Elon Musk's AI venture, xAI, is reportedly grappling with significant administrative and operational turmoil. Following its merger with SpaceX and subsequent rebranding as SpaceXAI, the company has seen a mass exodus of talent, including all 11 of its original co-founders.
The Claude Fixation
At the heart of the internal friction is Musk's reported obsession with Anthropic's Claude. Whenever the rival chatbot receives an update, Musk allegedly mandates that Grok match its capabilities. This focus is so pervasive that internal Slack channels were named after the competitor. Michael Nicolls, recently promoted to lead the company, emphasized in a staff memo that reaching Claude's performance levels is the primary mission.
Operational Failures and Disorganized Layoffs
Internal management is described as chaotic. In March, xAI aimed to reduce its workforce by 30%, but the process was so poorly handled that some employees were reportedly terminated without any official notification. Furthermore, recruitment has stalled; high-quality candidates were left in limbo because an understaffed HR department could not complete the necessary paperwork.
The 'White Flag' of Compute Sales
A striking paradox lies in xAI's infrastructure. Despite possessing massive computing power, the company was utilizing only 11% of its capacity as of April. In a move seen by some as a "white flag," xAI began selling its excess compute resources to its main rivals, including Anthropic and Google. While this generated billions in revenue, it directly empowered the very competitors Musk seeks to overtake.
"Our immediate goals are to reach Claude's performance," Nicolls reportedly wrote to staff.
Ethical Concerns and Internal Skepticism
Employee morale has been further dampened by controversies. The Grok chatbot was linked to the creation of non-consensual deepfake imagery, and promises to compensate employees for providing tax returns for AI training reportedly went unfulfilled. Some staff members have begun to question Musk's leadership, suggesting his experience in hardware and rocketry does not translate to the nuances of large language model development.
