Bloomberg is reporting that the Tesla Dojo chip project is being disbanded. The lead Peter Bannon and 20 others are going into another company Density AI.
The team recently lost about 20 workers to newly formed DensityAI, and the remaining Dojo workers are being reassigned to other data center and compute projects within Tesla according to Bloomberg.
Density AI is a newly formed stealth startup founded by Ganesh Venkataramanan, a former Dojo project lead who left Tesla in late 2023. The company is developing full-stack AI solutions, including chips, hardware, and software tailored for high-density computing in areas like automotive, robotics, and AI agents.
Tesla is merging Dojo 3 into its next-generation AI6 chip architecture. This could end Dojo as a standalone chip and project.
Brian Wang is a Futurist Thought Leader and a popular Science blogger with 1 million readers per month. His blog Nextbigfuture.com is ranked #1 Science News Blog. It covers many disruptive technology and trends including Space, Robotics, Artificial Intelligence, Medicine, Anti-aging Biotechnology, and Nanotechnology.
Known for identifying cutting edge technologies, he is currently a Co-Founder of a startup and fundraiser for high potential early-stage companies. He is the Head of Research for Allocations for deep technology investments and an Angel Investor at Space Angels.
A frequent speaker at corporations, he has been a TEDx speaker, a Singularity University speaker and guest at numerous interviews for radio and podcasts. He is open to public speaking and advising engagements.
The reason for dojo being disbanded is mainly the software issues, specifically it doesn’t have the cuda counterpart to quickly run the ai model like Nvidia gpu, as I believe ai6 is not for training it only be used as on vehicle chip, running trimmed model from GPU training.
It may be possible to compose super computer but once again it face cuda problem, some startup company may be possible to do this cuda work and Tesla may acquire them later.
Dividing resources in this hyper competitive race to ASI environment between two entirely different chip architectures when one of them was both more competitive and critical to all the core businesses was not viable.
Tesla needed to put everything into its AI line used for all its vehicle robots, Optimus robots, plans for distributed inference compute. AI5 and beyond can be used for training in big compute clusters both for X.ai and Tesla – but Dojo had no use in distributed compute/mobile robots.
A single chip line lets Tesla commit to long term demand for tens of millions of the most advanced chips, lowering per unit costs and scaling what they can achieve.
I admire Elon for his ability to try something and if it does not work to end it.
Dojo was a brilliant gamble… that failed. It would have been nice if it succeeded, but Elon saw that the external chips were a better, cheaper way forward.
Not to say that sometimes Elon does not have has his ‘love’ projects that he sticks to even though they failed (design of Cybertruck).
This would explain Elons comments during the last quarterly report. He stated that they should try to unify AI6 and Dojo 3, which was surprising considering that the AI-chips so far used a completely different technology than Dojo, so far. So either the AI-chips or the Dojo chips would have to make a complete turn around. I could not make it fit.
So the Dojo technology was scrapped and the team was (unsurprisingly) disgruntled and left for a new start-up. I agree with you, “d”, that it is an admirable trait to cut off “darling” projects that fail.
Dojo has been lagging Nvidia chips in terms of flops per W. Given Nvidias superior cadence – a new chip every year verus Teslas one chip every two years – it would seem unlikely that Tesla would catch up with the current Dojo architecture. So Elon made the right but difficult call to ax the Dojo architecture. Cudos to him!
Now the question is if AI6 will be competitive with future Nvidia generations of training chips, or if it will be for inference only. Time will tell, but it’s not boring!