Elon Musk Pulls the Plug on Tesla's Dojo Supercomputer Project
In a surprising strategic move, Elon Musk has announced the end of Tesla’s much-publicized “Dojo” supercomputer initiative — a project originally designed to train the company’s advanced artificial intelligence systems at an unprecedented scale. Calling the endeavor “a technological dead end,” Musk confirmed that Tesla will redirect its engineering power toward developing the next-generation AI5 and AI6 chips, which he believes hold far greater potential for the company’s future in autonomous driving and robotics.
From Big Ambitions to a Change in Course
Only weeks ago, Musk had spoken confidently about the second generation of Dojo being ready for mass deployment by 2026. Tesla’s first Dojo system blended NVIDIA GPUs with in-house D1 chips, and plans were in motion for Dojo 2, powered by the more advanced D2 processors. However, the vision has now shifted. According to Musk, emerging chip designs, particularly AI6, have outpaced the value proposition of scaling Dojo further, making the costly supercomputer build less attractive compared to integrating cutting-edge AI processors into Tesla’s fleet.
The Rise of AI5 and AI6
The upcoming AI5 chips, produced by semiconductor giants TSMC and Samsung, are intended to power Tesla’s Full Self-Driving (FSD) systems with greater efficiency and processing speed. Meanwhile, AI6 chips are being designed for on-vehicle inference and high-performance training — not just for cars, but also for Tesla’s humanoid robot projects. Musk envisions a unified hardware architecture where multiple AI5 and AI6 chips can be combined onto a single board, offering superior performance-per-watt and lowering costs compared to a massive centralized supercomputer.
What “Dojo 3” Could Mean
While the Dojo brand is officially discontinued, Musk hinted that this integrated chip platform could represent a new phase — something he jokingly referred to as “Dojo 3.” Unlike its predecessors, this system would not be a single colossal computer, but rather a distributed network of powerful, vehicle-mounted AI processors working in tandem to advance Tesla’s autonomous capabilities.
Industry Context and Investor Concerns
Tesla’s shift comes amid slowing electric vehicle sales and growing scrutiny over the safety of its self-driving technology. The company has faced delays and safety incidents in its autonomous taxi program in Austin, Texas, raising questions about the readiness of its FSD platform. At the same time, Tesla must reassure investors that AI — not just EVs — will be the company’s growth engine for the next decade.
The Road Ahead
By focusing on AI5 and AI6, Tesla is betting on a decentralized, chip-centric approach to AI scaling rather than investing in massive supercomputers. This pivot reflects a broader industry trend, where companies prioritize modular, power-efficient AI hardware that can be deployed widely rather than concentrated in one data center. If successful, this strategy could give Tesla an edge in the global race to perfect autonomous driving — and potentially revolutionize AI-powered robotics in the process.