Musk's xAI core team undergoes a major purge; 9 out of 11 founders have left! Why are top AI brains fleeing?

（Christian Szegedy (Image source: Internet)

Christian Szegedy will leave in February 2025.

Position:Former senior scientist at Google proposed the Inception network architecture.

Background: PhD in Mathematics from the University of BonnHe was a senior scientist at Google, renowned in the field of deep learning, and a major contributor to the Inception network architecture.

Igor Babuschkin will leave the company in August 2025.

Position:One of Musk's "right-hand men"

background:He has extensive experience in AI security and research and was an early core member of xAI.

(Greg Yang, image source: internet)

Greg Yang (杨格) will leave the company in January 2026.

Position:Grok Core Architect

Reason for leaving:Diagnosed with Lyme disease, which he attributed to his long hours of intense work at xAI, he decided to step back from day-to-day operations, become an informal consultant, and focus on his health.

background:During his time at xAI and Microsoft Research, he was dedicated to research on mathematical foundations and deep learning theory, and was one of Grok's principal architects.

(Image source: X)

Tony Wu (吴宇怀) resigned on February 10, 2026.

Position:Co-founder, a core expert in mathematical reasoning and symbolic AI, responsible for building the reasoning capabilities of Grok's large models.

background:Previously worked at Google, OpenAI, and other institutions, with in-depth expertise in automated reasoning and theorem proving.

At the Gori3 launch event in March 2025, the three people from left to right (Babushkin, Jimmy Bah, and Yu-Huai Wu) have all since left the company.

Jimmy Ba left the company on February 11, 2026.

Position:Co-founder, a renowned scientist who proposed the Adam optimization algorithm, was responsible for the development of AI tutoring functions and the Grok 4 model, reported directly to Musk, and was responsible for most of the company's business operations.

background:Known for proposing the Adam optimization algorithm, he has made outstanding contributions to the field of AI optimization and once held a key position at xAI, but many of his core responsibilities were later split up.

Toby Pohlen will leave the company at the end of February 2026.

Position:A former DeepMind researcher who led Project Macrohard (which aimed to create an AI agent capable of doing anything an office worker could do on a computer).

Reason for leaving:He left the company after only 16 days in office, the specific reasons for which are unknown, but the Macrohard project was temporarily suspended.

background:He was a researcher at DeepMind and has expertise in the field of AI agents.

Zihang Dai (戴子航) resigned in March 2026.

Position:Co-founder and one of the most senior members of the technical team, he previously worked at Google and holds a PhD in Computer Science from Carnegie Mellon University.

Reason for leaving:He did not explicitly state this, but he has publicly admitted that xAI has fallen behind in programming capabilities.

Background: Undergraduate degree from Tsinghua University.He holds a PhD in Computer Science from Carnegie Mellon University and previously worked at Google, where he was a core member of the xAI technology team.

Guodong Zhang resigned in March 2026.

Position:Co-founder, responsible for the two core product lines Grok Code and Grok Imagine, reporting directly to Musk.

background:The core team member of xAI is mainly responsible for machine learning optimization and training, and has professional experience covering top institutions such as Microsoft Research Asia, Google Brain, and DeepMind.

Among these departing employees are many Chinese, such as Wu Yuhuai, Jimmy Bach, Yang Ge, Dai Zihang, and Zhang Guodong. They are all leaders in the AI field, possessing top-notch academic backgrounds and industry experience. Their departure is undoubtedly a huge loss for xAI.

Translation: Over the past few years, many talented individuals have been rejected when applying to join xAI, without even receiving an interview. I sincerely apologize for this. I am working with the head of recruitment to review the company's interview records and will reconnect with candidates who demonstrated potential at that time.

Faced with such a large-scale loss of talent, Musk responded, "xAI was not built correctly in the first place, so it is now being rebuilt from the ground up." He also added that Tesla had experienced something similar in the past and unusually expressed deep regret for not hiring talented candidates in the past.

These words reveal some of the characteristics of Musk's "Musk management style".

1. Extreme exploitation and the "war room" culture

xAI operates around the clock, at a battle-like pace, under immense pressure. Former CFO Mike Liberatore once worked 120 hours a week, and General Counsel Robert Kiel described it as "working like a coal shovel." This "war room" culture may bring amazing explosive growth in the early stages of a startup, but in the long run, it puts a huge strain on the physical and mental health of employees.

(Image source: X)

2. The conflict between a flat organizational structure and de facto bureaucracy

A former employee, Benjamin De Kraker, revealed that xAI emphasized a "flat structure" and "proactive approach" when he joined, but when he tried to solicit suggestions for improvement from Grok through social media, he was severely criticized by his supervisor and his account was suspended. He bluntly stated, "This is one of the places with the most typical big-company mentality I've ever worked in." This inconsistency between words and actions severely dampened employee enthusiasm and creativity.

3. CEO-driven approach versus divergent technology roadmaps

Musk wields absolute power within the company, and his personal will often overrides the judgment of the technical team. One netizen commented, "The interviewer almost bluntly stated that certain parts of the company's models must align with Musk's ideas, and Musk might personally call at any time, demanding the team immediately change direction." In AI research, a field requiring long-term accumulation and trial and error, this "push-in" management style and excessive reliance on the CEO's personal will can lead to frequent shifts in the technological roadmap, and even miss crucial breakthroughs.

"Musk excels at engineering and productization, but the first half of AI research is more like scientific research, which requires patience and room for trial and error."

4. Resources are not lacking, but the conversion efficiency remains to be seen.

Musk has invested heavily in xAI: a large data center in Memphis, 200,000 dedicated AI chips, and data resources provided by the X platform. However, commentators have pointed out that "xAI has never lacked resources, but rather the ability to translate those resources into product and organizational efficiency." This illustrates that even with financial resources and top-tier computing power, it's difficult to stand out in the fierce AI competition if management and organizational efficiency are lacking.

(Image source: Internet)

The AI industry is currently in a "hard takeoff" phase, with an incredibly rapid pace of iteration that is dizzying. Even Elon Musk himself has admitted that "we're really struggling to keep up." The predicament of xAI reminds us that even with the most cutting-edge technology and abundant resources, a company with a chaotic organizational structure, inefficient management, and constant internal conflicts will find it difficult to translate potential into tangible results.