Auto Chip Rivalry Escalates: From Computing Power to Control of Smart Vehicles

Recently, competition in the automotive chip industry has intensified significantly. BYD unveiled its self-developed 4nm intelligent driving chip, "Xuanji A3," delivering approximately 700 TOPS of computing power per chip and up to 2,100 TOPS when three chips operate in parallel. The company emphasized the chip’s safety certification level, power efficiency, and algorithm compatibility. Other automakers—including NIO, XPeng, and Li Auto—are also advancing their own chip development or AI computing architectures, such as NIO’s "Shenji NX9031," XPeng’s "Turing AI Chip," and Li Auto’s proprietary AI inference architecture. Meanwhile, soaring memory chip prices have become a tangible pressure point for automakers and their supply chains. Surging demand from AI data centers for HBM, DRAM, NAND, and advanced packaging resources has led major global memory manufacturers—such as Samsung, Micron, and SK hynix—to prioritize AI computing customers and reallocate capacity accordingly. This shift is squeezing the automotive sector’s supply security and bargaining power, potentially triggering shortages in critical chip categories and impacting vehicle costs, configuration strategies, and delivery timelines. Technologically, the electronic/electrical (E/E) architecture of intelligent vehicles is evolving from distributed ECUs toward centralized computing and integrated cabin-driving systems. This transition demands chips that not only deliver high computing performance but also support functional safety, real-time isolation, redundancy design, and tight hardware-software co-optimization to handle complex tasks like end-to-end autonomous driving, multimodal interaction, and on-vehicle AI agents. Chips are no longer mere hardware components to be procured—they have become core elements of the vehicle’s intelligent architecture. If Tesla’s Supervised Full Self-Driving (FSD) system further expands and matures in China, it will intensify local competition in intelligent driving. Tesla’s end-to-end model system, validated by its global fleet data, could push Chinese automakers to shift their focus from “feature implementation” to competing on “system-level capabilities,” encompassing experience stability, generalization ability, iteration speed, and cost efficiency. Third-party chip companies are also repositioning themselves. At its 2026 Automotive Technology Summit, Qualcomm highlighted integrated cabin-driving solutions, Physical AI, and on-vehicle ecosystems, aiming to establish itself as a gateway for intelligent vehicle computing and software. Horizon Robotics offers an integrated suite of chips, algorithms, toolchains, and mass-production solutions but faces mounting pressure to reassess its business model amid the rising trend of in-house chip development by automakers. Beyond high-performance main chips, smaller components—such as automotive Ethernet PHYs, BMS analog chips, and 4D millimeter-wave radar chips—are gaining importance under centralized computing architectures, collectively forming the foundational technology layer of intelligent vehicles. Overall, competition in automotive chips has evolved beyond raw computing power metrics into a comprehensive battle over whole-vehicle intelligence capabilities, E/E architecture design, cost control, supply chain security, and the right to define next-generation autonomous driving.