Tesla Revives Dojo Supercomputer Project

Tesla is officially reviving its in-house supercomputer initiative, known as Dojo, according to a recent statement from CEO Elon Musk. The project had been shelved last year as the electric vehicle manufacturer prioritized the development of onboard AI chips.

The decision to return to the ambitious computing project comes after the company achieved a significant milestone in its hardware development. Musk indicated that the team is resuming work on the third generation of the supercomputer now that the design for the upcoming AI5 chip is finalized.

The Dojo project was effectively paused in the previous year when Tesla leadership made a strategic decision to consolidate resources. At that time, the company disbanded the specialized Dojo team to focus exclusively on the AI chips designed to operate directly within Tesla vehicles.

Musk previously explained that dividing resources between two distinct AI chip architectures was inefficient. The company's focus shifted to the AI5 and AI6 chips, which are optimized for inference—the process of running the Full Self-Driving software in real-time on the car's hardware.

However, with the vehicle-side chip design now in good shape, resources are once again available to tackle the massive computational needs of training the neural net.

Unlike the chips installed in cars, Dojo is built for a different purpose: training. The supercomputer is designed to ingest and process the immense volume of video recordings and other data collected from Tesla's global fleet of vehicles.

This data processing is critical for improving the neural net that powers the company's autonomous driving software. By analyzing real-world driving scenarios, Dojo helps the system learn and adapt to become safer and more reliable.

The renewed effort suggests Tesla is preparing for the next phase of autonomy, where the sheer volume of data requires specialized, high-performance computing infrastructure to turn that data into actionable intelligence.

While Tesla designs its own silicon, the manufacturing relies on external partners. The upcoming AI6 chips are slated for production at Samsung's facilities in Texas.

This manufacturing arrangement is backed by a massive financial commitment, reportedly valued at $16 billion. The deal underscores the scale of Tesla's ambition to secure a robust supply chain for its proprietary hardware.

Securing domestic manufacturing capacity for these advanced chips is a strategic move that aligns with broader industry trends toward localized high-tech production.

The vision for Dojo3 extends far beyond terrestrial data centers. Musk has floated the concept of a space-based AI compute architecture, proposing that data centers in orbit could offer significant advantages over ground-based facilities.

The theoretical benefits include:

• Uninterrupted access to solar energy
• Naturally cold vacuum environment reducing cooling costs
• Reduced latency for satellite data processing

Despite the increasing popularity of orbital data center concepts, the idea remains largely speculative. Experts in the field have expressed doubts regarding the feasibility and economic viability of deploying such complex infrastructure in space.

The return to Dojo3 represents a significant expansion of Tesla's technical roadmap. By resuming work on the supercomputer, the company is signaling that it is ready to handle the next generation of data processing required for fully autonomous driving.

However, the project carries inherent risks. Musk's history of making bold assertions that fail to materialize casts a shadow of uncertainty over the timeline and final capabilities of the space-based compute plans.

Observers will be watching closely to see if Tesla can translate these ambitious hardware plans into a functional supercomputer that delivers on the promise of full autonomy.