- NTT’s new chip performs real-time AI inference on 4K video using minimal power.
- It is designed for use in smart cameras, remote monitoring, and other edge computing applications.
What happened: NTT develops AI chip for real-time edge video processing
NTT has developed a new AI inference chip that processes 4K video in real time on low-power edge devices. The chip can handle 30 frames per second, making it suitable for smart cameras, drones, and other mobile or remote systems.
The chip uses NTT’s custom AI engine to reduce computation load while preserving detection accuracy. It also adjusts bit precision dynamically and processes video frame-by-frame, helping to cut power consumption without losing performance.
NTT plans to release the chip commercially in 2025 through its subsidiary, NTT Innovative Devices Corporation.
Also read: NTT Ltd.: Global tech solutions for digital transformation
Also read: Nokia, NTT, and Anritsu launch PoC to cut energy use in networks
Why it’s important
As more AI tasks shift to edge devices, there’s growing demand for energy-efficient hardware that can run advanced models outside traditional data centres. NTT’s chip supports that shift, enabling real-time decision-making in surveillance, public safety, and industrial automation — even in locations with limited power or bandwidth.
Its potential use in NTT’s broader networking initiatives could also improve speed and reduce energy use across future communications infrastructure.
