Ambarella Expands N1 Edge GenAI Family With SoC Targeted at On-Premise Multi-Channel VLM and NN Processing in Under 20 Watts

Product Announcement - Ambarella announced the N1-655 edge GenAI system-on-chip (SoC) at CES, capable of decoding 12x simultaneous 1080p30 video streams while processing video and running multiple multimodal vision-language models (VLMs) and traditional CNNs [1] - The N1-655 supports popular multimodal VLMs and large-language models (LLMs) with high AI processing performance, consuming only 20 watts of power, which is 10-100x lower than cloud processors [1] - The SoC can reliably run models like Phi, Gemma, LLaVA-OneVision, and Llama without an internet connection, targeting applications such as on-premise AI boxes, autonomous mobile robots (AMRs), and smart-city security video recorders [1] Product Family and Strategy - Ambarella is building a family of edge GenAI SoCs, with the N1-655 being the latest addition, designed to process GenAI models locally to improve privacy and reduce power consumption compared to cloud-based inference processing [2] - The company is leveraging cutting-edge fine-tuning techniques like QLoRA to reduce the footprint sizes of popular GenAI models, enabling their use in edge applications [3] Market Trends - There is a growing trend towards edge GenAI processing due to the high bandwidth costs, ongoing service fees, and power usage associated with cloud-based processing, as well as data security and privacy concerns [3] Developer Support - Ambarella is expanding its Cooper™ Developer Platform with the new Cooper Pro kit, powered by the N1-655, which provides industrial-class performance and will begin shipping after CES [4] Availability and Demonstrations - N1-655 samples are available upon request, and the SoC along with the Cooper Pro kit are being demonstrated at Ambarella's invitation-only exhibition during CES [5][6] Company Overview - Ambarella's products are used in various human vision and edge AI applications, including video security, ADAS, electronic mirrors, drive recorders, driver/cabin monitoring, autonomous driving, and robotics [7] - The company's low-power SoCs offer high-resolution video compression, advanced image and radar processing, and powerful deep neural network processing for intelligent perception, fusion, and planning [7]