Tesla's HW5 FSD Chip to Use TSMC’s 3nm Process, Mass Production Expected in 2026

Tesla is reportedly preparing to launch its next-generation Full Self-Driving (FSD) chip—internally referred to as AI5 or HW5—with mass production targeted for 2026. According to sources cited by Not a Tesla App, Notebookcheck, and Mydrivers, Tesla has partnered with Taiwan Semiconductor Manufacturing Company (TSMC) to fabricate the chip using its cutting-edge 3nm N3P process node, which promises significant improvements in performance and energy efficiency.

AI5: A Leap Forward in Compute Power

The new HW5 chip is expected to deliver a staggering 2,000 to 2,500 TOPS (trillions of operations per second), marking a four- to fivefold increase in performance compared to the current HW4 chip. Tesla’s current HW4 chip, introduced in 2023, is built using a 5nm process, while the older HW3—launched in 2019—was based on a much older 14nm process. This generational jump signifies Tesla’s commitment to staying at the forefront of autonomous driving hardware innovation.

While early production and system integration could begin as soon as 2025, mass manufacturing is not anticipated until 2026. This aligns with the broader industry timeline for high-volume 3nm chip production and testing for automotive-grade reliability.

Strengthening Ties with TSMC

Tesla's collaboration with TSMC has become increasingly pivotal as the EV maker pushes the boundaries of self-driving technology. Earlier in 2025, Taiwan’s Economic Daily News reported that Tesla’s China-based FSD chip upgrades had already utilized TSMC’s 5nm and even 4nm process technologies, further distancing itself from Samsung, which previously supplied older-generation FSD chips and memory components.

Moreover, Tesla’s in-house Dojo supercomputer—which underpins the training of its neural networks and FSD systems—also relies on TSMC's 7nm process and features InFO-SoW advanced packaging. This enables ultra-high-speed data throughput and custom thermal performance to meet the demands of Tesla's proprietary machine learning workloads.

Production Challenges and Robotaxi Delays

Despite the technological strides, Tesla is reportedly facing setbacks in other parts of its operations. Business Insider revealed that the company has temporarily halted production lines for the Cybertruck and Model Y at its Gigafactory in Austin, Texas, for at least one week. The reasons include upstream supply constraints and strategic realignments.

Additionally, Tesla's highly anticipated Robotaxi service—originally scheduled to launch on June 22 in Austin—has been officially postponed. According to Reuters, new state-level regulatory measures have led to a delay until at least September 2025, complicating Tesla's broader autonomous vehicle rollout timeline.

Conclusion

Tesla's adoption of TSMC's 3nm N3P process for its HW5 chip reflects its relentless pursuit of advanced silicon for self-driving applications. While performance benchmarks are promising, real-world deployment will depend on Tesla’s ability to navigate manufacturing scalability, regulatory hurdles, and operational logistics. With the AI5 chip on the horizon, Tesla appears to be laying the technical groundwork for the next era of autonomous electric vehicles.

Previous: What Is Wi-Fi 8 and Why It Matters: The Future of Wireless Connectivity