Ambarella, Inc. (NASDAQ: AMBA), a leader in edge AI semiconductors, announced the launch of its CV7 edge AI vision SoC at CES 2026, targeting applications across consumer, enterprise, automotive, and industrial markets. The CV7 supports AI-based 8K cameras, multi-imager security systems, robotics (including drones), industrial automation, and video conferencing devices.
The SoC is particularly suited for multi-stream automotive designs, supporting CNNs and transformer-based networks at the edge for AI vision gateways, fleet telematics, surround-view systems, and ADAS. It can process multiple video streams simultaneously—up to 8Kp60—while maintaining exceptional image quality and low power consumption.
Performance Upgrades and Efficiency Gains
Built on Samsung’s 4nm process technology, the CV7 consumes 20% less power than its predecessor, continuing Ambarella’s emphasis on high AI performance per watt. Its CVflow® third-generation AI accelerator delivers more than 2.5× the AI performance of the previous CV5 SoC, efficiently running multiple CNNs and transformer networks in tandem.
Ambarella’s algorithm-first architecture integrates its AI accelerator, image signal processor (ISP), video encoding, Arm cores, and I/Os into a single SoC. Compared with multi-chip solutions, this integration improves time-to-market, reduces BOM costs, and enables smaller, thermally efficient form factors.
Industry-Leading Imaging and Video Capabilities
The CV7 continues Ambarella’s tradition of high-performance image signal processing, including:
- High dynamic range (HDR) with improved contrast in bright/dark scenes
- 3D motion-compensated temporal filtering (MCTF)
- Dewarping for fisheye cameras
- Low-light performance down to 0.01 Lux
Hardware-accelerated video encoding supports H.264, H.265, and MJPEG, doubling encode performance over the CV5. This enables a single 4Kp240 stream, dual 8Kp30, or more than 4× 4Kp30 streams concurrently for enterprise multi-camera deployments.
Additional upgrades include a quad-core Arm Cortex-A73 CPU, doubling general-purpose processing performance, and a 64-bit DRAM interface for significantly higher memory bandwidth. These improvements make CV7 a robust platform for next-generation AI vision and vision-language models at the edge.
CEO Fermi Wang highlighted the SoC’s versatility: “The CV7 enables developers to deliver the most advanced imaging features and highest edge AI performance for consumer, enterprise, and industrial AIoT applications, all while reducing power consumption and thermal constraints.”
Power Tomorrow’s Intelligence — Build It with TechEdgeAI
