Tesla has recently completed the design of its AI5 chip, with CEO Elon Musk confirming that the chip has reached the tape-out stage, marking the final step before mass production. However, Musk clarified that the new chip will not be used to achieve a safety level better than human drivers for Full Self-Driving (FSD), as the existing AI4 chip is already capable of doing so.
Instead, the focus of the AI5 chip will be on Tesla’s future projects, including Optimus and supercomputer clusters. Musk expressed his gratitude to TSMC and Samsung for their support in bringing the chip to production, stating that AI5 has the potential to become one of the most produced AI chips ever. This strategic decision marks a shift in Tesla’s approach, as Musk emphasized that vehicles no longer require the latest silicon technology.
The AI4 hardware, which is currently deployed in hundreds of thousands of Teslas equipped with HW4, has already proven to deliver superior safety metrics compared to human drivers for FSD. As a result, AI5 will be used to accelerate the development of Optimus robots and massive training clusters similar to Tesla’s Dojo platform.
While AI4 is sufficient for achieving better-than-human safety for FSD, Musk hinted that it may enable unsupervised self-driving with certain limitations. Tesla’s FSD stack runs neural networks trained on real-world data, with AI4-equipped vehicles reportedly outperforming human drivers in various safety metrics. With dual-redundant AI4 chips providing ample computing power for driving tasks, Tesla can focus on developing advanced applications like dexterous robots and training at scale.
However, achieving unsupervised autonomy is not just a matter of computing power but also regulatory approval. Agencies like the NHTSA require extensive validation, liability frameworks, and public trust before granting approval for autonomous driving. While competitors like Waymo operate limited unsupervised fleets in geofenced areas, Tesla’s vision-only approach on a fleet scale presents a more ambitious challenge in terms of certification.
In conclusion, Musk’s decision to prioritize AI5 for future projects while leveraging the capabilities of AI4 for FSD demonstrates Tesla’s strategic direction towards capital-efficient innovation. If safety data continues to support the capabilities of AI4, unsupervised autonomy could become a reality sooner than expected. This pragmatic yet optimistic approach highlights Tesla’s commitment to pushing the boundaries of autonomous technology while ensuring the safety and relevance of its existing fleet.