Tesla CEO Elon Musk announced on Wednesday that the company is reviving its Dojo supercomputer project, which it halted last year. He had canceled the program less than a year ago to prioritize acquiring Nvidia chips, but now admits that decision was a mistake.
According to reports, Tesla's new AI5 chip is ready and will form the basis of a new supercomputer called Dojo3. The company has started hiring engineers for 9-month development cycles, aiming for the project to yield tangible results in a short period.
Musk stated that Tesla made a significant strategic error by shutting down Dojo. The team had worked on custom AI chips for years, but these efforts were put on hold because Nvidia's H100 chips appeared more attractive. Now, the company is returning to its internal chip and computing infrastructure.
New details were also shared regarding space-based data centers during the announcement. According to Musk, the main limitation is cooling, and space solves this problem. He indicated that it is "extremely possible" for Tesla to utilize orbital computing technologies in the future.
Following the announcement, Tesla shares surged. Investors view this as a significant step by the company to reduce its dependence on Nvidia's supply chains and pricing policies, and to build an independent infrastructure in the field of artificial intelligence.
Analysts note that by halting the Tesla Dojo project after spending billions of dollars, the company weakened its negotiating position. Amidst the high cost and limited availability of Nvidia chips, companies that canceled their own chip projects now realize they have lost their bargaining power.
Dojo3 is considered particularly important for autonomous vehicles. Tesla vehicles generate large volumes of training data, and custom-designed chips for this data could be more efficient than general-purpose Nvidia chips. However, developing new chips in 9 months is considered a very rapid timeframe, and some experts suggest it carries risks, implying either a quick solution has been found or delays may be inevitable.