Tesla, Inc. (TSLA.US) successfully taped out the A15 chip! Musk exclaimed: It will become one of the highest AI chip producers in the world in the future.

Elon Musk, CEO of Tesla, Inc. (TSLA.US) announced that Tesla, Inc.'s AI chip design team has successfully completed the sampling of the AI5 chip. This marks an important milestone in the chip's move towards the manufacturing phase. Musk posted on social media, saying, "Congratulations to Tesla, Inc.'s AI chip design team for completing the sampling of the A15 chip! The AI6, Dojo3, and other exciting chips are also in development." Musk also thanked partners Taiwan Semiconductor Manufacturing Co., Ltd. Sponsored ADR (TSM.US) and Samsung for their support in the chip production process, stating that the AI5 chip will become one of the highest volume AI chips globally in the future. It is reported that the AI5 is Tesla, Inc.'s fifth-generation AI chip, mainly aimed at ground applications. Its performance goals target industry-leading levels, with performance equivalent to NVIDIA Corporation's Hopper architecture in a single chip configuration, and comparable to NVIDIA Corporation's Blackwell architecture in a dual-chip configuration, while having lower costs and power consumption. This chip will be installed in all Tesla, Inc. models, future autonomous taxis (Cybercab), and Optimus humanoid Siasun Robot&Automation, forming the core of Tesla, Inc.'s ground computing power layout. Musk had previously mentioned that the AI5 chip is planned to enter mass production by 2027, aiming to replace the AI4 chips currently used in Tesla, Inc. vehicles. Sampling is a crucial step before mass production of chips, testing whether the designed chip is usable, and preparing for subsequent large-scale production. The successful sampling of the A15 chip means that the chip has taken a solid step towards mass production. Following the A15 chip is the A16 chip, which is currently in the development stage. Musk had mentioned last month that the sampling of the AI6 chip may be completed in December. The iteration cycle between the A15 and A16 chips is only 9 months, whereas the industry standard for chip iteration usually takes over 18 months. With its in-house chip development, autonomous driving algorithms, and closed-loop ecosystem for vehicle manufacturing, Tesla, Inc. can achieve rapid parameter tuning and real-time feedback, significantly reducing the development cycle to 9 months. The Tesla, Inc. AI5 and AI6 chips - serving as the unified "brains" for millions of autonomous taxis and Optimus humanoid Siasun Robot&Automation, as well as the Dojo 3 chip - tailored for the demanding yet highly potential space environment - are important "members" of the Terafab superchip factory project officially launched by Tesla, Inc. at the end of March. The Terafab project is expected to achieve an annual output of over 1 terawatt (1TW) of computing power, equivalent to approximately 50 times the current global annual output of AI chips. Musk had stated that this will be the largest chip manufacturing project in history. At the shareholder meeting in 2025, Musk first articulated the concept of the "Terafab" chip factory, expressing that the existing suppliers' capacity would not be able to meet future demands, necessitating the construction of a factory. He reiterated the plan in earnings calls and on social media multiple times, emphasizing the decisive significance of chip manufacturing for the core business. Musk mentioned that to support the continuous iteration of Tesla, Inc.'s Full Self-Driving (FSD) software and the large-scale deployment of Optimus Siasun Robot&Automation, Tesla, Inc. will require 100 to 200 billion chips annually, even if global foundries expand their production capacities optimally, their capacity would not be able to meet Tesla, Inc.'s explosive demand for AI chips in the future, hence the imperative need to build a large chip factory. Musk stated, "We are very grateful for the existing supply chain, we thank companies like Samsung, Taiwan Semiconductor Manufacturing Co., Ltd. Sponsored ADR, and others, hoping that they can expand their scale quickly. We will buy all their chips. I have already told them this, and the expansion rate that they can accept has a limit, which is far below our expectations, so we need to build Terafab." According to calculations, after the Terafab project is put into operation, Tesla, Inc.'s overall chip costs can be reduced by 50% to 70%, computing power electricity costs will only be 1/3 to 1/5 of ground photovoltaics, product iteration speed will increase by 2 to 3 times, greatly enhancing the cost advantage and technological leadership of Tesla, Inc. and its affiliated companies. However, entering the semiconductor production field also means that Tesla, Inc. will face high investment and high-risk dual challenges. Despite having over $40 billion in cash reserves, facing the potential total cost of the Terafab project reaching the hundreds of billions of dollars, Tesla, Inc. still faces immense financial pressure. Additionally, lacking experience in semiconductor manufacturing as an automotive manufacturer, starting from scratch to build a wafer fab with 2-nanometer process will face multiple challenges including equipment procurement, talent shortage, and yield control. Therefore, this grand plan will present Tesla, Inc. with significant financial and technological challenges. Currently, the Terafab project's official website is live, and positions related to the Tesla, Inc., SpaceX, and xAI systems are open for recruitment, indicating that the project has entered a substantive advancement stage. Industry analysts believe that Terafab is not only a key step in Tesla, Inc.'s vertical integration of the industry chain, but also a major exploration of the global computing power landscape and chip industry.