Author：Wall Street CN
Over the past week, xAI, Musk's AI company, has experienced a major personnel shake-up.
Half of the 12 co-founders have left, at least 9 engineers have publicly announced their departure, and even core executives responsible for inference and research security, Tony Wu and Jimmy Ba, have posted farewell messages, sparking discussion on the X platform.A trend of imitating "I left xAI".
Musk's response was very official: "This is an organizational restructuring that is necessary for xAI to scale up."
But the wording of the departing employees reveals a hidden meaning. Some said they wanted to accomplish big things with a small team, others felt "all AI labs are doing the same thing, it's too boring," and three even formed their own startup. So the question is, were these co-founders no longer a good fit for xAI, or is there a problem with xAI's direction?
Secrets hidden in the timeline
According to The New York Times, Musk gave an explanation at an all-hands meeting a few days ago.
Because we've reached a new scale, we're adjusting our organizational structure to make the company operate more efficiently at this size. In reality, when this happens, some people are better suited to the early stages of a company than the later stages.
In other words, the employees who left were not fast enough or well-suited to xAI.
A few hours later, he added on X:
xAI underwent restructuring a few days ago to improve execution speed. When a company grows rapidly, its organizational structure must evolve like a living organism. Unfortunately, this required parting ways with some people.
"Stop it, stop it," Musk used the word "unfortunate," but there was no real unfortunateness in his tone. He also directly recruited on BOSS Zhipin (a job search platform): "If you are interested in building an electromagnetic catapult on the moon, join xAI."
Looking at xAI's current list of departing employees, the most eye-catching departures are undoubtedly the two co-founders. Tony Wu, who was in charge of inference, posted a rather poetic statement upon leaving: "This is an era full of possibilities; a small team equipped with AI can move mountains and fill seas."
Jimmy Ba, head of research and security, put it more aggressively: "The recursive self-improvement loop is likely to be live within the next 12 months. It's time to recalibrate my gradients at a macro level. 2026 is going to be crazy."
There's another detail worth noting. Roland Gavrilescu, a former xAI engineer, left last year to found Nuraline, a platform for deployable AI agents. Then, on February 11th, he posted that he was leaving Nuraline to work on something new with other people who had left xAI.
On the surface, Musk attributes everything to the inevitability of scaling. But with multiple engineers following the co-founder, the situation is clearly more complex. Several key moments warrant attention.
"This is my last week at xAI," engineer Ayush Jaiswal said on February 6.
On February 7, former X employee Shayan Salehian resigned. He was previously responsible for product infrastructure and model post-training at xAI.
On February 9, technician Simon Zhai and co-founder Tony Wu announced their departure.
On February 10, co-founder Jimmy Ba, machine learning PhD Vahid Kazemi, multimodal project leader Hang Gao, and Macrohard founding team member Chace Lee, among others, all bid farewell.
This wave of resignations occurred at a time when xAI was at its most sensitive.
In the past 30 days, Grok Imagine has generated 6 billion images and 50 million videos every day, but it has also put xAI under regulatory scrutiny and even affected its merger with SpaceX. Against this backdrop, the collective departure of the co-founders is hard to say is just an organizational restructuring.
Furthermore, according to a person familiar with xAI's operations, Musk was dissatisfied with the data center expansion because the massive investment failed to give the AI models a sustainable competitive advantage. This frustration led to a new leadership structure and a major purge of personnel.
Four new teams and a high school student
At the company-wide meeting, Musk announced the four new teams that had been reorganized.
The first team was responsible for the Grok chatbot, including voice functionality.
The second team specializes in programming AI models, aiming to directly compete with Anthropic's Claude Code for large enterprise clients.
The third team worked on image generation, which is Grok Imagine.
The fourth team is the most magical: the Macrohard project.
To be fair, the name Macrohard itself is a mockery of Microsoft (Microsoft vs. Macrohard, one Microsoft, the other a giant). The goal is to "digitally simulate corporate organizations," allowing AI to do anything that any human with a computer can do.
"In the future, we should see rocket engines designed entirely by AI," said Toby Pohlen, the person in charge, at the meeting. So, can we expect AI to build rockets all by itself?
Another interesting personnel change is Diego Pasini. He graduated high school in 2023, won the xAI hackathon in 2024, and now heads the AI mentor team and Grokipedia. Having someone who graduated high school two years ago managing a project of this caliber just goes to show how Musk-like he is.
The all-hands meeting also revealed some X data. Musk said X now has 600 million monthly active users, compared to 237.8 million daily active users who could see ads when he acquired it in 2022. Nikita Bier, X's product manager, said that annual recurring revenue (ARR) from subscriptions has just surpassed $1 billion.
He simply said that more services, such as X Money banking functionality and a standalone chat app, would be added in the coming months, and that "daily active users will well exceed 1 billion." Well, another bold prediction.
Is the lunar AI factory for real?
The team reorganization is just the appetizer; what's truly worth paying attention to is Musk's lunar program.
According to The New York Times, Musk said at an all-hands meeting, "You have to go to the moon." His vision is to build a factory on the moon to produce AI satellites, and then use a giant electromagnetic catapult to launch the satellites into space. This would provide more computing power and energy support than data centers on Earth.
"It's hard to imagine what intelligence of that scale would think, but seeing it become a reality would be extremely exciting."
It sounds like science fiction, but Musk is serious. He even drew an even bigger picture: first, build a "self-sufficient city" on the moon, then go to Mars, and finally explore other star systems in search of extraterrestrial life.
Two former SpaceX executives, speaking anonymously, revealed that the moon has never been a priority for the company. However, in recent months, Musk has been posting extensively about the moon on SpaceX, clearly indicating a shift in focus.
All of this is predicated on the fact that xAI and SpaceX just announced their merger last week. The combined company is valued at $1.25 trillion, and SpaceX plans to IPO as early as June. Musk has tied the two companies together to advance the epic "space data center" project.
Getting back to the main point, xAI now has more than 1,000 employees. While losing a few co-founders won't be a major blow in the short term, in the AI field where talent is scarce and reputation is extremely important, this wave of departures is sending a dangerous signal.
At the end of the all-hands meeting, Musk said, "xAI is moving faster than any other company, and nobody is even close." That's true, but whether that speed is heading in the right direction is something only the remaining co-founders really know.
This article is sourced from:
