Intel introduces next-gen AI-enhanced SDV SoC with multi-node chiplet architecture

At Auto Shanghai 2025, Intel introduced its second-generation software-defined vehicle (SDV) system-on-chip (SoC), featuring the automotive industry’s first multi-node chiplet architecture. Designed for intelligent, connected vehicles, the SoC supports customizable compute, graphics, and AI capabilities. Intel also announced new collaborations with ModelBest and Black Sesame Technologies.
The SDV SoC uses a multi-node chiplet architecture, integrating silicon for each functional block. It delivers up to 10x AI performance for generative and multimodal AI and up to 3x graphics performance for enhanced human-machine interfaces. The SoC supports 12 camera lanes for expanded camera input and image processing.
Intel’s collaboration with ModelBest integrates ModelBest’s GUI Intelligent Agent with the SDV SoC and Intel Arc graphics, enabling an on-device large language model for offline, AI-enhanced voice control and personalized interactions. The agent processes natural language in complex scenarios, leveraging ModelBest’s experience with Intel’s AI PC acceleration program.
With Black Sesame Technologies, Intel is developing a central compute platform combining Black Sesame’s advanced driver-assistance systems (ADAS) with the SDV SoC and Intel Arc graphics. This platform merges ADAS and immersive cockpit experiences into an energy-efficient system with stable, high-speed, low-latency connections.
These developments highlight Intel’s focus on providing scalable compute power for energy-efficient, software-defined vehicles, supporting the automotive industry’s shift toward intelligent mobility.
For more details, visit https://www.intel.com.