A vision chip runs on-device video analytics with edge AI and hardware security, reducing cloud use in CCTV systems. Here is what it brings.
Mindgrove Technologies plans to release its V2600 system-on-chip in late 2026 for CCTV and surveillance systems that require built-in security and on-device processing. The chip is designed to run video analytics at the edge, limiting the need to send data to the cloud. It targets surveillance deployments facing tighter regulatory requirements around data protection and system security. Commercial availability is expected by the end of 2026.
At the core of the V2600 is a single-core RV64-GCV RISC-V CPU running at 1.2 GHz. The SoC integrates compute, imaging, video processing, and hardware security blocks on one die. This setup enables real-time video analytics and local decision-making without transferring sensitive footage to external servers.
The chip supports H.264 and H.265 encoding and includes dual-channel MIPI CSI-2 interfaces for camera input. It can be used in multi-camera surveillance systems, biometric devices, and other connected vision products. Processing data locally helps reduce latency and limit data exposure.
The chip also supports a range of peripheral interfaces for integration into smart cameras, industrial systems, automotive platforms, and IoT devices. The architecture is built to handle vision workloads at the edge, where response time and data integrity matter.
Security functions are implemented at the hardware level. The SoC includes a cryptographic engine and one-time programmable memory for secure boot, key storage, and device authentication. By combining compute, vision processing, and embedded security, the chip supports end-to-end protection within a single platform.
V2600 is the first chip in the company’s MG Vision series, which will address applications such as secure CCTV, automotive systems, industrial automation, biometric access control, smart cities, smart TVs, and other connected devices. It follows the company’s earlier Secure IoT microcontroller and reflects its plan to build secure semiconductor products in India.
“By tightly integrating compute, vision processing, and hardware-level security, we enable developers to build high-performance vision applications without exposing sensitive data to the cloud,” said Shashwath TR, co-founder and CEO of Mindgrove Technologies. “This not only improves system responsiveness but also materially strengthens end-to-end security.”