Get ready for a thrilling ride into the world of Tesla's cutting-edge technology! We've got some exciting news that will blow your mind and spark some serious debate.
Tesla's New Chip: A Game-Changer for Autonomous Driving?
In a recent development, Tesla's Electronic Parts Catalog has hinted at the release of a new HW 4.5 FSD computer, potentially revolutionizing the way their vehicles think and make decisions.
The Findings Unveiled
Our eagle-eyed observer, Todd deRego, spotted a new component listed in Tesla's official catalog, indicating a potential upgrade to their FSD system. The part, named 'CAR COMPUTER - LEFT HAND DRIVE - PROVISIONED - HARDWARE 4.5', carries a price tag similar to other FSD computers, suggesting it's an enhanced version rather than a brand new addition.
The 3-SoC Mystery
Longtime Tesla firmware hacker, Green, has shed some light on this mystery. According to him, Tesla's firmware has long referenced a 3-SoC (System-on-Chip) design. Traditionally, Tesla's FSD computers have used a dual-SoC setup, providing standard redundancy. But with a 3-SoC architecture, the possibilities are endless.
Why Go for 3?
The third node could unlock a whole new world of capabilities. Firstly, it boosts raw throughput, enabling larger and more complex inference models. As FSD v14 and future models grow exponentially, this extra computing power is crucial. Additionally, a 3-chip system introduces Triple Modular Redundancy (TMR), ensuring fault tolerance and an extra layer of safety. Imagine a scenario where one chip hallucinates an obstacle, but the other two see a clear road - TMR would allow the system to 'vote' and continue driving smoothly.
A Bridge to AI5
With Tesla's next-generation FSD chip, AI5, slated for production later this year, there's a gap to fill. Hardware 4.5 could be the perfect bridge, ensuring that millions of Tesla cars aren't left behind as neural networks evolve. By introducing a slightly more capable iteration now, Tesla ensures their vehicles stay ahead of the curve.
Unsupervised Robotaxi Rides: A Historic Moment
In a groundbreaking move, Tesla has begun offering unsupervised Robotaxi rides to the public in Austin, Texas. As of January 22, 2026, riders can hail a Model Y Robotaxi with no one in the vehicle. Until now, Tesla's Robotaxi tests included a human safety monitor, but now, that safety net is gone.
A New Era of Autonomous Transit
With this launch, Tesla joins the elite club of companies operating truly driverless public transit, alongside Waymo and Zoox. However, Tesla's approach differs significantly. While Waymo relies on expensive LiDAR arrays and pre-mapped geofences, Tesla achieves the same with cameras and AI. If this pilot continues successfully, it validates Tesla's strategy and paves the way for further expansion across Texas and other US states.
AI5: The Future is Here
Tesla's next-generation chip, AI5, is set to enter production in the second half of 2026. According to reports from Korea, Samsung's foundry in Taylor, Texas, is preparing for mass production. AI5 will be a game-changer, advancing FSD capabilities and rivaling NVIDIA's $30,000 chip in performance.
The Stakes are High
Elon Musk has admitted that Tesla's future depends on the success of AI5. The current AI4 computer delivers impressive results, but AI5 takes it to the next level. With its impressive compute density, AI5 will power Tesla's autonomous vehicles and their humanoid robot, Optimus. Tesla's commitment to FSD Unsupervised and the development of Cybercab units equipped with AI4 chips showcases their confidence in this technology.
A Supercomputer in Your Car
Elon Musk has quantified the performance of AI5, comparing it to NVIDIA's industry-standard GPU, the H100. A single SoC AI5 is equivalent to one H100, while a dual AI5 system rivals NVIDIA's B100/B200 Blackwell. This means future Tesla vehicles will carry a supercomputer-equivalent GPU, rivaling many powerful AI server nodes.
The Future is Bright
With AI5 architecture finalized, Tesla's focus shifts to Dojo 3 (AI6). While AI5 powers the cars, Dojo powers the server and inference training. The hardware to run neural nets is almost ready, and now the focus is on teaching those networks.
So, what do you think? Is Tesla's new chip a game-changer? Will it revolutionize autonomous driving? And what about their unsupervised Robotaxi rides? Join the discussion and share your thoughts in the comments!
