● On the self-driving and semiconductor industries
☆Tesla is expected to become a very important industry from now on. It will also be connected to the last post, On-device AI.
☆This will be included in the book you are currently writing.
Tesla will unveil the robotaxi on Oct. 10. Elon Musk claimed in an investor day presentation back in April 2019 that he would have 1 million robotaxi on the road by 2020. More than five years after the claim, the concept of a self-driving taxi was finally unveiled. According to Automotive News, the venue for the event is the Warner Bros. studio site in Los Angeles, California. The Warner Bros. site is an ideal place to demonstrate the product’s functionality, free from problems on public roads, as it consists of a network of interconnected private roads, so the presentation is expected to proceed normally.
Tesla posted its FSD launch roadmap on Tesla AI’s X (formerly Twitter) account, which promotes its AI technology, the day before. It is raising expectations, with plans to launch full self-driving (FSD) in Europe and China in the first quarter of next year only leaving regulators to approve it. Tesla’s roadmap also included plans to launch a version of the v13 in October, which increases the distance the vehicle travels autonomously by six times more than before, until driver intervention is required during FSD operation.
First, to explain Tesla’s autonomous driving, Tesla intends to integrate its advanced driver assistance system, Autopilot, with full self-driving (FSD), thereby minimizing driver intervention and ultimately fully autonomous driving. Autopilot is the concept of advanced driver assistance (ADAS) and is embedded in basically all Tesla vehicles. Tesla implements camera sensor-based autonomous driving technology and provides wireless software updates through over-the-air (OTA) technology. FSD is an advanced autonomous driving technology that aims to be fully autonomous. FSD offers more advanced features than the Autopilot system, and is designed to allow the vehicle to drive independently without driver intervention. Tesla recognizes the surroundings of the vehicle through a sensor fusion including eight cameras, 12 ultrasonic sensors, and a front radar. In the processing of vast amounts of data collected from sensors and determining the driving decisions of the vehicle based on them, Tesla uses its own independently designed FSD chip. When you purchase a vehicle, you can choose the FSD package as an additional option to the price of the vehicle, and it’s also available as a monthly subscription model. As of 2024, the FSD package has a temporary purchase price of around $15,000. It costs around $299 per month for a monthly subscription, and in some markets, you can subscribe to the Enhanced Autopilot feature separately, which costs around $99 per month. Enhanced Autopilot includes basic autonomous driving features and excludes some advanced features of FSD.
What’s scary about Tesla is that it’s in the business of being paid for this time as a reward while saving humanity’s time. Tesla’s electric vehicles will ultimately accelerate their penetration for autonomous driving. While the electric vehicle market passes through Casm, Tesla can increase its market share by pushing back on volume. In a bad situation, other OEMs have to bear the brunt of increasing production, but Tesla can push back on volume at the expense of this loss to cover its margins with SW. Tesla is now all set for its evolution into a software company. Both Apple and Nvidia have experienced significant gains in stock prices as HW companies in the past, evolving into SW. We expect Tesla to follow that path in 2025.
Tesla released the FSD Beta 12 in August 2023. It incorporates a complete Tesla vision system and improves urban self-driving capabilities to improve left-turn and complex intersection processing performance. Tesla is further refining its FSD capabilities to be recognized for full self-driving through continuous updates. Recently, in July, the FSD 12.5 version was released, which should be applied to future Cybertrucks as well. This FSD 12.5 version will most likely be the first to surpass human capabilities in public. This time, the FSD 12.5 version of the MPI (mile per intervention) is 448 miles, a significant increase of three to four times over the previous version. When the MPI is 448 miles, it means that the driver will have to travel 448 miles (about 700 kilometers) with the FSD to intervene once. The reason for the evolution from the 12.5 version is that we learned by increasing the model’s per liter by five times based on HW4.0. The existing model, FSD 12.4, learned HW3.0 as the main model, and this new model, FSD 12.5, learned HW4.0 as the main model, which Elon Musk says has room left for HW4.0 performance to increase eight times more in the new parameter model. After the completion of the larger cluster of more than 100,000 H100,000 currently being built next to Giga Texas, it is expected to be a level-up with increased parameters again.
[Mobility and car semiconductors]
The growth of smartphone and PC-led markets in the semiconductor industry will shift to mobility in the future. The semiconductor market is expected to grow little from about $210 billion in 2022 to about $120 billion by 2027 in the smartphone and PC industries, but the automotive semiconductor market is expected to grow from $70 billion in 2022 to about $120 billion in 2027, accounting for 60% of the combined smartphone and PC markets. If there are approximately 200 semiconductors installed in internal combustion engines, the number of semiconductors installed in electric vehicles will increase exponentially to 1000, and the number of semiconductors installed in autonomous vehicles will increase to 2000.