SoundHound AI, Inc. (NASDAQ: SOUN) announced its participation in NVIDIA GTC 2026, where the company will demonstrate what it describes as the first multimodal, multilingual AI platform operating entirely on edge devices at Booth #1844.

The platform runs the complete AI processing stack locally within vehicle architecture, eliminating the need for cloud connectivity that typically supports complex AI requests. The system combines voice, vision, and reasoning capabilities that operate offline while maintaining what the company states is 100% uptime and enhanced privacy.

At the conference, SoundHound will showcase three main technologies. The Agentic+ platform supports multiple protocols including MCP and A2A, allowing various types of AI agents to function together locally. The company will demonstrate vehicle control, navigation, and information services without remote server access.

A collaboration with NVIDIA utilizes the DRIVE AGX Orin platform to power the edge-based voice AI system. The demonstration will show generative AI responses functioning in areas without network connectivity.

The company will also present Vision AI capabilities integrated into conversational interfaces, enabling vehicles to process visual information such as landmark identification and gesture recognition while operating offline.

"For the very first time, our multimodal, multilingual agentic AI platform allows OEMs to turn on intelligent agents that can see, hear, and act – while maintaining user privacy along with the speed of edge computing," said Keyvan Mohajer, CEO and Co-Founder of SoundHound AI.

SoundHound AI provides voice and conversational AI solutions across retail, financial services, healthcare, automotive, telecommunications, smart devices, and restaurant industries. The company's products include Smart Answering, Smart Ordering, Dynamic Drive-Thru, and the Amelia Platform for enterprise AI agents.