Tesla has ended its in-house Dojo SuperComputer program and is shifting AI chip development only to inference chips that run real-time AI models in vehicles and robotics. CEO Elon Musk saw the X's movement and said splitting resources into two chip designs “makes no sense.”
Dojo is designed to train Tesla's autonomous driving software using video and sensor data collected from the EV fleet. Once considered a key differentiator of Tesla's fully autonomous ambitions, the company will focus its AI efforts on AI5 and AI6 reasoning chips. The mask says it handles both real-time operations and some training functions.
Sign up for CBT News' Daily Newsletter and deliver the latest industry stories directly to your inbox.
Strategic Pivot is facing softer demand for EVs, especially in Europe, as thousands of people are slowly fired and a wave of execution departures. The company is now making itself a focus on artificial intelligence and robotics, and has signed a $16.5 billion deal with Samsung to help produce the next generation of chips.
This is why it's important
For those selling or competing with Tesla, this shows the company's growing commitment to deploying autonomous driving capabilities more quickly. Inference chips allow real-time autonomy. This means Tesla accelerates updates to its FSD systems, creating stronger selling points and differentiation in the EV and Premium segments. It also reflects the integration of costs and operations at a time when Tesla's price reductions and margin pressures are felt throughout the automotive industry.
Important takeouts:
- The Dojo program will end after an internal review
Musk has left the Dojo Supercomputer team with key personnel like Peter Bannon. - Shift to development of inference chips for autonomy
Tesla focuses solely on chips such as AI5 and AI6, powers real-time autonomous driving and robotics, and integrates the AI roadmap. - Operational integration amidst market pressure
The move is part of a broader Tesla restructuring, including layoffs, executive turnover and strengthening AI investments to offset delayed EV demand. - Samsung tapped on a $16.5 billion AI chip deal
Tesla will partner with Samsung to produce next-generation AI chips with the aim of becoming an AI5 by 2026, but the production timeline for AI6 remains unknown. - Fast FSD deployment could potentially rebuild the EV competition
A more focused AI chip strategy will allow Tesla to deploy advanced autonomous driving capabilities faster, increasing the interests of other automakers.
