla to say out loud. And it lines up with an ugly fact: Of the 14 executives Tesla paraded on stage at its 2023 Investor Day as proof of its "significant bench strength," half are gone. Tesla has also lost the heads of its Cybertruck and Model Y programs and a string of leaders from Cybercab, the program Musk is now pulling people toward.
The Cybertruck isn't even the first casualty this year. In January, Tesla announced the end of the Model S and Model X, and the Fremont line that built them is being converted to make Optimus robots. By the middle of 2027, a company that had five consumer vehicle programs at the start of this year will be down to essentially one platform: the Model 3 and Model Y.
Grok Bot hits a wall
In August, SpaceX's AI arm launched Grok Bot, an AI agent that runs tasks on its own virtual computer in the background: research, writing, browsing, multistep jobs that keep going after you close the tab. It took off. On SpaceX's October earnings call, President Gwynne Shotwell said it had reached about 800,000 paying weekly users, and Musk said it was "the fastest growth of a product that I've ever seen in my life."
RELATED: Elon Musk's next target? America's self-entitled phone companies.
Then, around the start of November, users started running into walls. Weekly usage allowances were running out far faster than anyone expected. Some users reported their meters jumping from 56% to 100% within an hour. Others claimed their effective limits had been cut roughly in half. The company never published the old limits or the new ones and gave no warning.
When a user asked Musk on Nov. 1 whether the usage pool could be reset, he gave a blunt answer: "The cost of serving Grok Bot was unsustainably high relative to revenue."
The token problem
Every AI company runs on tokens, the chunks of text a model reads and writes. Every token costs compute, and compute costs money. A chatbot answering a question burns a modest number of them. An agent like Grok Bot, grinding away for an hour browsing the web and calling the model over and over, burns orders of magnitude more. Flat-rate subscriptions ($30, $100, $200 a month) don't come close to covering a heavy user, a problem the whole industry is wrestling with.
Then Musk made the math worse. On Oct. 6, he announced that Grok Bot would use "the best back end model for any given task, including Claude Opus 5.5, MidJourney, Suno and other leading APIs." In other words, some of the work done by SpaceX's AI agent would run on Anthropic's top model, paid for by the token, while subscribers kept paying a flat monthly fee. It's also a strange thing for an AI company to admit: that for a lot of jobs, a competitor's model beats its own.
Less than four weeks later, Musk was calling Grok Bot's serving costs unsustainable. SpaceX hasn't said the Anthropic bill is what broke the budget, and neither company has disclosed terms. But you don't need to be a CFO to see how paying a rival by the token to power an all-you-can-eat subscription could get out of hand in a hurry.
So SpaceX did the math, didn't like it, and quietly turned down the faucet on its hottest product. To be fair, it is not alone. OpenAI has reportedly tightened its own metering this year. But Musk's pitch has always been that his companies play by different rules, that he can simply outbuild everyone else. This is Musk's AI business getting caught by the same economics as everyone else and handling it the way everyone else does: by rationing.
So is he losing it?
The cynical read is that Musk is stretched across too many companies and too many fights, and it's starting to show. He's running SpaceX, which now includes xAI and X, plus Tesla, plus Neuralink and the Boring Company, while posting to X at a pace that would get most of us fired. Tesla's GAAP net income fell 43% year over year last quarter. A man can only be in so many places at once, and Musk himself just told us the same is true of his engineers.
The charitable read is that this is Musk doing what Musk does: cutting loose anything that isn't the next big bet and putting every chip on autonomy and AI. This year he retired the Model S, the car that made Tesla, to free up a factory for robots. He's never been sentimental about the past.
The difference is that the old bets were on technology somebody had already proven. Other people had built electric cars and orbital rockets. Twitter was already one of the biggest social networks in the world the day he bought it. The new bet is a robotaxi that has to beat Waymo, which still logs roughly seven times more driverless miles per week, and an AI product that, by Musk's own admission, costs too much to run relative to what it brings in.
Maybe he will pull it off again. He's made a career out of making people like me eat our words. This time, though, I wouldn't bet the farm on him.
Continue reading...
[H/T Blaze Media]