Earlier this week, Tesla ended Venture Dojo, which was a partnership between Tesla and Dojo computer systems. This introduced an finish to the automaker’s efforts to construct in-house chips for its driverless know-how objectives.
Tesla has been targeted on constructing driverless automobiles for nearly a decade now. So why did Tesla finish its makes an attempt to provide chips in-house for its autonomous autos?
The choice got here as a result of Tesla CEO Elon Musk shifted to counting on Nvidia and AMD for compute chips, together with Samsung for manufacturing. Based on Musk, Tesla was closing Venture Dojo as a result of it will be inefficient for the enterprise to separate its assets and scale two distinct AI chip designs. Dojo wouldn’t be included in Tesla’s shopper items as a result of it was supposed to coach the corporate’s autonomous driving programme.
So, what was Tesla Dojo?
Dojo was the automaker’s effort to construct customized supercomputers designed to coach Tesla’s neural community for driverless automobiles. A neural community features like a simulated mind, supporting Tesla’s aim of launching Robotaxi and Full Self-Driving (FSD) automobiles. This community is put in throughout a big fleet of Tesla automobiles at the moment, enabling some automated driving features that also require a human to be attentive behind the wheel.
Tesla’s determination to close down Dojo, which Musk had been speaking about since 2019, marks a serious strategic shift. Musk had beforehand described Dojo because the cornerstone of Tesla’s AI ambitions and its push for full self-driving, due to its potential to “course of really huge quantities of video knowledge.” He even talked about it briefly through the firm’s second-quarter earnings name.
Why does Tesla want a supercomputer?
The first motive is Tesla’s vision-only philosophy. For FSD to work, its neural networks should determine and categorise objects within the setting, then make driving choices in actual time. This requires coaching on huge volumes of driving knowledge in order that, when activated, FSD can constantly collect and course of visible knowledge at speeds corresponding to human depth and movement detection. In essence, Tesla desires to construct a digital mannequin of how the human mind and visible cortex perform.
To attain this, Tesla should retailer, course of, and run hundreds of thousands of simulations on video knowledge gathered from autos worldwide. Though Tesla relied on Nvidia to run its Dojo coaching laptop, the goal was to cut back reliance on a single provider and enhance efficiency by boosting bandwidth and decreasing latency.
Story continues beneath this advert
What’s a supercomputer?
Tesla’s Dojo system was supposed to function an AI coaching platform for FSD. A supercomputer consists of hundreds of smaller computer systems, or nodes. Every node has a central processing unit (CPU) and a graphics processing unit (GPU). Whereas the CPU manages the node total, the GPU handles extra complicated duties, comparable to splitting workloads into many elements and processing them in parallel. GPUs are important for machine studying duties like FSD coaching in simulation.
Musk’s broader imaginative and prescient is for Tesla to develop into an AI firm able to making self-driving automobiles by mimicking human imaginative and prescient. Most different firms within the autonomous driving sector use a mix of sensors—lidar, radar, cameras—and high-definition maps. Tesla believes it may possibly obtain full autonomy with cameras alone, paired with superior neural networks to analyse knowledge and make prompt driving choices.
What’s taking place now?
In August 2024, Musk started selling Cortex, Tesla’s new AI coaching supercluster being constructed at its Austin headquarters to deal with real-world AI challenges. This shifted focus away from Dojo.
Bloomberg reviews that Tesla will now rely extra on Nvidia and different companions comparable to AMD for computing and Samsung for chip fabrication. On 28 July 2025, Tesla and Samsung signed a $16.5 billion settlement to provide AI6 inference processors for high-performance AI coaching, FSD, and Tesla’s Optimus humanoid robots.
Story continues beneath this advert
In the meantime, Dojo’s lead, Peter Bannon, is leaving the corporate, and the remaining staff members will probably be reassigned to different knowledge centre and compute tasks inside Tesla.