Key Takeaways
- Complete Exodus. All 11 original xAI co-founders have departed. The final two — Manuel Kroiss, who led pretraining, and Ross Nordeen, Musk’s “right-hand operator” — left this week as confirmed by Business Insider and TechCrunch.
- Musk Admits Failure. Following the departures, Musk posted on X that xAI “was not built right first time around, so is being rebuilt from the foundations up” — an unusually direct acknowledgement.
- SpaceX Merger Context. The exodus accelerated after SpaceX acquired xAI in February 2026 in an all-stock deal. The merger brought xAI, X (formerly Twitter), and SpaceX under one corporate umbrella, valued collectively at $1.25 trillion.
- The Talent Gap. The departed founders came from Google, DeepMind, Microsoft, and OpenAI. They are leaving into the most competitive AI talent market ever — where Meta is offering packages reportedly worth up to $300 million over four years.
The Complete Departure Timeline
Understanding the xAI co-founder exodus requires the full sequence. It was not a sudden collapse — it was an accelerating cascade over 14 months.
February 2025: Christian Szegedy, who came from Google, departs. The first signal of instability, noted at the time but not widely covered as a pattern.
August 2025: Igor Babuschkin leaves to launch a venture capital firm.
Mid-2024: Kyle Kosic departs and joins OpenAI — directly to a competitor.
January 2026: Greg Yang steps back from his role, citing health reasons including a battle with Lyme disease.
February 2, 2026: SpaceX acquires xAI in an all-stock deal. The merger valuation: SpaceX at $1 trillion, xAI at $250 billion, combined at $1.25 trillion.
February 10, 2026: Tony Wu, who led xAI’s reasoning team and was described as one of the most operationally central co-founders, announces his departure on X. His post: “It’s time for my next chapter.”
February 11, 2026: Jimmy Ba, who directed research and safety efforts, resigns within 24 hours of Wu’s announcement. Ba’s post: “Grateful to have helped cofound at the start.”
Late February 2026: Toby Pohlen follows.
March 2026: Zihang Dai and Guodong Zhang depart, per Business Insider reporting. Musk posts his public acknowledgement about xAI needing to be rebuilt.
March 27, 2026: Manuel Kroiss informs colleagues he is leaving.
March 28, 2026: Ross Nordeen, the last remaining co-founder, departs. TechCrunch confirms: all 11 original co-founders are now gone.
Direct Answer: Have all xAI co-founders left the company? Yes. As of March 28, 2026, all 11 original co-founders of xAI have departed the company. The final two to leave were Manuel Kroiss, who led the pretraining team, and Ross Nordeen, Musk’s right-hand operator, both of whom departed this week. The departures follow the SpaceX acquisition of xAI in February 2026. Elon Musk has publicly stated that xAI “was not built right the first time around” and is being rebuilt from its foundations.
Who They Were and Where They Came From
The original xAI founding team, assembled by Musk in 2023, was drawn from the world’s most competitive AI research organisations:
From Google/Google Brain: Christian Szegedy (one of the architects of modern neural network theory), Guodong Zhang, Greg Yang.
From DeepMind: Toby Pohlen, Kyle Kosic.
From OpenAI: Jimmy Ba (had worked on foundational alignment research).
From Microsoft Research: Zihang Dai.
From various: Tony Wu (reasoning), Igor Babuschkin (previously at OpenAI), Manuel Kroiss (pretraining), Ross Nordeen (operations, previously Tesla).
This was not a team assembled for optics. These were researchers and operators with track records at the most respected AI institutions in the world. The fact that all of them have now left — and that several moved directly to competitors — is a substantive signal about conditions inside xAI under Musk’s post-merger management.
Why It Happened: The Three Explanations
No single explanation accounts for all 11 departures, but three factors appear repeatedly across reporting.
1. The SpaceX Merger Changed the Company
The February acquisition brought xAI under SpaceX’s corporate structure. For researchers who joined xAI as an independent AI company focused on research and model development, the merger represented a fundamental change in what they were working on. SpaceX is a hardware and aerospace company. Its culture, priorities, and management cadence are different from an AI research lab. Several departures appear to have accelerated immediately after the merger announcement.
2. Musk’s Management Style in Research Environments
Musk’s management approach has produced extraordinary outcomes in hardware engineering — SpaceX’s rapid iteration, Tesla’s manufacturing scale. The same approach appears less effective in AI research, where the most valuable talent has abundant alternatives and low tolerance for instability.
Tony Wu’s departure in February came alongside reports of internal tensions over model performance demands. Jimmy Ba’s same-day resignation amplified the signal. When two senior researchers leave within 24 hours of each other, something structural has gone wrong.
Musk’s own admission — that xAI “was not built right first time around” — is worth taking seriously. This is not a leader who typically volunteers acknowledgements of failure. The departures appear to have created enough external pressure to produce an unusually candid public statement.
3. The AI Talent Market in 2026
The eleven researchers who built xAI are leaving into the most competitive AI talent market in history. Meta is reportedly offering packages worth up to $300 million over four years to retain top researchers. OpenAI, Google DeepMind, and Anthropic are all expanding their research teams aggressively.
Kyle Kosic went directly to OpenAI. Where the others land will reveal a great deal about the industry’s hiring priorities in 2026. For talent this calibre, finding a new role is not a concern. The question is which organisation benefits most from each departure.
What xAI Still Has
The loss of the founding research team is significant but does not render xAI inoperative. Several assets remain:
Colossus Supercomputer. The Memphis, Tennessee facility runs more than 200,000 NVIDIA H100 GPUs — one of the largest AI training clusters in the world. Infrastructure at this scale gives xAI meaningful capability even without its original research leadership.
Grok Distribution. Grok has access to X’s user base — hundreds of millions of accounts — as a built-in distribution channel. No other frontier AI lab has an equivalent consumer distribution mechanism.
SpaceX Integration. Access to SpaceX’s engineering talent, capital, and physical infrastructure creates operational resources that most AI startups cannot match.
Existing Models. Grok 4.20, released this week with a four-agent multi-reasoning architecture, demonstrates that xAI continues to ship product. The question is whether the research leadership rebuilding around that product can match what departed.
New Hiring. Musk has publicly stated he is recruiting from other AI firms, including two hires from coding startup Cursor this month. The rebuild is underway.
The Sovereignty and Governance Angle
The xAI exodus is a case study in a question that sovereign AI users care about: who actually controls the AI you are depending on?
xAI was founded by 11 researchers and Elon Musk. Of those 12 people, only Musk remains — and the company is simultaneously being restructured under SpaceX ownership, preparing for an IPO, and acknowledged by its founder to have been built incorrectly.
For users or organisations that chose Grok because of its capabilities, the relevant question is: what continuity exists between the Grok that was built by those 11 researchers and the Grok that will be built by the team assembled by Musk’s post-merger company?
The same question applies in a different form to any cloud AI dependency. Research leadership changes. Acquisitions happen. Products are restructured. Companies prepare for IPOs and shift their priorities. The only AI system whose governance you can be certain of is one you run yourself.
FAQ
Does this affect Grok’s availability or existing features? There is no indication of immediate disruption to Grok’s consumer or enterprise availability. xAI continues to ship updates — Grok 4.20 launched this week. The longer-term question is research capability and competitive positioning without the founding team.
Where are the departed co-founders going? Kyle Kosic is confirmed at OpenAI. Other destinations have not been publicly confirmed at the time of writing. The AI talent market in 2026 means all of them will have options.
Is the SpaceX IPO still happening? SpaceX is reportedly preparing for a public listing with a projected $1.75 trillion valuation. The integration of xAI into SpaceX is part of what analysts see as a strategy to strengthen the IPO valuation by adding AI capabilities to SpaceX’s aerospace and infrastructure story.
Does Tesla’s $2 billion xAI investment create a shareholder conflict? Tesla shareholders filed a lawsuit alleging that Musk directed shareholder capital into his own private venture. The lawsuit gained traction in March after Musk publicly acknowledged that xAI’s coding tools were not competitive with Anthropic’s Claude Code or OpenAI’s Codex. Legal proceedings are ongoing.
Should I continue using Grok? The co-founder departures are a governance and research trajectory concern, not an immediate safety or functionality issue. If you use Grok for specific capabilities it does well — real-time X platform data, certain reasoning tasks — those capabilities still exist. For sovereignty-focused users, the governance uncertainty reinforces the case for local AI alternatives wherever feasible.
Related Articles
- SoftBank Borrows $40 Billion to Bet on OpenAI: What the Loan Signals About an IPO
- Gemini 3 Deep Think vs Grok 4.20: Two Approaches to Frontier Reasoning
- Anthropic’s Claude Paid Subscriptions Have More Than Doubled in 2026
- Paperclip AI: Build a Sovereign AI Company With a CEO, CTO, and Full Agent Team
- OpenAI Kills Sora: The AI Video App That Never Found Its Audience
Sources & Further Reading
- MIT Technology Review — AI Section — In-depth coverage of AI research and industry trends
- arXiv AI Papers — Pre-print research papers on AI and machine learning
- EFF on AI — Civil liberties perspective on AI policy
