Tesla has terminated its in-house Dojo supercomputer program, a project once touted as central to its AI-driven self-driving ambitions. The decision follows the loss of key engineers, escalating development expenses, and a strategic pivot toward external chip partnerships. The company will now rely more on suppliers like Nvidia, AMD, and Samsung, potentially accelerating short-term progress but raising questions about long-term autonomy. The closure marks a shift in Tesla’s AI strategy amid declining sales, intensifying competition, and ongoing legal and regulatory challenges.
Project Termination and Strategic Pivot
Reports confirmed that Tesla disbanded the entire Dojo supercomputer team. The system was intended to train machine learning models for Autopilot and “Full Self-Driving” (FSD) features using vast datasets from the company’s global vehicle fleet. CEO Elon Musk had previously called Dojo a “beast” but acknowledged last year that it was a “long shot” with limited probability of success.
Musk stated it made little sense for Tesla to scale two different AI chip designs, indicating the company would now focus efforts elsewhere.According to company sources, the remaining Dojo engineers will be reassigned to other data center projects. Tesla’s AI chip roadmap will now prioritize the AI5, AI6, and subsequent chips, which Musk said would be excellent for inference and “pretty good” for training. These chips will be integrated alongside external solutions rather than replace them entirely.
It doesn’t make sense for Tesla to divide its resources and scale two quite different AI chip designs.
— Elon Musk (@elonmusk) August 8, 2025
The Tesla AI5, AI6 and subsequent chips will be excellent for inference and at least pretty good for training. All effort is focused on that.
Talent Exodus to Competitors
A major factor that drove the project to the closure as cited by industry reports was the massive loss of human capital. About 20 team members working at the Dojo, including the leader of the project at Tesla, abandoned Tesla and joined DensityAI, an AI-based data center services startup. The departures were characterized by Bloomberg sources to a larger trend of senior executive departures in the technology department of Tesla.
Not only Dojo is experiencing loss of talent. Tesla lost the veteran engineer Pete Bannon who came in from Apple and headed its chip development. It was observed by analysts that the war on AI talent has heightened throughout Silicon Valley with companies trying to poach older engineers with huge compensation packages. The exodus has not only stagnated the development of Tesla proprietary hardware but jeopardized its potential to maintain an in house AI infrastructure.
Increasing expenses and technical issues.
Dojo was aimed to compete with supercomputers created by Nvidia and AMD, using custom silicon to deal with large volume computer vision information. The system however was also expensive to scale up. The production of custom chips and the establishment of data centers cost a lot of funds, which became more difficult to keep up with the drop in sales and growth of mean global competition in the arena of the electric vehicle market.
Technical obstacles were also a factor as sources close to the matter told CNBC. The hardware offered by Tesla had delays in achieving its performance benchmarks which slows its competitive advantage compared to current GPU based systems. The profile of cost associated with the project along with the uncertain pay-off caused executives to reevaluate the viability of the project regarding the priorities of the company at large.
Partnerships with Chipmakers
Tesla’s shift away from Dojo coincides with deeper collaborations with established semiconductor manufacturers. The company has already bought some AI semiconductors from Samsung worth 16.5 billion. The company will go on using Nvidia whose GPUs are being used to train AI at the Tesla. Musk has suggested that external collaboration would enable Tesla to gain vector with regard to chip design, towards a single scalable architecture as opposed to “division of resources”.
Wells Fargo analysts told Yahoo News Australia that such a step is good news to chip suppliers because this will lead to Tesla being more dependent on off-the-shelf products and thereby increasing the demand. Industry watchers speculate that this joint venture model could result in a more rapid pace of AI training upgrades, but lessen the control Tesla has on proprietary systems, an important determinant of its long-range competitive needs in the autonomous driving space.
Broader Implications for Tesla’s AI and FSD Programs
This shutdown of Dojo is sensitive to Tesla drivers’ work on autonomous driving. Its pilot robotaxi in Austin, Texas has had technical problems and safety accidents, including near misses as reported by the local press. The squeeze is also legal, as a recent Florida jury stated that Tesla was partly responsible in an Autopilot crash, which resulted in an almost $250 million penalty.There is an increase in regulatory inquiry concerning the safety and marketing of the Tesla FSD system. The competitors such as Waymo and Cruise are entering more deeply into the robotaxi industry, which makes the competition more intense regarding autonomous mobility leadership.
In the absence of its own supercomputer such as Dojo, Tesla may be required to adjust its AI plan in order to keep pace with the competition that already possesses specialized AI resources.However, Tesla market capitalization still remains at more than a trillion dollars sustained by a 10% rise of share price in the last one month. Investors are more concentrated on the short-term expenditure and the possible benefits of efficiency in the use of external AI hardware. Nonetheless, industry analysts warn that this might put the future of Tesla in abandoning its compute strength at the expense of possibly losing its own development capacity.