Tesla has officially disbanded its Dojo supercomputer team, dismantling a central piece of its custom AI infrastructure strategy. CEO Elon Musk ordered the shutdown, and the project lead, Peter Bannon, is departing the company, reports Bloomberg.
Approximately 20 Dojo team members have already left to join the spin-off startup DensityAI. The remaining engineers are being reassigned to broader data-center and compute initiatives within Tesla.
Tesla’s pivot away from Dojo reflects a shift from developing bespoke training hardware toward leveraging standardized inference chips. Musk stated that it “does not make sense to divide resources on two quite different AI chip designs,” pledging to consolidate efforts around his AI5 and AI6 chips, which will be produced via external partners such as Samsung and TSMC.
For engineering teams, the implications are significant:
- Strategic realignment: Tesla is scaling back vertical integration on AI training hardware, instead doubling down on scalable, widely available inference chip architectures.
- Talent migration: The loss of Dojo’s core team to startups like DensityAI underscores growing competition in AI hardware talent.
- Resource optimization: By focusing on AI5/AI6 development and outsourcing training infrastructure, Tesla aims to reallocate capital and engineering bandwidth toward its stronger product avenues.
This move signals Tesla’s recalibration of its AI strategy—from ambitious in-house infrastructure to pragmatic, partner-based chip development—reshaping how its engineering teams must adapt.