Socionext has secured a strategic licensing deal for aiMotive’s aiWare neural network accelerator IP, aiming to fast-track the design of energy-efficient, high-performance custom SoCs for next-gen automotive ADAS.

This partnership highlights a shared focus on cutting-edge chiplet architectures and integrated AI solutions tailored for increasingly complex vehicle safety systems.

Socionext has recently entered into a strategic licensing deal with aiMotive, giving itself access to aiMotive’s cutting-edge aiWare IP and development tools. The goal here? To speed up the process of designing and tape-outing next-gen customized SoCs for automotive advanced driver-assistance systems (ADAS). This partnership really emphasizes work around custom chiplet tech structures, showing that both companies share a common drive to push the frontiers of automotive AI hardware.

As Socionext’s announcement points out, integrating aiMotive’s aiWare neural network accelerator—renowned for its low power draw and high efficiency—makes it possible to deliver very optimized semiconductor solutions. These are tailored to meet the increasingly complicated demands of ADAS applications. Márton Fehér, aiMotive’s Senior Vice President of Semiconductor Engineering, called this partnership a game-changer for aiWare’s prospects. He highlighted Socionext’s deep experience in automotive SoC design and its solid industry relationships as major factors that will help scale up their AI acceleration tech in both power and production volume.

Now, Socionext isn’t a newcomer to this field—far from it. They’ve built a solid track record of providing custom SoCs across various markets, especially focusing on energy-efficient, high-performance processors aimed at automotive use. Takehiro Kamada, their Executive Vice President, stressed that this partnership broadens what Socionext can offer in terms of custom SoCs and helps meet the rising market demand for solutions that combine smart AI accelerators with rich software development kits (SDKs) and supporting toolchains. That kind of all-in-one support? Well, it’s becoming increasingly important for automotive clients who seek smooth integration of AI technology into their systems.

Speaking of aiWare, the core IP developed by aiMotive is specifically crafted for accelerating neural networks in cars. It can handle a wide range of AI inference tasks—like CNNs, LSTMs, RNNs, and Transformer-based models—and can reach up to 256 Effective TOPS per core with an impressive efficiency, reportedly hitting up to 98%. Its clever dataflow architecture minimizes the need for external memory bandwidth by processing as much data as possible on-chip, which not only conserves power but also boosts performance—key factors when designing automotive safety systems that need to be both reliable and efficient.

Plus, the latest version, aiWare4+, offers new features like wavefront processing, better programmability, and compliance with safety standards like ISO 26262 (ASIL B). It’s also designed with future trends in mind—supporting formats like FP8, enabling more granular structured sparsity, and scalable multitasking. That kind of versatility means it can serve a spectrum of automotive AI needs, from edge processors embedded in sensors to powerful central units managing complex autonomous features.

This licensing agreement and the resulting development collaboration fit neatly into aiMotive’s broader strategy. Beyond aiWare, they also supply aiSim, a virtual validation suite that supports the entire conceptual-to-production journey, and aiData, a sophisticated data pipeline that simplifies rolling out automated driving features. These tools are all about shaving down time to market and cutting development costs for automakers and Tier 1 suppliers working on advanced ADAS and autonomous driving systems.

All in all, the partnership between Socionext and aiMotive feels like a significant step forward in the automotive semiconductor scene. It’s promising to see how such collaborations can accelerate innovation—addressing the complex needs of modern ADAS and AI-powered driving assistance. As automotive systems become smarter and more automated, these kinds of alliances—blending specialized IP with custom SoC development expertise—are pretty much essential for making roads safer and vehicles more intelligent, don’t you think?


Note: References are available for further reading but are not directly included here to keep the flow natural.

Source: Noah Wire Services