FEV and Microsoft announced a collaboration on Thursday, July 9, 2026, to integrate NVIDIA GPU-accelerated compute and AI model microservices into vehicles, enabling multimodal voice, text, and gesture interactions without a permanent internet connection. The partnership focuses on using small language models (SLM), specifically Microsoft’s Phi-4-mini-instruct in Microsoft Foundry, powered by NVIDIA DRIVE AGX accelerated compute, to allow drivers to configure vehicle functions like the dashboard or individual profiles via voice commands. The system will also serve as a local backup for cloud-based large language models (LLMs), ensuring robustness and availability of AI-powered functions. Performing inference directly in the vehicle keeps central functions accessible with limited or no internet connection and reduces backend and infrastructure costs by supplementing or partially replacing cloud-based LLMs. Thomas Hülshorst, Group Vice President Intelligent Mobility and Software at FEV, stated the collaboration showcases how small, efficient language models can transform in-vehicle experiences. Boris Scholl, Vice President of Engineering at Microsoft, added that the companies are shaping the future of intelligent, voice-driven interfaces for automotive deployment.