Customer care
+34 913 67 09 92
Follow us

Tesla Begins Production of Advanced Dojo Processor

Tesla Boosts Innovation with Its Own Processor Tesla, the electric car giant, has begun manufacturing its processor known as the “Dojo Training Tile.” This marks a significant milestone in its efforts to internally develop its AI training compute capacity. Although it already possesses a NVIDIA GPU-based supercomputer, one of the most powerful in the world, the new Dojo computing system has been custom-built using proprietary chips and infrastructure.

Development and Capabilities of Dojo Tesla’s first-generation Dojo supercomputer went operational last summer, specifically designed for AI machine learning and, more precisely, for video training using video data from its vehicle fleet. Concurrently, Tesla was working on the sequel: a system-on-wafer processor comprising a 5×5 array of processor chips placed on a carrier wafer and interconnected using TSMC’s integrated fan-out (InFO) technology for wafer-scale interconnections (InFO_SoW).

Production and Future Plans At TSMC’s North American Technology Symposium, it was confirmed that the next-generation Dojo chip has entered production. Furthermore, it was announced that both companies are working on a third-generation Dojo, expected to be released in 2027. This new processor appears to be intended for use in Tesla’s planned $500 million Dojo cluster in New York.

Implications and Future Expectations With these developments, Tesla not only reinforces its leadership in technological innovation but also positions itself as a key player in the field of AI computing. The scalability and performance of the Dojo promise to revolutionize large-scale data processing, especially in AI applications that require intensive analysis of large volumes of visual data.