Tesla has reportedly shut down its Dojo supercomputer, the ambitious AI training system once hailed by Elon Musk as essential to achieving full self-driving capability. Despite years of development and high expectations, Dojo has failed to deliver on its promise, leading to its abrupt cancellation. The project, designed to process vast amounts of video data from Tesla vehicles to improve its autonomous driving algorithms, was intended to give the company a competitive edge in AI. However, internal challenges, technical limitations, and delays in progress have undermined its viability. With Dojo now decommissioned, Tesla faces uncertainty over how it will advance its self-driving technology without a dedicated, high-performance AI training platform. The shutdown raises questions about the company’s long-term AI strategy and whether it will pivot toward leveraging external AI infrastructure or investing in alternative approaches to achieve its autonomy goals.
Tesla is dismantling its Dojo supercomputer project, marking a significant strategic retreat from its long-held vision of building custom AI hardware in-house to power its autonomous driving and robotics ambitions. According to Bloomberg, the departure of Dojo’s lead, Peter Bannon, and the reassignment of the remaining team to other data center and compute initiatives signal the end of the project. This move follows a broader exodus of key talent, including Ganesh Venkataramanan, former Dojo head, and 20 other engineers who left to found DensityAI, a new AI startup focused on developing chips and infrastructure for robotics, AI agents, and automotive applications. Venkataramanan, Bill Chang, and Ben Floering are leading the effort, which is expected to emerge from stealth soon. The decision comes after years of delays, technical hurdles, and leadership turnover. The Dojo project, first announced in 2021, was meant to be Tesla’s “secret sauce” — a vertically integrated AI system combining custom D1 chips and a massive supercomputer to process vast amounts of video data from Tesla’s vehicle fleet. Elon Musk had long touted Dojo as essential to achieving full self-driving, even predicting in 2023 that it could add $500 billion to Tesla’s market value. However, progress stalled, and key figures like Jim Keller and Venkataramanan eventually left, undermining confidence in the project’s viability. Now, Tesla is pivoting toward reliance on external partners. Musk confirmed on X that future AI chips will be optimized for inference — the process of running AI models — and “at least pretty good” for training. The company is doubling down on partnerships with Nvidia, AMD, and Samsung. Most notably, Tesla recently signed a $16.5 billion deal with Samsung to manufacture its AI6 inference chips, designed to power everything from Full Self-Driving (FSD) to the Optimus humanoid robot and data center applications. Taiwan’s TSMC is also expected to produce Tesla’s AI5 chips. This shift reflects a pragmatic response to the immense capital and technical demands of in-house chip development. As Musk acknowledged during the Q2 earnings call, the company is seeking convergence between its next-gen AI chips and the Dojo roadmap, suggesting a streamlined, more efficient approach. The move may also be a response to investor pressure, especially after Tesla’s board offered Musk a $29 billion compensation package to focus on the company rather than his other ventures, including the AI startup xAI. Despite the setback, Tesla’s stock rose over 2.5% on the news, indicating investor relief that the company is cutting losses on a costly, underperforming project. The Dojo’s demise underscores the challenges of building proprietary AI infrastructure in a field dominated by established players like Nvidia. While Tesla’s early promise of vertical integration has faded, its shift to strategic partnerships could allow it to remain competitive in the AI and robotics race without the massive R&D burden. However, the talent drain remains a deeper concern. Executives like Milan Kovac (Optimus), David Lau (software), and Omead Afshar (Musk’s close aide) have also departed, raising questions about leadership stability. Musk’s growing focus on xAI and his polarizing public persona have further strained the company’s culture and brand, potentially accelerating the departure of top talent. In the end, Tesla’s decision to abandon Dojo is a clear signal: the era of in-house AI hardware may be over for the automaker. Whether this pivot will help Tesla catch up in the AI race or expose its weaknesses remains to be seen.
