Visteon and Nvidia build AI arbitration layer for SDVs

Visteon has announced an edge-to-cloud AI arbitration architecture for software-defined vehicles, developed in collaboration with Nvidia. The platform dynamically distributes AI workloads between on-device hardware and cloud infrastructure, managing latency, privacy and connectivity requirements in real time.

The architecture builds on the AI-ADAS Compute Module unveiled at CES 2026, which is powered by Nvidia DRIVE AGX Orin and DriveOS. The latest iteration integrates Nvidia Nemotron open models and Nvidia NIM microservices, extending the platform with generative AI capabilities across speech, vision, retrieval-augmented generation (RAG) and large language model functions. Latency-sensitive functions execute at the vehicle edge, while larger reasoning models for fleet learning and predictive maintenance run in the cloud.

Visteon delivers the system as a fully integrated solution combining AI model training and optimisation with a complete automotive-grade AI operating system (AIOS) stack. Automakers retain control over data residency, model configuration and regulatory compliance strategy, while Visteon manages the integration layer including model fine-tuning, workload orchestration and lifecycle management across edge and cloud environments.

In automotive applications, the architecture is designed to power agent-based in-cabin assistants with contextual awareness, safety monitoring and personalised user experiences.

In a statement, Visteon said the platform is engineered for global scale and will be applied across its product portfolio to expand flexible compute strategies in AI-enabled software-defined mobility.

Source: Visteon