Tesla is making a notable pivot in its artificial intelligence strategy by reviving the development of its Dojo3 chip, which had previously been put on hold. This announcement, made by CEO Elon Musk over the weekend, indicates a shift from the chip”s original design intended for self-driving car systems to a new focus on “space-based AI compute.”
The decision to restart the Dojo3 project comes just five months after Tesla effectively ended its Dojo program, disbanding the team that had been working on the supercomputer. This team was led by Peter Bannon, who left the company, leading to around 20 Dojo employees transitioning to DensityAI, a new AI infrastructure firm founded by former Dojo director Ganesh Venkataramanan along with ex-Tesla staff.
In recent statements, Musk revealed that rather than collaborating with other chip manufacturers like Nvidia and AMD for computational power, Tesla has decided to halt plans to outsource silicon production and aims to develop its own chips. This strategic turnaround coincides with the company”s progress on its internal chip development timeline, particularly with the AI5 chip, which Musk noted is “in good shape.”
The AI5 chip is produced by TSMC and was designed to enhance Tesla”s automated driving capabilities and the operations of its Optimus humanoid robots. Last year, Tesla completed a significant agreement with Samsung to produce the AI6 chips, which are intended to power Tesla vehicles and facilitate advanced AI training in data centers.
Musk”s recent comments highlight the ambitious nature of the AI7/Dojo3 project, describing it as a “moonshot” initiative aimed at enabling space-based AI computations. To support this renewed effort, Musk is looking to rebuild the workforce that was previously disbanded, actively recruiting engineers through a call to action on social media.
As the timeline aligns with Nvidia“s upcoming presentation of its Alpamayo AI system, which competes with Tesla”s Full Self-Driving (FSD) software, Musk acknowledged the challenges of tackling complex driving scenarios, expressing hope for the success of competing technologies.
The concept of off-planet data centers has gained traction among leaders in the AI industry, with discussions around the need for such facilities given the pressures on terrestrial power grids. Notably, Sam Altman of OpenAI has also shown interest in orbital data centers. Musk”s ties to SpaceX provide a unique advantage, as the company can facilitate the launch of a constellation of computational satellites capable of harnessing solar energy continuously.
Musk has plans to utilize SpaceX”s forthcoming IPO to finance his vision of deploying these satellites, aiming to create a robust infrastructure for space-based computing. This strategic move could significantly impact the landscape of AI and computing technologies in the coming years.
