Intelligent computing power in the automotive cloud becomes the key to competition among automakers

Around 2015, the technology for sensing targets and obstacles matured, and the demand for vehicle-side computing chips began to emerge; after 2020, the perception modules were gradually unified into neural networks, and the cloud computing power of automobile companies increased significantly; with the rise of end-to-end autonomous driving in the past two years, the demand for computing power has further increased rapidly. The growth of Tesla's cloud computing power can prove this trend. In 2019, Tesla had less than 1,500 GPUs; in 2021, it had a cluster of more than 5,760 A100 GPUs, with a single cluster computing power of 1,800 PFLOPS; in August 2023, a cluster of 10,000 H100 GPUs was launched; in the first quarter of 2024, it had reached the equivalent computing power of 35,000 H100 GPUs.