Tesla tapes out AI5 chip
Tesla announced it has completed the tape‑out of its next‑generation AI5 processor as it prepares AI6 and Dojo3 projects. The company framed this as part of its push into custom silicon for autonomous driving and AI workloads. (ibtimes.com.au)
Tesla said its AI5 chip has reached tape-out, the step where a processor design is frozen and sent to a factory for fabrication. (ibtimes.com.au) Elon Musk posted the update on April 15, 2026, writing that Tesla had taped out AI5 and that “AI6, Dojo3 & other exciting chips” were also in development. He separately thanked Taiwan Semiconductor Manufacturing Company and Samsung, saying AI5 “will be one of most produced AI chips ever.” (ibtimes.com.au) A tape-out does not put the chip into cars. It means Tesla’s engineers have finished the layout and handed it to manufacturing partners for first silicon, testing and later production. (electrek.co) Tesla’s current self-driving computers already run custom in-house chips, and the company says its artificial intelligence work depends on “inference hardware,” or processors that turn trained software into real-time driving decisions inside vehicles. Tesla says those chips support Full Self-Driving, humanoid robots and other systems. (tesla.com) The timing matters because Tesla’s own 2025 fourth-quarter update said production of its in-house AI5 and AI6 inference chips is planned for 2027 and 2028. That puts AI5 at least a year away from large-scale deployment even after this design milestone. (assets-ir.tesla.com) Tesla is also tying its chip work to a broader push beyond passenger cars. In its January 28, 2026 shareholder update, the company said it was shifting from a “hardware-centric business” toward a “physical AI company” built around robotaxis, Optimus robots and training infrastructure. (assets-ir.tesla.com) That strategy is already visible in products Tesla has put on sale. Tesla’s Robotaxi page says autonomous rides are now being offered in Austin, Texas, starting with Model Y, while Cybercab remains a future vehicle for the same network. (tesla.com) Tesla’s artificial intelligence page says the company is designing chips to maximize “performance-per-watt,” a measure of how much computing work a processor can do for the electricity it uses. In a car, that tradeoff affects heat, cost and how much computing hardware can fit behind the dashboard. (tesla.com) The tape-out also lands later than Tesla’s earlier roadmap implied. Electrek reported that the milestone came nearly two years after Tesla had originally said AI5 would already be in vehicles, highlighting the long gap between chip design goals and mass production. (electrek.co) For now, Tesla has a finished AI5 design, outside foundries lined up, and later chips already being named before AI5 reaches volume production. That leaves 2026 as a year of fabrication and testing, not yet a year of AI5-powered cars on the road. (electrek.co)