Chip Design Firm Achieves Key Automotive Standard for Neural Processing Unit IP, Setting Stage for Safer Edge AI Applications
VeriSilicon, a prominent player in platform based, comprehensive, and integrated circuit design services and semiconductor IP licensing, has significantly advanced its position in the automotive and industrial edge computing markets. The Shanghai based company announced the successful attainment of ISO 26262 ASIL B certification for its VIP9000NanoOi-FS Neural Processing Unit (NPU) Intellectual Property (IP). This certification, issued by SGS-TÜV Saar, an internationally recognized testing, inspection, and certification body, is a crucial endorsement of the IP’s functional safety capabilities, particularly for mid level electronic components in vehicles.
The move underscores a wider industry trend where the integration of Artificial Intelligence into safety critical systems, such as Advanced Driver Assistance Systems (ADAS) and autonomous vehicles, necessitates rigorous compliance with global functional safety standards. For a company like VeriSilicon, which has established a strong foundation in semiconductor IP with its technology adopted in over 140 chip designs by 91 customers, securing this certification for its NPU portfolio is a strategic imperative that solidifies its role as a key enabler for next generation smart systems.
The Strategic Imperative of Automotive Safety Standards
The ISO 26262 standard defines functional safety for electrical and electronic systems in road vehicles. Achieving Automotive Safety Integrity Level B (ASIL B) means the VIP9000NanoOi-FS IP meets a demanding level of risk reduction for potential system failures. This is not merely a technical achievement but a commercial necessity for any component destined for the increasingly sophisticated electronic architectures of modern cars. Given the explosive growth of edge AI, especially in applications that involve real time decision making like object and scene detection, the ability to guarantee determinism and safety is paramount.
VeriSilicon’s history of innovation in silicon design positions this new certification as a natural progression. The company’s focus on heterogeneous computing and its extensive library of silicon IP have been central to its operations. This newest certification signals the firm’s commitment to expanding its Functional Safety (FuSa) portfolio to meet the stringent demands of the automotive industry, which relies on suppliers to de risk their complex System on Chip (SoC) designs.
Engineering Excellence for Edge AI Acceleration
The VIP9000NanoOi-FS IP, part of VeriSilicon’s VIP9000 series, is engineered specifically for high performance, low power inference acceleration in demanding environments like the vehicle cabin or industrial robotics. Its architecture is optimized for seamless SoC integration, balancing high quality AI inference with minimal power consumption and silicon footprint. This is a critical engineering challenge, as AI processing demands significant computational power, which often conflicts with the size and power constraints of edge devices.
The NPU IP’s capability to natively support a wide array of AI models, from convolutional neural networks (CNNs) to increasingly complex large language models (LLMs), addresses the diverse needs of next generation applications. Its hardware accelerated AI inference is vital for enabling real time functions such as augmented reality overlays, sophisticated visual inspection systems, and low latency video surveillance. Furthermore, the IP’s design incorporates a programmable SIMD vision pipeline and supports multiple data types, ensuring high performance while maintaining the deterministically low latency required to satisfy ASIL B requirements. This architectural focus on determinism is what differentiates safety certified IP from general purpose accelerators.
The Software Ecosystem as a Competitive Differentiator
In the complex world of semiconductor IP, the accompanying software ecosystem often dictates the speed and ease of customer adoption. VeriSilicon has provided comprehensive software stack and development tools that are fully compatible with industry standards. The VIP9000NanoOi-FS offers hardware support for the OpenVX API and is compatible with major frameworks, including OpenCL 3.0, OpenCL 1.2 Full Profile, OpenVX 1.3, and OpenVX 1.2 Neural Network Extension, along with various OCL, OVX, and neural network extensions. This breadth of compatibility is crucial, as it allows developers to leverage existing expertise and tooling, significantly reducing the time to market for safety critical AI solutions.
The confidence in this technology is evidenced by its existing market penetration. With nearly 200 million units of VeriSilicon’s AI IP already shipped globally, the company’s assertion of proven reliability and performance carries considerable weight. Weijin Dai, VeriSilicon’s executive vice president and general manager of the IP division, noted that the certification reinforces the company’s pledge to functional safety in NPUs. As AI becomes ubiquitous in autonomous driving and intelligent edge applications, providing safety compliant IP is essential for industry progress, allowing customers to confidently deploy AI solutions in safety critical environments. This deep commitment to safety and reliability is a core tenet of VeriSilicon’s strategy to service the high growth, high value automotive semiconductor market. The broader strategic move is to ensure that as AI accelerates the complexity of vehicular systems, the underlying hardware foundations remain unassailably secure and reliable. For more information on the company’s offerings, visit www.verisilicon.com.
