Tesla’s pursuit of autonomous driving relies heavily on significant computational capabilities. Today, TSMC has confirmed the initiation of manufacturing for Tesla’s next-gen Dojo supercomputer training chips, promising a substantial leap in computing power by 2027.

According to a TechNews report, Elon Musk’s strategy emphasizes the crucial role of software in achieving profitability, necessitating both efficient algorithms and robust computational prowess. Tesla employs a dual hardware approach, acquiring tens of thousands of NVIDIA H100 GPUs while concurrently developing its proprietary supercomputing chip, Dojo. TSMC, reportedly responsible for producing these supercomputer chips, has affirmed the commencement of production.

As previously outlined in another TechNews report, Tesla’s focus primarily centers on the demand for autonomous driving, resulting in the introduction of two AI chips: the Full Self-Driving (FSD) chip and the Dojo D1 chip. The FSD chip powers Tesla vehicles’ autonomous driving systems, while the Dojo D1 chip fuels Tesla’s supercomputers, acting as a versatile CPU for constructing AI training chips that drive the Dojo system.

During the North American technology conference, TSMC provided comprehensive insights into semiconductor technology and advanced packaging, facilitating system-level integration across entire wafers and delivering ultra-high computing performance. TSMC announced the commencement of production for next-gen Dojo training modules for Tesla, aiming to surpass current systems’ computing power by over 40 times by 2027.