NPUs enable a wide range of edge AI applications, including multi-modal generative AI models and deliver more secure, cost-effective and energy efficient AI processing at the edge.
Advanced AI at the edge enables critical decisions to be made locally and independently from the cloud, ensuring faster responses, improved data privacy, and reduced costs.
Kinara’s NPUs and comprehensive software are able to deliver energy-efficient AI performance across a range of neural networks, including conventional AI, as well as generative AI, to address the rapidly growing AI needs of industrial and automotive markets.
The acquisition will enable NXP to provide complete and scalable AI platforms, from TinyML to generative AI, by bringing discrete NPUs and robust AI software to NXP’s portfolio of processors, connectivity, security, and advanced analogue solutions.
Already partners, the acquisition will help NXP to make it easier to pair Kinara’s NPUs with NXP’s portfolio of industrial and IoT processors making it possible to deliver more scalable AI platforms for a variety of industrial and automotive AI inference needs.
“The industrial market is going through a transformation, with new innovations like generative AI helping to deliver major improvements in efficiency, sustainability, safety and predictability, and in many instances, unlock new use cases and functionality,” said Rafael Sotomayor, executive vice president and general manager, Secure Connected Edge at NXP. “Adding Kinara’s AI capabilities to our broad intelligent edge portfolio creates a scalable platform for new classes of AI-powered systems. Together, we can help our customers simplify complexity and accelerate time to market as they create transformative AI systems.”
Kinara’s discrete NPUs, including the Ara-1 and Ara-2, are being used for emerging AI applications in vision, voice, gesture, and a variety of other generative AI-powered multi-modal implementations.
Both devices feature an architecture that enables mapping of the inference graphs for efficient execution on Kinara’s programmable proprietary neural processing units for maximising edge AI performance. This programmability ensures adaptability as AI algorithms continue to evolve from CNNs to generative AI and new approaches such as agentic AI in the future.
Ara-1 is a first generation discrete NPU, capable of advanced AI inferencing at the edge. Ara-2, capable of up to 40 TOPS (Tera Operations Per Second), is the company’s second generation NPU and is optimised for achieving system-level high performance for generative AI.
The Ara-1 and Ara-2 NPUs can be integrated with embedded systems to enhance their AI capabilities, including upgrading existing in-field systems.
Kinara also provides a complete software development kit enabling customers to optimise AI model performance and streamline the deployment. Its AI software portfolio includes extensive model libraries and model optimisation tools, which will be integrated into NXP’s eIQ AI/ML software development environment to enable customers to create end-to-end AI systems.
