Key Points
- Tesla dissolves the Dojo AI‑training supercomputer team.
- Peter Bannon, head of Dojo, departs; about 20 ex‑team members formed DensityAI.
- Elon Musk says Tesla will focus on AI5 and AI6 chips for both inference and training.
- AI5 and AI6 chips will be manufactured by Samsung after a $16.5 billion deal.
- Tesla will lean on external vendors like NVIDIA for training silicon.
- AI5 production slated for 2026; AI6 to follow.
- Grok chatbot from xAI is now integrated into Tesla vehicles.
- Robotaxi pilot continues with mixed results.
Dojo Team Disbanded
Tesla announced that the team behind its proprietary Dojo AI‑training supercomputer is being broken up. Remaining engineers will be reassigned to other internal projects, while Peter Bannon, the head of Dojo, is exiting the company. The restructuring follows the departure of roughly 20 Dojo employees who have already created a new venture called DensityAI.
Shift to AI5 and AI6 Chip Strategy
CEO Elon Musk responded on the platform X, noting that it does not make sense for Tesla to split resources between two distinct AI‑chip designs. He emphasized that the upcoming AI5 and AI6 chips, which will be manufactured by Samsung under a $16.5 billion agreement, are “excellent for inference and at least pretty good for training.” Musk said all effort will be focused on these chips, which are intended to power real‑time decision‑making in Tesla vehicles and robots.
Implications for Training Infrastructure
The decision to shut down Dojo effectively ends Tesla’s long‑standing plan to build an in‑house training architecture. Going forward, the company will rely more heavily on external vendors such as NVIDIA for training‑specific silicon, continuing to invest billions in those solutions. Production of the AI5 chip is targeted to begin in 2026, with the AI6 chip expected to follow.
Broader AI Initiatives
Even as Dojo is wound down, Tesla remains active in other AI endeavors. The firm’s xAI division has launched the Grok chatbot, which is now available inside Tesla vehicles. Additionally, Tesla continues to pilot its Robotaxi fleet, though results have been mixed.
Overall Impact
By consolidating its AI hardware efforts on the AI5 and AI6 platforms, Tesla aims to streamline development, reduce duplication, and accelerate deployment of AI capabilities across its product lineup. The move signals a strategic pivot from building a separate training supercomputer toward leveraging a unified chip roadmap and external training partners.
Source: engadget.com