NTT has introduced the world’s first AI inference large-scale integration (LSI) chip that enables real-time 4K video processing at the edge, specifically designed for low-power and power-constrained devices.
This innovative chip allows for real-time AI inference from ultra-high-definition video without the need for compression, ultimately enhancing efficiency and accuracy. It opens up possibilities for applications like beyond visual line-of-sight drone operations for infrastructure inspections, extending the detection range from 30m to 150m.
NTT’s LSI chip utilizes a proprietary AI inference engine, significantly reducing computational complexity and power consumption to below 20 watts.
Kazu Gomi, president and CEO of NTT Research, highlights the potential of combining low-power AI inferencing with ultra-high-definition video, stating, “NTT’s LSI chip, which we believe to be the first of its kind to achieve such results, represents a significant advancement in enabling AI inference at the edge and for power-constrained terminals.”
The plan is to make the chip commercially available by fiscal year 2025, expanding its applications and supported inference models.
This technology aligns with NTT’s IOWN initiative, which leverages high-speed, low-latency networks to tackle modern infrastructure challenges.
NTT is collaborating with NTT DATA to integrate the chip with Attribute-Based Encryption (ABE), ensuring secure and flexible data sharing.
Related
AI inference | AI/ML | chip | edge computing | NTT | NTT Data | video processing
