That's ridiculous!! Lao Ma has come to reduce the dimensionality and strike again! The AI5 chip has been announced to be completed, and the computing power behind Grok and various AI applications has taken off directly! Elon Musk announced that Tesla's AI5 chip has been successfully taped out and is simultaneously advancing the development of the AI6 and Dojo 3 supercomputer projects. This incident established Tesla's role transformation in the computing hardware chain, officially shifting the company from Nvidia's GPU purchaser to an AI chip developer directly involved in the underlying architecture design. Data and indicators show that the AI5 chip aims to replace AI4 hardware and achieve a performance improvement of approximately 40 times. Its single SoC system computing power is benchmarked against Nvidia's Hopper architecture, with dual chip configurations aligned with Blackwell. By removing the general graphics processing module and optimizing for neural networks, the expected power consumption of AI5 is controlled within the range of 150 watts to 250 watts, which is lower than the 700 watts of Nvidia H100. On the supply chain side, Tesla has placed a $16.5 billion 2-nanometer process order with Samsung, simultaneously activating TSMC's Arizona factory production capacity and introducing Intel's EMIB technology to perform multi die packaging testing for Dojo large modules. In August 2025, Tesla disbanded the Dojo project due to the loss of core personnel, and its computing power training relied entirely on external hardware. With the AI5 design finalized, Tesla will quickly restart the Dojo 3 in 2026. Despite Musk's statement that the company is operating a computing power cluster equivalent to 81000 H100 and will continue to use Nvidia products, Nvidia's exclusivity in Tesla's future incremental orders has been broken as AI5 enters the 2027 production milestone. Musk made it clear in his previous internal statement that "solving AI5 is a survival issue for Tesla. AI5 is no longer a single on-board component, but is set as the central computing power of the entire ecosystem, synchronously driving FSD fully autonomous driving, Optimus humanoid robots, and factory automation networks. According to the production capacity plan, AI5 chips that have not been consumed by terminal hardware will be directly deployed in Tesla's self built data center for model training. Tesla has completed the self-developed wafer and packaging layout of customized silicon wafers by crossing the division of labor system between Samsung, TSMC, and Intel. Its business model has moved away from the traditional evaluation system of car companies that focuses on sales and manufacturing gross profit, and has shifted towards establishing a closed-loop network based on autonomous computing infrastructure and directly connected to edge intelligent terminals.