Now, Tesla has formally stopped its go big or go Dojo AI training supercomputer project. This closure goes against the company’s purported goal of becoming the first to reach full self-driving. If the decision is made in August 2024, it could be just as Elon Musk is making his own pivot to a new initiative, Cortex. This unexpected shift raises questions about the future of Tesla’s AI ambitions and its impact on the company’s market valuation.
Dojo was purpose–built as a new, enormous AI training supercluster at Tesla’s headquarters in Austin, Texas. The initiative’s focus is to meet pressing, real-world challenges with AI. Just this year, Morgan Stanley predicted that Dojo’s successful implementation might add an extraordinary $500 billion to Tesla’s market value. This growth, aside from some volume expansion, would largely be driven by unlocking new revenue opportunities in robotaxis and software as a service. Musk had previously highlighted Dojo’s potential, stating it would allow Tesla to “process truly vast amounts of video data,” which is essential for advancing autonomous driving technology.
As Tesla prepared for its robotaxi reveal in October, Musk indicated that the AI team would “double down” on Dojo. Yet for all these claims and promises, the project’s future was seemingly thrown into doubt after this month’s announcement. Interestingly, on Tesla’s second-quarter earnings call, Musk seemed to short sell Dojo, hinting at a lost interest in the project.
The recent departure of Dojo’s lead, Peter Bannon, doesn’t make things any clearer. After his departure, Tesla plans to repurpose the rest of his team to other data center and high-performance computing initiatives across the company. In particular, it raises questions about the automaker’s recently announced plans to increase its reliance on external tech partners. For computing power, they’ll look to Nvidia, processing support to AMD, and chip fabrication to Samsung.
Musk has always emphasized Tesla’s role as an AI and robotics company. He encourages the shareholders to buy into this bigger picture. Given the recent small robotaxi rollout in Austin that came at the end of June, red flags were raised. Output of the previously written paragraph. Model Y vehicles with a dummy in the front passenger seat exhibited driver inattentiveness. This makes the case that Tesla is far from ready to deploy autonomous vehicles at scale.
Tesla is taking on a lot of hard things. They are working on a next-generation D2 chip to remove any and all information flow bottlenecks that dragged down its predecessor. The internalization of X.AI would fit Musk’s vision of making the company’s AI efforts more integrated, industrialized, and ultimately, controlled.
Yet the turf around Dojo has changed dramatically. DensityAI, a new startup created by former Dojo head Ganesh Venkataramanan and ex-Tesla employees Bill Chang and Ben Floering, has risen to fill that void. This move represents a possible pivot in direction for key personnel that used to be at the core of Tesla’s AI talent.
According to industry observers, the sudden termination of Dojo is raising troubling questions not only about Tesla’s execution and priorities but the company’s strategic direction.
“Thinking about Dojo 3 and the AI6 inference chip, it seems like intuitively, we want to try to find convergence there, where it’s basically the same chip.” – Elon Musk