In a major leap in Edge AI processing, NTT Corporation announced a groundbreaking AI inference chip that can process real-time 4K video at 30 frames per second—using less than 20 watts of power. This new Large-scale Integration (LSI) chip is the world’s first method to implement such high-performance AI video inference in power-constrained environments, making it a breakthrough for edge computing applications.
Revealed during NTT Upgrade 2025 The chip in Summit in San Francisco is designed specifically for deployment at edge devices—the magnets are physically close to data sources such as drones, smart cameras and sensors. With traditional AI systems relying on cloud computing for inference, the chip brings powerful AI capabilities directly to the edge, greatly reducing latency and eliminating the need to transmit ultra-high definition videos to centralized cloud servers for analysis.
Edge Computing and Cloud Computing: Why It Is Important
In traditional cloud computing, data from devices such as drones or cameras are sent to data centers that are remote data centers (usually hundreds or thousands of miles away). While this approach provides almost unlimited computing power, latency is introduced due to data transmission, which is problematic for real-time applications such as autonomous navigation, security monitoring, and real-time decision making.
In contrast, the edge computing process performs data locally or nearby on or near the device itself. This reduces latency, preserves bandwidth, and enables real-time insights in limited or intermittent Internet connectivity environments. It also enhances privacy and data security by minimizing the need to transmit sensitive data over public networks.
NTT’s new AI chips fully include this edge-first philosophy, which is to conduct real-time 4K video analysis directly inside the device without relying on the cloud.
A new era of real-time AI for drones and devices
With this chip installed, the drone can detect 150 meters (492 feet) of people or objects, which is a legal height limit for Japanese drones. This is a huge improvement on traditional real-time AI systems, which are usually limited to a 30-meter range due to their low resolution or processing speed.
This advance makes many new use cases work:
-
Infrastructure inspection In hard to reach places
-
Disaster Response Areas with limited connectivity
-
Agricultural monitoring Across a vast field
-
Security and surveillance No constant cloud uplink
All of this is done with less than 20 watts of chips – dramatically less than the hundreds of watts required for a GPU-powered AI server, which is not practical for mobile or battery-powered systems.
Inside the chip: NTT’s proprietary AI reasoning engine
The performance of LSI depends on NTT’s custom AI inference engine that ensures high-speed, accurate results while minimizing power usage. Key innovations include:
-
InterFrame Relevance: By comparing sequential video frames, the chip can reduce redundant calculations, thereby improving efficiency.
-
Dynamic bit precise control: This technique can use fewer bits to adjust the numerical accuracy required on the fly to simplify the task and thus preserve energy without compromising accuracy.
-
Native Yolov3 execution:Chip supports direct execution You only watch V3 onceis one of the fastest real-time object detection algorithms in machine learning.
These combined capabilities allow chips to provide robust AI performance in environments previously considered power or bandwidth limitations for advanced inference.
Commercialization and vision pathways
NTT plans to commercialize chips in fiscal 2025 through its operating company NTT Innovative Devices Corporation.
Researchers are already exploring its integration with innovative optical and wireless networks (IOWN), the next generation infrastructure vision of NTT, aiming to overhaul the digital backbone of modern society. In IONN data-centric infrastructure (DCI), the chip will utilize an all-optical network for ultra-low latency, high-speed communication, supplementing the local processing power it brings to Edge devices.
NTT is also with NTT Data, Inc. Cooperate to combine chip functionality with attribute-based encryption (ABE) technology, which enables secure, fine-grained access control for sensitive data. Together, these technologies will support AI applications that require speed and security, such as healthcare, smart cities and autonomous systems.
The legacy of innovation and a vision for the future
This AI inferred chip is the latest demonstration of NTT’s mission, which is to enhance a sustainable, smart society through deep technological innovation. As a global leader with revenue of over $92 billion, 330,000 employees and $3.6 billion in annual R&D, NTT provides over 75% of the services to Fortune Global 100 companies and millions of consumers in 190 countries.
Whether it is drones that go beyond visual routes, cameras that detect events in real time without cloud dependencies, or protect data flows through attribute-based encryption, NTT’s new chip sets the stage for AI to be on the next boundary of Edge’s AI, that is, intelligence will encounter immediately.
