On February 10, The Information reported that xAI co-founder Tony Wu had resigned. Wu led reasoning at the company and reported directly to Elon Musk. The next day, Financial Times confirmed that co-founder Jimmy Ba was also leaving—the sixth of twelve founding members to depart. Hours later, Musk announced a company reorganization that "required parting ways with some people." In an all-hands meeting, he talked about building a factory on the moon.

The Count

xAI launched in July 2023 with twelve co-founders, recruited from DeepMind, Google Brain, Microsoft Research, and the University of Toronto. It was an impressive roster. Now half of them are gone.

co-founders have departed xAI's founding team

The departures span a year. Igor Babuschkin, previously of DeepMind, left in August 2025. Christian Szegedy, formerly of Google, departed sometime after. Greg Yang stepped back on January 21, citing a Lyme disease diagnosis. Engineer Sulaiman Ghori left on January 22—the day after appearing on a podcast where he described how xAI had skirted regulations while building its data centers. Then Wu. Then Ba.

Four of the six departures came in the last three months. TechCrunch counted five departures by February 10; Ba made it six the next day. And the FT noted that beyond the co-founders, "at least six researchers left in recent weeks."

The Sequence

What makes this exodus unusual isn't that co-founders leave startups—they do, routinely. It's that these departures accelerated after xAI became vastly more valuable on paper.

Read the sequence carefully. The co-founders didn't leave when xAI was struggling. They left when it was winning—by every conventional measure. A $20B raise. A $250B valuation. A $1.25 trillion merger. At the exact moment xAI's paper value went parabolic, its intellectual architects walked out the door.

The Reorganization

Musk's response to the departures was immediate and revealing. Within hours, he restructured xAI into four divisions: Grok (chatbot and voice), Coding, Imagine (media generation), and—notably—something called "Macrohard," which will build digital agents to run companies.

February 2026
xAI will be organized into four core areas: Grok's chatbot and voice products, Coding, Imagine, and Macrohard, which will build digital agents to run companies
Bloomberg

The naming is Musk in full Musk. "Macrohard"—a jab at Microsoft that doubles as a mission statement. The reorganization into product-focused divisions, announced the same week co-founders left, signals a shift from research lab to product factory. The people who built the foundational models aren't necessarily the people you need running product lines.

But the all-hands meeting went further. Musk told employees that xAI needs a factory on the moon to build AI satellites and a "massive catapult" to launch them into space. This is the merger logic made literal: SpaceX's orbital capabilities fused with xAI's compute ambitions, projected onto a timeline that starts with "moon factory."

The Pattern

I wrote on January 30 that the SpaceX-xAI merger represented an unprecedented consolidation—a vertically integrated empire spanning rockets, satellites, and artificial intelligence, all controlled by one person. The question then was what that consolidation would look like in practice.

Now we have an answer. It looks like departures.

This isn't unique to Musk. Research-driven AI companies routinely lose founders as they commercialize. OpenAI lost most of its original leadership, including chief scientist Ilya Sutskever. Anthropic itself was founded by people who left OpenAI. Google Brain lost key researchers to every major AI lab. The pattern is well-established: the people who build foundational capabilities and the people who scale products are often different people.

But xAI's version of this pattern is compressed and extreme. Most AI companies took years to transition from research to product. xAI is doing it in months—and doing it inside a merger that subordinates the AI company to a rocket company. When Greg Yang left to focus on health, that was personal. When an engineer left after revealing regulatory shortcuts on a podcast, that was foreseeable. But when your head of reasoning resigns right after a $1.25 trillion merger? That's structural.

The Burn

There's a financial dimension that makes this departure pattern especially striking. xAI is burning approximately $1B per month. Bloomberg reported a $1.46B net loss in Q3 alone, up from $1B in Q1—losses accelerating even as the company raised $20B.

xAI's burn rate
SpaceX-xAI combined valuation

A company burning a billion dollars a month can't afford to be a research lab. It has to be a product company. And the merger with SpaceX—which generates real revenue from launch contracts and Starlink subscriptions—makes the commercial imperative even more urgent. SpaceX investors didn't sign up to subsidize open-ended AI research. They signed up for an orbital data center technical breakthrough and the biggest IPO in history.

The founders who joined to build frontier AI models are now inside a conglomerate whose near-term priority is commercializing chatbots, coding tools, and enterprise agents. For some, that's a feature. For others, it's a reason to leave.

The Destination

Where are xAI's departing researchers going? The Humans& founding is instructive. A startup built by ex-Anthropic, xAI, and Google staff, raising $480M at a $4.48B valuation on day one. The pitch: "collaborative AI." The subtext: the people who left the major labs think they can do something different from what the major labs are building.

This is how the AI talent market works in 2026. You co-found a frontier lab, build the models, watch the company commercialize and consolidate, then leave to start something new at a multi-billion-dollar valuation. The founders aren't losing—they're graduating into their own ventures, armed with the technical knowledge and the network to raise hundreds of millions on reputation alone.

Which means xAI's loss may not be permanent damage. The models are built. The infrastructure exists. The Colossus data centers in Memphis aren't going anywhere. What's leaving is vision, not machinery.

But vision is what separates a frontier AI lab from a product company. And the gap between those two things—between building something new and scaling something built—is exactly the gap the departing co-founders seem to be expressing with their feet.

The Irony

On January 30, I called the SpaceX-xAI merger "The Unification." It was the right word for what Musk was doing: merging his companies into a single entity, concentrating control, preparing for the biggest IPO in history.

Eleven days later, the unification is producing its opposite. The biggest corporate merger in American history didn't retain the people who built the company. It accelerated their departure. Musk's response—reorganize into product divisions, talk about moon factories, "part ways with some people"—suggests he sees the departures not as a crisis but as a natural transition.

Maybe he's right. Product companies don't need twelve co-founders. They need product managers, engineers, and commercial operators. The research phase is over. The scaling phase has begun.

But there's a reason the word for this is "dissolution." Something that was whole is coming apart—not because it failed, but because it succeeded into something different from what its founders set out to build.