Tesla to Restart Dojo3 Project After Progress on AI5 Chip

Tesla Inc. will restart development of its Dojo3 project after progress on the AI5 chip, chief executive Elon Musk said, marking a renewed effort to build in-house computing systems for large-scale artificial intelligence workloads.

Dojo is Tesla’s internal supercomputer initiative, created to support AI training for vehicle vision systems, automation software, and robotics programs.

Development slowed in 2025 when the company paused parts of the effort and reassigned engineers, citing limitations in earlier designs and shifting priorities in chip development.

According to Musk, recent work on the AI5 chip has reached a stage that allows Tesla to move forward again with Dojo3.

• Tesla to restart Dojo3 project after progress on AI5 chip
• Elon Musk says Tesla will rebuild Dojo teams to expand in-house AI computing
• Dojo revival underscores Tesla push to control its full AI training stack

The company plans to restart hiring for engineering roles and rebuild teams focused on designing the next version of the system. The objective is to reduce long-term dependence on external suppliers of AI hardware while improving cost efficiency and control over performance.

Dojo was originally positioned as an alternative to GPU-based systems, using custom chips optimized for processing video data and training large neural networks. These networks underpin Tesla’s Full Self-Driving software and related automation features. Progress has been uneven, raising questions among investors about timelines and returns, but the company has maintained that in-house infrastructure remains strategically important.

The restart comes as competition in autonomous driving intensifies globally. Carmakers and technology firms are investing heavily in data centers, chips, and software to support advanced driver-assistance systems and vehicle automation. Tesla is also expanding work in robotics, including humanoid robots, which rely on similar training systems.

Tesla is expected to continue using third-party hardware alongside Dojo3 rather than replacing it entirely. Future chip programs are also underway as part of a longer-term plan to align training systems with in-vehicle computing.

Also Read: OpenAI to Test Ads on ChatGPT Free and Go Tiers in the U.S.

The revival of Dojo3 signals Tesla’s continued focus on owning more of its AI stack as it scales automation and mobility technologies.