What happened

Elon Musk announced that Tesla is restarting work on Dojo3 — the third generation of its in‑house supercomputer for training autonomous driving systems. The project was wound down in 2025 and the team was disbanded to redirect resources toward developing the AI5 and AI6 chips.

"Returning to Dojo became possible after completing the AI5 project"

— Elon Musk, founder and CEO of Tesla

Why it matters

The decision is not merely technical: it shows how companies have balanced local, in‑vehicle data processing and centralized training of large models. Essentially Tesla completed the phase of creating compact, on‑vehicle chips (AI5), and is now returning to the infrastructure for mass training (Dojo3), which has different performance and energy requirements.

AI5 vs AI6 and Samsung's role

According to Musk, AI5 are optimized for processing data directly in the vehicle — fast and energy‑efficient, but not suitable for training large models. Production of AI6 was entrusted to Samsung: the chips will be manufactured at the Texas plant under a deal worth approximately $16 billion. This signals a long‑term bet on large‑scale production and diversification of supply chains.

Orbital Dojo — a technological fad or a strategic move?

Musk suggested the possibility of placing Dojo3 in orbit — arguing that this would allow better cooling and power from solar energy. The idea is attractive technically, but raises questions of latency, launch costs and data security. Analysts note that orbital data centers may make sense for specific tasks, but are not a universal solution for all model‑training needs.

Monetization of data: autopilot subscription and available models

At the same time Tesla is moving the autonomous driving feature to a monthly subscription model. That means a steady flow of data and money — key resources for large‑scale model training in Dojo3. The availability of the Model Y Standard Long Range in Europe with a range of 657 km also increases the fleet of active sensors generating training data.

"A space data center will allow better cooling of equipment and power it via solar energy"

— Elon Musk, founder and CEO of Tesla

What this means for Ukraine

Tesla's trends are important not only for the automotive industry: they underscore the value of control over chips, compute infrastructure and data flows. For Ukraine this suggests three key lessons:

• Chip independence — local manufacturing capacity or reliable production partners (as in the case of Samsung) reduce risks during force‑majeure events.
• Data and monetization — stable sources of data (services, subscriptions) enable investment in model training and bringing products to market.
• Infrastructure flexibility — the balance between edge (in the vehicle) and training centers (Dojo) is important for military and civilian applications where latency, energy consumption and reliability are critical.

Conclusion

Tesla's return to Dojo3 is not driven by a desire to have it for its own sake, but follows a logical cycle — first develop energy‑efficient chips for the product, then scale up training of large models. For Ukraine, the story is a reminder that technological sovereignty and partnerships (in microelectronics and compute) are not abstract concepts but practical elements of security and the economy. As Dojo3 gains momentum, it's worth watching not only the technical details but also which business models and partners will determine who gains access to the world's new computing power.