Tesla CEO Elon Musk recently announced that with the design of the AI5 chip now complete, Tesla (TSLA) will reboot the development of its Dojo 3 supercomputer project. The Dojo project aims to provide massive computing power for autonomous driving systems and AI models through self-developed chips and systems, reducing reliance on external suppliers.

Analysts believe the AI5 chip will support more complex Full Self-Driving (FSD) algorithms. The progress of the chip’s mass production will directly influence the rollout speed of Tesla’s FSD features and could become a critical pillar for its robotics business.

Musk’s Announcement

On January 19, Musk posted on the social media platform X that Tesla will restart the development of the Dojo 3 supercomputer project following the completion of the AI5 chip design.

Simultaneously, he posted recruitment information seeking talent interested in “developing the world’s highest-volume chip,” requiring applicants to summarize key technical challenges they have solved in three bullet points.

In subsequent posts, Musk emphasized: “Solving the AI5 chip issue is critical for Tesla. Therefore, I had to have both teams focus on this chip’s development, and I have personally spent every Saturday on it for several months.”

Musk noted that AI5 will be an exceptionally powerful chip: a single SoC’s performance is roughly equivalent to Nvidia’s Hopper(NVDA) class, while a dual-chip configuration approaches Blackwell levels—but at a significantly lower cost and power consumption. “With AI5 progressing smoothly, we finally have some bandwidth to restart the R&D for Dojo 3,” he stated.

This marks a strategic reversal. In August 2025, reports suggested that Tesla had fully suspended the Dojo project, leading to the departure of project lead Peter Bannon. At the time, the move was interpreted as Tesla abandoning its self-developed autonomous driving chip plan.

Musk previously explained that it made little sense for Tesla to divide resources between two vastly different AI chip designs. He noted that Tesla’s AI5, AI6, and subsequent chips would excel in inference and perform well in training, and that all efforts would be concentrated there. He added that integrating multiple AI5/AI6 chips onto a single circuit board for supercomputer clusters could reduce networking complexity and costs by several orders of magnitude.

First mentioned in 2019, the Dojo project carries Tesla’s grand vision for AI. It is designed to optimize neural network models and process autonomous driving video data. Morgan Stanley previously estimated that a fully operational Dojo could potentially add billions of dollars to Tesla’s valuation.

A Decisive Battle for Autonomous Driving

Musk recently revealed that the AI5 chip for FSD is nearing design completion, while AI6 is in its early stages. Tesla aims to complete design cycles for AI7, AI8, and AI9 within a nine-month cadence.

According to previously disclosed data, the AI5 chip will deliver 2,000–2,500 TOPS of computing performance—roughly five times that of the current HW4 chip—enabling more sophisticated FSD algorithms.

Sampling and small-scale deployment of the AI5 chip are scheduled for 2026, with full mass production expected in 2027. As AI5 nears the finish line, Tesla has initiated early work on AI6, which is expected to launch in 2028. AI6 will likely continue Tesla’s foundry partnership with Samsung Electronics, utilizing a modular architecture deeply integrated with the Dojo supercomputer ecosystem to create synergy across vehicles, robots, and supercomputing.

Recent reports indicate that Samsung Electronics is accelerating preparations for AI5 production at its U.S. facilities, recruiting experienced engineers to stabilize yields and ensure a smooth manufacturing process for Tesla.

Shift to FSD Subscription Era

Alongside hardware updates, Tesla’s FSD strategy is undergoing a major shift. Musk recently announced that starting February 14, Tesla will discontinue the one-time purchase option for FSD in favor of a monthly subscription model.

This marks the end of a decade-long era of one-time buyouts (previously $8,000 in the U.S. and 64,000 RMB in China). Analysts point out that this “SaaS” (Software as a Service) approach aims to lower the barrier to entry, increase penetration, and generate recurring revenue.

Musk has hinted that the next version of FSD will achieve full autonomy, even admitting that Robotaxi driverless testing has already begun. Morgan Stanley views FSD 14.3 as a potential “Steam Engine Moment” for autonomous driving, which is shaping up to be a primary investment theme for Wall Street in 2026.