Linkedin Tesla is ending its in-house Dojo supercomputer program, shifting its AI chip development exclusively toward inference chips that run real-time AI models in vehicles and robotics. CEO Elon Musk confirmed the move on X, saying it “doesn’t make sense” to split resources between two chip designs. Dojo was designed to train Tesla’s autonomous driving software using video and sensor data collected from its EV fleet. While it was once viewed as a key differentiator in Tesla’s full self-driving ambitions, the company will now concentrate its AI efforts on its AI5 and AI6 inference chips, which Musk says will handle both real-time operation and some training functions. Sign up for CBT News’ daily newsletter and get the latest industry stories delivered straight to your inbox. The strategic pivot comes amid broader restructuring efforts at Tesla, which has laid off thousands, seen a wave of executive departures, and faced softening EV demand, particularly in Europe. The company is now realigning itself around artificial intelligence and robotics, and has inked a $16.5 billion deal with Samsung to help produce its next-generation chips. Here’s why it matters For those selling or competing with Tesla, this signals the company’s increasing commitment to deploying self-driving features more rapidly. Inference chips are what enable real-time autonomy, which means Tesla may accelerate updates to its FSD system, creating stronger selling points and differentiation in the EV and premium segments. It also reflects cost and operational consolidation during a time when Tesla’s price cuts and margin pressure are being felt across the auto industry. Key takeaways: