NTT Corporation has unveiled a new low-power large-scale integration (LSI) chip capable of real-time AI inference processing on ultra-high-definition (4K, 30 fps) video for edge and power-constrained devices. The LSI incorporates an NTT-developed AI inference engine that significantly reduces computational load while maintaining detection accuracy, allowing for advanced use cases such as beyond visual line of sight (BVLOS) drone operations for public infrastructure inspection. The announcement was made during Upgrade 2025, NTT’s annual research summit in San Francisco.
The LSI enables object detection from drones flying up to 150 meters (492 feet), the maximum legal drone altitude in Japan—five times the operational range of conventional AI inference systems. Operating at less than 20 watts of power, the chip executes advanced algorithms like YOLOv3 using techniques such as interframe correlation and dynamic bit-precision control. This level of processing power makes it suitable for a range of edge applications including public safety, industrial monitoring, and live events where traditional server-class GPUs are impractical due to size, heat, and power requirements.
NTT plans to commercialize the LSI through its subsidiary, NTT Innovative Devices Corporation, within fiscal year 2025. The company is also exploring integration of the LSI into its IOWN (Innovative Optical and Wireless Network) architecture, particularly within the data-centric infrastructure (DCI) layer. In addition, NTT is collaborating with NTT DATA on merging the chip with Attribute-Based Encryption (ABE) technologies to enable secure, fine-grained access control for edge AI systems.
• NTT introduces AI inference LSI for 4K video processing at 30 fps under 20 watts
• Enables BVLOS drone operations and edge deployments with limited power budgets
• Executes YOLOv3 object detection algorithm with high efficiency and accuracy
• Uses NTT’s proprietary inference engine featuring interframe correlation and dynamic bit-precision
• Commercial availability expected within fiscal year 2025 via NTT Innovative Devices Corporation
• LSI supports future applications in IOWN’s data-centric infrastructure and integrates with ABE security technologies
“The combination of low-power AI inferencing with ultra-high-definition video holds an enormous amount of potential, from infrastructure inspection to public safety to live sporting events,” said Kazu Gomi, president and CEO of NTT Research.
