Tesla has wound down its in-house Dojo supercomputer program and disbanded the team behind it, people familiar with the matter told Bloomberg. Peter Bannon, vice president of hardware design engineering and the project’s leader, is leaving the company after Elon Musk ordered the initiative halted. Roughly 20 engineers from Dojo recently left to form DensityAI, a data-center hardware startup, while remaining staff are being reassigned to other compute, data-center and silicon efforts inside Tesla. The decision ends a multiyear attempt to develop custom chips and wafer-scale processors aimed at training Tesla’s autonomous-driving and robotics models. Tesla will instead step up reliance on outside suppliers, using Nvidia and AMD processors for AI compute and tapping Samsung to manufacture its next-generation AI5 and AI6 chips. “It doesn’t make sense for Tesla to divide its resources and scale two quite different AI chip designs,” Musk wrote on X, adding that future chips would be “excellent for inference and at least pretty good for training.” Dojo had been touted as a potential competitive moat—Morgan Stanley once suggested it could add as much as $500 billion to Tesla’s valuation—but its shutdown underscores broader restructuring at the automaker amid slowing electric-vehicle sales and executive turnover. Tesla shares slipped as much as 0.7% in pre-market trading after the news.