So, remember when Elon Musk kept talking about Dojo — the in-house supercomputer that was supposed to be Tesla’s secret weapon in the race to full self-driving cars? Well, that chapter just closed.
What Happened?
Tesla has officially broken up the team behind its ambitious Dojo project, according to a Bloomberg report. That includes Peter Bannon, who led the initiative and is now leaving the company. The rest of the team is being reshuffled to work on other compute and data center tasks inside Tesla.
This move came quietly, but it marks a major shift for a company that once predicted Dojo would open up an entirely new wave of AI-driven revenue.
Why It Matters
Dojo wasn’t just another chip project. It was Tesla’s bet on making its own AI training chips and systems — hardware designed specifically to crunch the video data needed to power self-driving cars. Musk had spoken about Dojo since 2019, calling it essential to Tesla’s AI journey. Back in 2023, Morgan Stanley even estimated Dojo could unlock $500 billion in additional market value for Tesla.
Now? That dream has been shelved.
Photo by Karsten Winegeart on Unsplash
A Shift in Strategy
The writing might’ve been on the wall. Around mid-2024, talk of Dojo started to fade. Musk shifted attention toward something new: a different AI training system called Cortex, currently being built out at Tesla’s HQ in Austin. It’s being billed as a “giant” supercluster aimed at solving real-world AI problems. That shift likely set the stage for Dojo’s quiet exit.
Interestingly, during Tesla’s Q2 earnings call, Musk hinted that a single unified chip — possibly something like the AI6 inference chip — could make Dojo’s hardware redundant.
A Brain Drain and a Startup Boom
Another clue this was coming? The exodus of around 20 former Dojo engineers who’ve now formed their own AI startup: DensityAI.
Led by ex-Dojo chief Ganesh Venkataramanan, along with ex-Tesla employees Bill Chang and Ben Floering, DensityAI is coming out of stealth soon and is working on its own chips, software, and data center hardware designed for AI, robotics, and automotive applications.
If you’re wondering whether top AI talent is still getting poached, the answer is yes—and sometimes they’re doing the poaching themselves.
Photo by Start Digital on Unsplash
The New Plan
Instead of making every AI chip in-house, Tesla’s now leaning more on external players.
- Just last month, Tesla signed a massive $16.5 billion deal with Samsung to manufacture its new AI6 chips.
- They’re also expected to keep working with Nvidia and AMD as hardware partners for training and inference workloads.
This represents a pivot from a vertically integrated approach (build everything yourself) to a more hybrid model: focus on what you’re best at, and outsource the rest.
Where Does This Leave Tesla’s AI Ambitions?
It’s still very much in play. Musk recently went to bat for a $29 billion pay package meant to keep him focused on Tesla and its AI efforts. Meanwhile, Tesla continues to push its robotaxi goals—despite a limited rollout in Austin that received mixed reviews for its driving behavior.
And while Dojo is done, Tesla’s AI work is far from over. The company seems laser-focused now on its Cortex platform and scaling the AI6 chip across everything from autonomous vehicles to its Optimus robots.
What’s Next?
We’ll likely hear more about Cortex over the next year, especially given how central AI still is to Musk’s long-term vision for Tesla. With new chips, new partnerships, and a shakeup of the original AI team, the next phase of Tesla’s AI push is well underway. Just not in the shape anyone expected.
If you were wondering whether Tesla’s going all-in on AI: the answer is still yes. Just not the Dojo way.
Keywords: Tesla Dojo shutdown, Elon Musk AI chip, Cortex AI supercluster, DensityAI, Tesla AI6 chip, Tesla Samsung chip deal, full self-driving Tesla, Tesla AI strategy 2025, Peter Bannon Tesla, Ganesh Venkataramanan startup
