Software-defined vehicles: the AI-driven race to unlock chip potential
Written by Nuying Huang, Taipei; Rodney Chan, DIGITIMES Asia Tuesday 5 November 2024
Credit: AFP
Software-defined vehicles (SDVs) are at the forefront of automotive electronic innovation, and the surge in generative AI further amplifies software’s transformative power. Yet, as automotive technology and AI advance, a multitude of software and algorithms flood the field, prompting a fundamental question: which software genuinely defines and shapes the hardware’s core?
AutoSys General Manager Chang-feng Hsu believes that determining a software’s effectiveness often rests with semiconductor companies. Software that maximizes the potential of a chipmaker’s design ultimately defines success. This pursuit of the “software unicorn” drives global automotive chipmakers to continuously seek software that can fully leverage their hardware.
According to Hsu, the true synergy between hardware and software doesn’t occur just by purchasing a chip and writing algorithms. It begins with a contract that grants the software firm access to a chipmaker’s resources, following a thorough evaluation of the software provider’s capabilities. AutoSys, for instance, collaborates with firms like NXP and Taiwanese chipmakers Sunplus and iCatch Technology, using processors that deliver more than expected—sometimes producing four functions on a chip originally designed for two. This creates additional value for Tier 1 suppliers and automakers, enhancing the cost-performance ratio.
Global competition is fierce, particularly in China’s automotive market, where firms specializing in 360-degree vision and advanced driver-assistance systems (ADAS) have undergone rigorous testing. Surviving firms are often top-tier suppliers or OEM-backed powerhouses, yet they may still grapple with scalability and overcommitment issues.
In the end, Hsu argues that real-world performance is the ultimate test. For 360-degree vision systems, success hinges on two crucial elements: flawless image stitching and effective blind-spot management. When software isn’t optimized, it can result in blurry images, “ghosting” effects, or visuals that unexpectedly disappear. To prevent these issues, algorithms need to adapt smoothly across different chip architectures, giving Tier 1 suppliers and automakers the flexibility to choose software that fits a range of vehicle models.
AutoSys itself illustrates this with a recent unexpected order from a US commercial vehicle client in 2024, underscoring how critical it is for chipmakers to choose compatible partners amid complex collaborations.
Some component and subsystem suppliers, aiming to add value, are also exploring the software space, leveraging their existing automotive channels but encountering high costs and industry barriers. Thus, strategic partnerships with chipmakers and software firms have become essential, enabling seamless, cost-effective upgrades and expanding the scope for software-driven advancements.