Elon Musk’s AI venture xAI has suffered a dramatic exodus: the bulk of its founding team has left and Musk has publicly acknowledged that the firm was not properly built, announcing a complete rebuild from the ground up. Of the original founding group, only Manuel Kroiss and Ross Nordeen remain on staff, while a string of high‑profile departures this year included Guodong Zhang, Haotian Liu and Zihang Dai alongside several senior engineers and researchers.
The departures accelerated in 2026 after internal disputes over technical direction and management style. Key projects have been disrupted: Macrohard, an ambitious AI agent initiative also branded internally as Digital Optimus, has been reported stalled after successive leadership losses, and an internal annotation effort involving more than 600 contractors was abruptly halted when researchers found pervasive model defects.
The personnel crisis comes as Musk doubled down on integrating xAI with his other companies. Tesla disclosed a $2 billion investment in xAI in January and SpaceX moved to fold xAI under its ownership, a reorganisation that industry reports valued at roughly $1.25 trillion on a post‑deal basis. Musk’s public message — that xAI must be rebuilt from scratch, echoing Tesla’s own chaotic early years — undercuts the narrative he pitched to investors and has already become fodder for legal challenges from Tesla shareholders.
Technical disagreements appear central. xAI teams reportedly favoured workflows that rely on static screenshots and staged interactions for agent training, while Tesla’s systems emphasise continuous, real‑time video streams and closed‑loop control similar to its FSD approach. That divergence has prompted pressure from Musk for xAI projects to adopt Tesla’s engineering model and has led to tensions over staffing, data pipelines and research methods.
The talent flight has broader consequences for the AI sector. Several departing engineers are said to be spinning up new ventures, potentially seeding direct competitors, and the loss of senior researchers risks slowing progress on flagship features such as autonomous software agents and multimodal coding tools. At the same time, the apparent handover of certain projects to Tesla and SpaceX suggests Musk is consolidating critical capabilities within his better‑resourced firms rather than trusting xAI to remain an independent R&D outpost.
For investors and regulators the episode raises uncomfortable questions about governance and diligence. Tesla shareholders have already sued in Delaware, alleging that the creation of xAI diverted personnel, GPU allocations and strategy away from Tesla in ways that benefited Musk’s private interests. Musk’s admission of foundational flaws in xAI, coming after public disclosures of the Tesla investment and SpaceX acquisition, amplifies those claims and injects new uncertainty into ongoing litigation.
xAI’s next steps are uncertain. A public pronouncement of a ground‑up rebuild gives Musk latitude to reconfigure teams and processes, but it also signals that earlier technical assumptions and management choices have failed. Whether consolidation into Tesla and SpaceX will accelerate development or simply bury xAI as a captive R&D lab depends on whether the reorganisation can retain or replace departing talent and resolve the deep technical disagreements that precipitated the exodus.