The agreement extends Ceva's customer base beyond traditional semiconductor companies and device OEMs to include software platform companies that are increasingly designing custom silicon to optimize performance, power and area (PPA) and the overall user experience.

Leading technology platform companies increasingly recognize that custom AI silicon is essential to optimize performance, power efficiency and full-stack control at scale. For companies that own both the operating system and the hardware platform, co-designing silicon and software creates a decisive advantage: tighter OS-to-silicon optimization enables greater performance and power efficiency that off-the-shelf processors cannot deliver, particularly in portable edge computing devices where thermal and battery constraints are unforgiving. Just as CPUs defined general-purpose computing and GPUs accelerated graphics and parallel workloads, AI acceleration is emerging as a third foundational layer of the computing stack, driving a new generation of custom inference silicon and positioning NPUs as a core architectural element of future intelligent computing platforms.

The customer selected NeuPro-M to provide scalable, power-efficient AI acceleration for advanced on-device inference workloads. The architecture enables efficient execution of generative AI, multimodal AI, emerging agentic AI workloads and other machine learning applications while operating within the power, area and thermal constraints of intelligent edge computing devices. NeuPro-M enables customers to integrate advanced AI capabilities directly into custom silicon architectures, providing the flexibility to co-optimize performance, power efficiency and user experience across the full hardware and software stack. As part of the program, Ceva collaborated closely with the customer to implement advanced neural network optimizations tailored to its target AI workloads, further improving inference efficiency and performance.