For years, Elon Musk has spoken of the promise of Dojo, the AI supercomputer that was imagined to be the cornerstone of Tesla’s AI ambitions. It was vital sufficient to Musk that in July 2024, he stated the corporate’s AI staff would “double down” on Dojo within the lead-up to Tesla’s robotaxi reveal, which occurred in October.
After six years of hype, Tesla determined final month to shut down Dojo and disband the staff behind the supercomputer in August 2025. Inside weeks of projecting that Dojo 2, Tesla’s second supercluster that was meant to be constructed on the corporate’s in-house D2 chips, would attain scale by 2026, Musk reversed course, declaring it “an evolutionary dead end.”
This text initially got down to clarify what Dojo was and the way it might assist Tesla obtain full-self driving, autonomous humanoid robots, semiconductor autonomy, and extra. Now, you’ll be able to consider it extra as an obituary of a challenge that satisfied so many analysts and buyers that Tesla wasn’t simply an automaker, it was an AI firm.
Dojo was Tesla’s custom-built supercomputer that was designed to coach its “Full Self-Driving” neural networks.
Beefing up Dojo went hand-in-hand with Tesla’s purpose to achieve full self-driving and convey a robotaxi to market. FSD (Supervised) is Tesla’s superior driver help system that’s on lots of of hundreds of Tesla automobiles at the moment and may carry out some automated driving duties, but it surely nonetheless requires a human to be attentive behind the wheel. It’s additionally the idea of comparable expertise powering Tesla’s restricted robotaxi service that the corporate launched in Austin this June utilizing Mannequin Y SUVs.
Whilst Dojo’s raison d’être began to come back to life, Tesla did not attribute its self-driving successes — controversial as they had been — to the supercomputer. In reality, Musk and Tesla had barely talked about Dojo at everywhere in the previous 12 months. In August 2024, Tesla started selling Cortex, the corporate’s “giant new AI training supercluster being built at Tesla HQ in Austin to solve real-world AI,” which Musk has stated would have “massive storage for video training of FSD and Optimus.”
In Tesla’s This autumn 2024 shareholder deck, the corporate shared updates on Cortex, however nothing on Dojo. It’s not clear whether or not Tesla’s Dojo shutdown impacts Cortex.
Techcrunch occasion
San Francisco
|
October 27-29, 2025
The response to Dojo’s disbanding has been combined. Some see it as one other instance of Musk making guarantees he can’t ship on that comes at a time of falling EV gross sales and a lackluster robotaxi rollout. Others say the shutdown wasn’t a failure, however a strategic pivot from a high-risk, self-reliant {hardware} to a streamlined path that depends on companions for chip improvement.
Dojo’s story reveals what was on the road, the place the challenge fell quick, and what its shutdown alerts for Tesla’s future.
A recap of Dojo’s shutdown
Tesla disbanded its Dojo staff and shut down the challenge in mid-August 2025. Dojo’s lead, Peter Bannon, left the corporate as properly, following the departure of round 20 staff who left to begin their very own AI chip and infrastructure firm referred to as DensityAI.
Analysts have identified that shedding key expertise can shortly derail a challenge, particularly a extremely specialised, inside tech challenge.
The shutdown got here a few weeks after Tesla signed a $16.5 billion deal to get its next-generation AI6 chips from Samsung. The AI6 chip is Tesla’s wager on a chip design that may scale from powering FSD and Tesla’s Optimus humanoid robots to high-performance AI coaching in knowledge facilities.
“Once it became clear that all paths converged to AI6, I had to shut down Dojo and make some tough personnel choices, as Dojo 2 was now an evolutionary dead end,” Musk posted on X, the social media platform he owns. “Dojo 3 arguably lives on in the form of a large number of AI6 [systems-on-a-chip] on a single board.”
Tesla’s Dojo backstory
Musk has insisted that Tesla isn’t simply an automaker, or perhaps a purveyor of photo voltaic panels and power storage programs. As a substitute, he has pitched Tesla as an AI firm, one which has cracked the code to self-driving automobiles by mimicking human notion.
Most different firms constructing autonomous car expertise depend on a mixture of sensors to understand the world — like lidar, radar and cameras — in addition to high-definition maps to localize the car. Tesla believes it could obtain totally autonomous driving by counting on cameras alone to seize visible knowledge after which use superior neural networks to course of that knowledge and make fast choices about how the automotive ought to behave.
The pitch has been that Dojo-trained AI software program will finally be pushed out to Tesla prospects through over-the-air updates. The dimensions of FSD additionally means Tesla has been in a position to rake in hundreds of thousands of miles price of video footage that it makes use of to coach FSD. The concept there’s that the extra knowledge Tesla can gather, the nearer the automaker can get to really attaining full self-driving.
Nevertheless, some trade specialists say there may be a restrict to the brute drive method of throwing extra knowledge at a mannequin and anticipating it to get smarter.
“First of all, there’s an economic constraint, and soon it will just get too expensive to do that,” Anand Raghunathan, Purdue College’s Silicon Valley professor {of electrical} and laptop engineering, instructed TechCrunch. Additional, he stated, “Some people claim that we might actually run out of meaningful data to train the models on. More data doesn’t necessarily mean more information, so it depends on whether that data has information that is useful to create a better model, and if the training process is able to actually distill that information into a better model.”
Raghunathan stated regardless of these doubts, the pattern of extra knowledge seems to be right here for the short-term no less than. And extra knowledge means extra compute energy wanted to retailer and course of all of it to coach Tesla’s AI fashions. That was the place Dojo, the supercomputer, got here in.
What’s a supercomputer?
Dojo was Tesla’s supercomputer system that was designed to perform as a coaching floor for AI, particularly FSD. The identify is a nod to the area the place martial arts are practiced.
A supercomputer is made up of hundreds of smaller computer systems referred to as nodes. Every of these nodes has its personal CPU (central processing unit) and GPU (graphics processing unit). The previous handles general administration of the node, and the latter does the complicated stuff, like splitting duties into a number of components and dealing on them concurrently.
GPUs are important for machine studying operations like those who energy FSD coaching in simulation. In addition they energy massive language fashions, which is why the rise of generative AI has made Nvidia probably the most helpful firm on the planet.
Even Tesla buys Nvidia GPUs to coach its AI (extra on that later).
Why did Tesla want a supercomputer?
Tesla’s vision-only method was the principle motive Tesla wanted a supercomputer. The neural networks behind FSD are educated on huge quantities of driving knowledge to acknowledge and classify objects across the car after which make driving choices. That signifies that when FSD is engaged, the neural nets have to gather and course of visible knowledge constantly at speeds that match the depth and velocity recognition capabilities of a human.
In different phrases, Tesla means to create a digital duplicate of the human visible cortex and mind perform.
To get there, Tesla must retailer and course of all of the video knowledge collected from its automobiles around the globe and run hundreds of thousands of simulations to coach its mannequin on the information.
Tesla relied primarily on Nvidia to energy its present Dojo coaching laptop, but it surely didn’t need to have all its eggs in a single basket — not least as a result of Nvidia chips are costly. Tesla had hoped to make one thing higher that elevated bandwidth and decreased latencies. That’s why the automaker’s AI division determined to provide you with its personal {custom} {hardware} program that aimed to coach AI fashions extra effectively than conventional programs.
At that program’s core was Tesla’s proprietary D1 chips, which the corporate stated had been optimized for AI workloads.
Inform me extra about these chips
Tesla, like Apple, thinks {hardware} and software program ought to be designed to work collectively. That’s why Tesla was working to maneuver away from the usual GPU {hardware} and design its personal chips to energy Dojo.
Tesla unveiled its D1 chip, a silicon sq. the dimensions of a palm, on AI Day in 2021. The D1 chip entered into manufacturing round July 2023.
The Taiwan Semiconductor Manufacturing Firm (TSMC) manufactured the chips utilizing 7 nanometer semiconductor nodes. The D1 has 50 billion transistors and a big die dimension of 645 millimeters squared, in accordance with Tesla. That is all to say that the D1 guarantees to be extraordinarily highly effective and environment friendly and to deal with complicated duties shortly.
The D1 wasn’t as highly effective as Nvidia’s A100 chip, although.
Tesla had been engaged on a next-gen D2 chip that aimed to resolve data move bottlenecks. As a substitute of connecting the person chips, the D2 would have put the complete Dojo tile onto a single wafer of silicon.
Tesla by no means confirmed what number of D1 chips it ordered or acquired. The corporate additionally by no means supplied a timeline for a way lengthy it might have taken to get Dojo supercomputers operating on D1 chips.
What did Dojo imply for Tesla?
Tesla’s hope was that by taking management of its personal chip manufacturing, it would in the future have the ability to shortly add massive quantities of compute energy to AI coaching packages at a low price.
It additionally meant not having to depend on Nvidia’s chips sooner or later, that are more and more costly and onerous to safe. Now, Tesla goes all-in on partnerships — with Nvidia, AMD, and Samsung, which is able to construct its next-gen AI6 chip.
Throughout Tesla’s second-quarter 2024 earnings name, Musk stated demand for Nvidia {hardware} was “so high that it’s often difficult to get the GPUs.” He stated he was “quite concerned about actually being able to get steady GPUs when we want them, and I think this therefore requires that we put a lot more effort on Dojo in order to ensure that we’ve got the training capability that we need.”
Dojo was a dangerous wager, one which Musk hedged a number of instances by saying that Tesla may not succeed.
In the long term, Tesla toyed with the concept of making a brand new enterprise mannequin primarily based on its AI division, with Musk even saying throughout a Q2 2024 earnings name that he noticed “a path to being competitive with Nvidia with Dojo.” Whereas D1 was extra tailor-made for Tesla laptop imaginative and prescient labeling and coaching — helpful for FSD and Optimus coaching — it wouldn’t have been helpful for a lot else. Future variations must be extra tailor-made to general-purpose AI coaching, Musk stated.
The issue that Tesla might need come up in opposition to is that the majority AI software program out there was written to work with GPUs. Utilizing Dojo chips to coach general-purpose AI fashions would have required rewriting the software program.
That’s, except Tesla rented out its compute, much like how AWS and Azure hire out cloud computing capabilities — an concept that excited analysts. A September 2023 report from Morgan Stanley predicted that Dojo might add $500 billion to Tesla’s market worth by unlocking new income streams within the type of robotaxis and software program providers.
Briefly, Dojo chips had been an insurance coverage coverage for the automaker, however one which may have paid dividends.
How far did Tesla Dojo get?
Musk usually supplied progress reviews, however lots of his objectives for Dojo had been by no means reached.
As an example, Musk instructed in June 2023 that Dojo had been on-line and operating helpful duties for just a few months.” Across the identical time, Tesla stated it anticipated Dojo to be one of many prime 5 strongest supercomputers by February 2024 and had deliberate for whole compute to achieve 100 exaflops in October 2024, which might have required roughly 276,000 D1s, or round 320,500 Nvidia A100 GPUs.
Tesla by no means supplied an replace or any data that may counsel it ever reached these objectives.
Tesla and Musk made quite a few different pledges for Dojo, together with monetary ones. As an example, Tesla dedicated in January 2024 to spend $500 million to construct a Dojo supercomputer at its gigafactory in Buffalo, New York, and has already spent $314 million of that, per a 2024 report.
Simply after Tesla’s second-quarter 2024 earnings name, Musk posted images of Dojo 1 on X, saying that it might have “roughly 8k H100-equivalent of training online by end of year. Not massive, but not trivial either.”
Regardless of all of this exercise — notably by Musk on X and in earnings calls — point out of Dojo abruptly ended August 2024. And speak switched to Cortex.
In the course of the firm’s fourth-quarter 2024 earnings name, Tesla stated it accomplished the deployment of Cortex, “a ~50k H100 training cluster at Gigafactory Texas” and that Cortex helped allow V13 of supervised FSD.
In Q2 2025, Tesla famous it “expanded AI training compute with an additional 16k H200 GPUs at Gigafactory Texas, bringing Cortex to a total of 67k H100 equivalents.” Throughout that very same earnings name, Musk stated he anticipated to have a second Dojo cluster working “at scale” in 2026. He additionally hinted at potential redundancies.
“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,” Musk stated.
A couple of weeks later, he reversed course and disbanded the Dojo staff.
TechCrunch confirmed in late August 2025 that Tesla nonetheless plans to commit $500 million to a supercomputer in Buffalo — it simply gained’t be Dojo.
This story initially revealed August 3, 2024. The article was up to date for a remaining time September 2, 2025, with new details about Tesla’s choice to close down Dojo.
