In a surprise turn of events, Tesla Dojo supercomputer shutdown reports have revealed that the electric car company is splitting up its high-profile in-house AI training team. The action is a dramatic departure from Tesla's plan for creating artificial intelligence infrastructure for autonomous car technology and robotics.

Dojo's Ambitious Beginnings

Tesla's Dojo venture was first revealed in 2019 and was officially revealed during the firm's AI Day event in 2021. Dojo was to be the next generation of supercomputers based on the D1 chip, which was made by Tesla, to process enormous amounts of video and sensor data collected from Tesla's global fleet.

The goal was to accelerate the development of Tesla's Full Self-Driving (FSD) system and make AI-based robotics projects such as the Optimus humanoid robot possible. CEO Elon Musk even hinted that Dojo might be profitable as a cloud service, renting out its processing power to other companies.

The Shutdown Option

Now, after years of gestation, the Tesla Dojo supercomputer shutdown is finally in progress. Sources report the Dojo team has been disbanded, and project lead Peter Bannon has exited the company. Approximately 20 key engineers have allegedly left to work at a new AI startup, DensityAI, founded by ex-Dojo architect Ganesh Venkataramanan.

The rest of the Dojo staff are being repurposed to work on other AI initiatives at Tesla, primarily inference computing and not training at scale.

Elon Musk justifies the move

Elon Musk broke down the switch on X (formerly Twitter), writing that having two completely different AI chip designs was not effective. Tesla will instead consolidate its hardware efforts around its future AI5 and AI6 chips.

These chips, in Musk's opinion, will perform inference and some form of training, basically making Dojo's unique architecture unnecessary. He referred to the previous Dojo plan as "an evolutionary dead end," but added that the concepts in "Dojo 3" would live on in the AI6 system-on-chip boards.

Changing to Partnerships

The shutdown of Tesla Dojo supercomputer also marks a shift towards outsourcing hardware partners. Tesla inked a $16.5 billion agreement with Samsung to produce AI6 chips and will also use Nvidia and AMD hardware to conduct AI computing.

Industry observers think that this will reduce expenses, speed up deployment, and allow Tesla to compete more effectively in the expanding market for autonomous driving. By shedding some of its AI hardware demands, Tesla can focus more on software, processing information, and expanding its robotaxi plans.

What It Means for Tesla's AI Future

Although Dojo was presented as a one-trick competitive advantage, its closure doesn't mean Tesla is abandoning AI. Tesla's autonomous initiative is still a major focus area, and the robotaxi network has begun to deploy in a limited capacity in cities like Austin.

Yet the move shows Tesla prioritizes cost-effectiveness and adaptability over manufacturing each technology component in-house. Aligning with the behemoth chip manufacturers and GPU sellers will keep Tesla in the lead without the disadvantages of designing, testing, and manufacturing fully proprietary systems.

Market Response

Following the Tesla Dojo supercomputer shutdown announcement, Tesla stock went up, indicating investors are okay with this shift in strategy. Most see this as a reasonable move, reducing R&D expenses while Tesla remains competitive in both electric vehicle and AI sectors.

The Bigger Picture

Tesla's move is illustrative of a larger trend in the technology industry: fewer and fewer companies are designing whole AI supercomputers from the ground up. Rather, companies are increasingly looking to specialized chips and cloud infrastructure from established semiconductor companies.

For Tesla, its Dojo project didn't fully realize its ultimate full vision, but what it discovered will shape the company's next generation of AI computing. Whether the shift to AI5 and AI6 chips will deliver on what Musk promised is to be seen in the coming months.