xAI's Cofounder Exodus: What Losing Half Your Founding Team Means

From The Bit Baker newsletter — February 14, 2026

When Elon Musk launched xAI in July 2023, he put together 12 researchers pulled from the top of the AI talent pool: OpenAI, Google DeepMind, Microsoft, the University of Toronto. The mission statement was equal parts ambitious and hazy -- "to understand the true nature of the universe." Less than three years in, half the founding team has walked out.

The latest exits came fast. Reasoning team lead Tony Wu posted his departure on February 10. Jimmy Ba, who ran research and safety, followed a day later. At least nine other engineers left during the same week -- former OpenAI researcher Vahid Kazemi, Simon Zhai, Shayan Salehian, Andrew Ma among them. That brings the total to six of the original 12 cofounders gone.

Musk tackled it head-on at an all-hands. He called it a deliberate reorganization meant "to improve speed of execution," framing the departures as "push, not pull." Departing staff offered a different read, pointing to frustrations about autonomy and where the company was headed.

Why It Matters

These aren't random engineers. Jimmy Ba trained under Geoffrey Hinton -- one of deep learning's three foundational figures -- and his work directly shaped how Grok handles safety and alignment. Tony Wu ran the reasoning team, which is the single most competitive capability in today's AI race. Losing both in the span of 48 hours goes well beyond normal turnover.

The exits track a pattern that's been building for months. Igor Babuschkin, the engineer most responsible for building Grok, left in July 2025 to launch a venture capital firm. Christian Szegedy, who spent 12 years at Google before joining xAI, departed in February 2025 -- first to Morph Labs, then to start Math Inc. Kyle Kosic was the first cofounder out the door, heading back to OpenAI in April 2024. Each exit carried away institutional memory you can't backfill with a job listing.

Timing tells its own story. xAI recently completed an all-stock merger with SpaceX, pricing SpaceX at about $1 trillion and xAI at $250 billion. That followed xAI's acquisition of X (formerly Twitter) in an all-stock deal last March. Both moves look like groundwork for a potential $1.5 trillion SpaceX IPO later this year.

The Bigger Picture

Cofounders leaving startups isn't rare. The speed and concentration at xAI, though, is striking -- and it reveals something about how the company actually operates.

Musk runs his companies with aggressive timelines, minimal hierarchy, and constant pivoting. That works in hardware businesses where the founder's vision is the product roadmap. At a research-driven AI lab, it can collide with the culture that draws top scientists. Ba and Wu came from academic settings where independence and intellectual freedom aren't perks -- they're baseline expectations. A "move fast, I decide" environment grinds against people who are used to running their own programs.

The SpaceX merger piles on another set of complications. Folding a 300-person AI startup into a 13,000-person aerospace company brings bureaucratic overhead, culture shock, and the inevitable question of who's actually steering. Some of the departing engineers pointed to exactly these frictions.

Competitively, xAI has the money to absorb individual losses. Cofounders Guodong Zhang and Manuel Kroiss are still there, and the war chest is enormous. But researchers at this tier are scarce, and they don't leave empty-handed -- they take ideas, unpublished work, and entire professional networks with them. Kosic returning to OpenAI is the most direct example of one lab's loss feeding a rival's bench.

Then there's the financial angle. Musk installed Anthony Armstrong -- a former Morgan Stanley banker who'd been advising the White House -- as CFO in October 2025. That's a textbook pre-IPO hire. If xAI is now optimizing for financial milestones over research output, the departure of its scientific leaders might not be a problem management is trying to solve. It might be a consequence they've accepted.

What to Watch

• Grok's next update. Quality and timing will show whether the talent drain has slowed xAI's research pipeline. Any gap compared to OpenAI or Anthropic's release cadence will be hard to miss.
• Where the cofounders end up. Multiple departures landing at the same competitor -- or banding together to start a new lab -- would signal a deeper rift. Babuschkin's VC firm could also fund projects that compete directly with xAI.
• The IPO clock. If SpaceX goes public on schedule, the exodus may fade into a footnote. If the IPO hits delays, the questions about xAI's stability get louder -- and that $250 billion valuation starts looking vulnerable.

References

1. Business Insider -- xAI key departures list
2. TechCrunch -- Musk suggests xAI exits were push not pull
3. Observer -- xAI founding member exodus
4. Fortune -- Half of xAI founding team has left
5. Observer -- Elon Musk launches xAI
6. Contrary Research -- xAI company profile
7. Wikipedia -- xAI (company)
8. The Register -- Musk xAI departures