Cerence Shares Jump 30% After NVIDIA Expands Partnership on Automotive AI Capabilities

Cerence, a pioneer in voice AI and enhanced AI-powered experiences for the automotive industry, has announced an expanded collaboration with NVIDIA to advance the capabilities of its CaLLM™ family of language models.

The shares jumped 30% premarket after the partnership announcement, while declining 60% in 2024.

The collaboration leverages NVIDIA's AI Enterprise software platform, including TensorRT-LLM and the NeMo framework, to optimize performance for both cloud-based and embedded automotive language models. Specifically, Cerence's cloud-based Cerence Automotive Large Language Model (CaLLM) and its CaLLM Edge embedded small language model are powered by NVIDIA AI Enterprise, an end-to-end, cloud-native software platform. Additionally, aspects of CaLLM Edge are powered by NVIDIA DRIVE AGX Orin.

Integrating agentic frameworks with in-car conversations in both cloud and embedded forms requires a comprehensive, cross-disciplinary effort combining hardware, software, and UX domain expertise. Working alongside NVIDIA hardware and software engineers, Cerence AI has enhanced its ability to meet production timelines and productize generative AI innovation for automotive applications. This collaboration aims to improve in-vehicle assistant performance through NVIDIA's accelerated computing and SoCs, implement automotive-optimized NVIDIA NeMo Guardrails, and optimize agentic architecture via NVIDIA DRIVE AGX Orin.

The integration focuses on delivering faster in-vehicle assistant performance, developing automotive-optimized implementations of NVIDIA NeMo Guardrails to navigate the nuances of in-car interaction, and implementing and optimizing an agentic architecture on CaLLM Edge. This partnership enables Cerence to deliver enhanced driver experiences with advanced performance, reduced latency, improved privacy, and robust security measures.

By optimizing the performance of our CaLLM family of language models, we are delivering cost savings and improved performance to our automaker customers, who are running quickly to deploy generative AI-powered solutions to their drivers, said Nils Schanz, Executive Vice President, Product & Technology, Cerence AI. As we advance our next-gen platform, with CaLLM as its foundation, these advanced capabilities will deliver faster, more reliable interaction to drivers, enhancing their safety, enjoyment, and productivity on the road.

Large language models are offering vast, new user experiences, but complexities in size and deployment can make it difficult for developers to get AI-powered solutions into the hands of end users, said Rishi Dhall, Vice President of Automotive, NVIDIA. Through this expanded collaboration, Cerence AI is deploying advanced NVIDIA AI and accelerated computing technologies to optimize its LLM development and deployment.