Tesla made a big change in how it works with AI chips. The company closed its Dojo AI supercomputer project. Now, Tesla will focus only on special chips called inference chips. These chips help make real-time decisions for self-driving cars and robots.
What Happened?
Tesla’s supercomputer project, called Dojo, was shut down. The Dojo team was disbanded and its leader, Peter Bannon, left the company. Elon Musk said Tesla should not split its effort between two different chip designs. Instead, the company will pour all effort into its Tesla AI5 and AI6 chips. These are made for quick, real-time decisions and are good at training too.(Reuters, Tom's Hardware, Statesman)
Why Tesla Made the Change
Tesla decided it wouldn’t make sense to build two different chip systems:
-
Dojo was made for training—teaching AI with huge amounts of video and sensor data.
-
The new AI5 and AI6 chips are for inference—making decisions fast inside cars or robots.(Applying AI, Reuters)
Elon Musk explained that it is better to focus on one chip platform. He said the AI5 and AI6 chips will be “excellent for inference and at least pretty good for training.”(Inc.com, Reuters)
What Was Dojo?
Dojo was Tesla’s big in-house training project. It used custom chips and a unique supercomputer design to process millions of terabytes of video from Tesla vehicles. The goal was to improve Full Self-Driving (FSD) systems. Analysts once valued its potential at as much as $500 billion.(Wikipedia, Inc.com, Reuters)
However, managing both a training supercomputer and separate chips was costly and complex. That is one reason Tesla decided to make a strategic shift.(Reuters, Tom's Hardware)
Who is Leading the Change
Peter Bannon was the leader of the Dojo team. After the project shut down, he left Tesla. About 20 team members have joined a new startup called DensityAI. Others have been reassigned to new projects at Tesla.(Tom's Hardware, Statesman)
Focus on AI5 and AI6 Chips
Tesla’s new plan focuses on its AI5 and AI6 chips:
-
AI5 chip: Expected in development by end of 2026. Meant for both inference and some training.(Inc.com, Reuters)
-
AI6 chip: Created through a major deal with Samsung worth $16.5 billion. These chips will be made in Samsung’s U.S. factory in Texas.(Indiatimes, PC Gamer, Investors.com, Top AI Tools List - OpenTools)
Musk wants Tesla to help make production efficient. The AI6 chips will serve Tesla’s cars, robots, and future smart systems.(Indiatimes)
Investor Reaction
Tesla’s stock rose about 2–2.5% after the news. Investors think the shift away from Dojo is smart and cuts risk from a costly internal project.(Barron's, Statesman)
What This Means for Tesla and Tech
-
Tesla will now use fewer resources and be more focused on making real-time chips.
-
It will rely on trusted partners—Samsung, Nvidia, and AMD—for training compute power.(Tom's Hardware, Applying AI)
-
The change makes Tesla’s AI strategy easier to manage and more cost-effective.
-
It helps Tesla move faster in bringing real self-driving features to roads.
Quick Recap
-
Tesla ended Dojo such supercomputer work.
-
Peter Bannon and some team members left.
-
Tesla is focusing on AI5 and AI6 inference chips.
-
AI6 will be made by Samsung via a $16.5B deal.
-
Tesla stock rose slightly as investors saw smart strategy.
-
The company is leaning on partners for training chips.
Final Thoughts
Tesla's decision to drop Dojo and lean into inference chips shows a smart realignment of technology strategy. It allows Tesla to stay focused, reduce cost, and push forward with practical AI that powers self-driving cars and robotics. This move may allow faster progress—and smarter technology—for users in the near future.