Tesla adds FSD supercomputing capacity at Giga Texas
Musk confirms capacity expansion for FSD training but Dojo progress in unclear
EV market leader Tesla is installing a supercomputer cluster as part of its ongoing expansion of its Texas gigafactory, CEO Elon Musk has revealed, as the firm positions itself to prioritise its AI and autonomy software as the core driver of future growth.
The move is part of the firm's planned pivot towards AI and robotics, which is to be led by its Full Self Driving (FSD) ADAS software, as Musk revealed a new mission statement for the company at the end of Q1 which could see traditional auto manufacturing relegated amongst Tesla' primary business plays.
"The rear portion of the factory extension [...] will be a super dense, water-cooled supercomputer cluster," Musk wrote on social media site X in response to a photo posted showing building work at Giga Texas.
Armed with an aging product line-up and with few signs that its EV deliveries will pick up their prior pace, Musk has been unequivocal about Tesla's investment thesis lying in its software plays, telling analysts last month that "if somebody does not believe Tesla is going to solve autonomy, I think they should not be an investor in the company".
FSD version 12.4 recently began rolling out to Tesla employees, but the software is still only a Level 2 ADAS system which requires driver attention at all times. The latest upgrade even introduces cabin cameras being used to monitor the driver's attentiveness.It is still unclear, however, whether Tesla will make use of its in-house supercomputer Dojo, or if the new capacity will be comprised of computing power bought from suppliers such as Taiwanese firm Nvidia. While Tesla says Dojo is performing certain limited training tasks, the company has been ordering hardware such as GPUs from Nvidia, which Musk has described as "hedging our bets".
Tesla is set to invest $500mn over the next five years to develop its new Dojo supercomputer at its Riverbend site in Buffalo, NY, although the it is possible that Tesla's relatively sudden acceleration of its AI and software plans since this announcement has brought the need for computing capacity forward.
Tesla is set to unveil a robotaxi product in August this year, although much remains unknown about what precisely the company will launch. Musk announced this launch date in April amid widespread consternation at Tesla's decision to pivot away from an affordable next-generation BEV produced entirely from scratch, in favour of far less certain software ventures.
Tesla's future success appears to lie in being able to continually improve is FSD to the point of Level 4 autonomy, which will not be achieved by the August product launch. Enormous computing capacity is needed to train software on billions of miles of driving data that Tesla collects from its cars currently on the road. To handle some of this data capacity, the firm is also reportedly exploring data centre sites in China, with Musk hinting that further expansion to Tesla's compute power is on the cards.
“It is possible, with the right architectural decisions, that Tesla may in the future have more compute than everyone else combined,” Musk said during the firm's Q1 results call.