NTT Unveils Breakthrough AI Inference Chip for Real-Time 4K Video Processing at the Edge

In a big leap for Edge AI processing, NTT Corporation Has announced a groundbreaking AI-Inference chip that can process real-time 4K video with 30 frames per second with less than 20 watts. This new Large -scale integration (LSI) Chip is the first in the world to reach such powerful AI-Video Investment in power-restricted environments, making it a breakthrough for Edge Computing applications.

Unveiled during NTTs Upgrade 2025 meeting In San Francisco, the chip is specially designed for implementation in peripherals – hardware that is physically close to the data source, such as drones, smart cameras and sensors. In contrast to traditional AI systems that depend on cloud computing for inference, this chip brings powerful AI options directly to the edge, reducing the latency and the need to send Ultra-High-Definition video to centralized cloud servers for analysis.

Edge Computing vs. Cloud Computing: Why it matters

In traditional cloud computing, data from devices such as drones or cameras are sent to remote data centers – often hundreds or thousands of kilometers away – where it is processed and analyzed. Although this approach offers virtually unlimited computing power, it introduces delays as a result of data transfer, which is problematic for real -time applications such as autonomous navigation, security monitoring and live decision -making.

Edge Computing processes, on the other hand, data locally, on or near the device itself. This reduces latency, retains bandwidth and makes real -time insights possible, even in environments with limited or intermittent internet connectivity. It also improves privacy and data security by minimizing sensitive data on public networks.

The new AI chip from NTT embraces this EDGE-first philosophy full-time real-time 4K video analysis directly within the device, without trusting the cloud.

A new era for real -time AI on drones and devices

With this chip installed, a drone people or objects can detect a maximum of 150 meters (492 feet) – the statutory high -limit for drones in Japan. That is a dramatic improvement compared to traditional real-time AI systems, which are generally limited to a range of 30 meters due to a lower resolution or processing speed.

This progress makes a large number of new use cases possible, including:

• Infrastructure inspections in hard -to -reach places

• RAMP -Reaction In areas with limited connectivity

• Agricultural monitoring About wide fields

• Security and supervision Without constant cloud -uplinks

All this is achieved with a chip that consumes less than 20 watts dramatically lower than the hundreds of watts needed by GPU-driven AI servers, which are impractical for mobile or battery-driven systems.

In the chip: your own AI -insertion engine of NTT

The performance of the LSI depends on the tailor-made AI-insference engine of NTT, which ensures fast, accurate results, while the power is minimized. Main innovations are:

• Interframe correlation: By comparing sequential video frames, the chip reduces redundant calculations, improving efficiency.

• Dynamic BIT-PreCision control: This technique adjusts the numerical precision that is needed, with fewer bits for simpler tasks, the preservation of energy without accuracy.

• Native Yolov3 version: The chip supports direct implementation of You only watch V3 onceOne of the fastest real-time object detection algorithms in machine learning.

With these combined functions, the chip can deliver the AI ​​performance in environments that are previously considered to be limited or bandwidth for advanced inference.

Path to commercialization and the vision in the candidate

NTT is planning to commercialize the chip within the tax year 2025 through its work company, NTT Innovative Devices Corporation.

Researchers are already investigating their integration into the Innovative Optical and Wireless Network (Iown)—Tt’s next generation of infrastructure vision aimed at overhauling the digital backbone of modern society. Within Iown’s Data-centric infrastructure (DCI)The chip would benefit from the All-Photonics Network for Ultra-Lage latency, high speed communication, in addition to the local processing power it offers to peripherals.

NTT also works together with NTT Data, Inc. To combine the possibilities of the chip with its Attribute-based coding (ABE) Technology that makes safe, fine -grained access control possible about sensitive data. Together, these technologies will support AI applications that require both speed and security – such as in health care, smart cities and autonomous systems.

An inheritance of innovation and a vision for the future

This AI Inference Chip is the newest demonstration of NTT mission to enable a sustainable, intelligent society through deep technological innovation. As a world leader with more than $ 92 billion in income, 330,000 employees and $ 3.6 billion in annual R&D, NTT serves more than 75% of the Global 100 companies Fortune and millions of consumers in 190 countries.

Whether it is drones that fly past the visual facial line, cameras that detect events in real-time without cloud dependence, or protecting data flows with attribute-based coding, NTT’s new chip forms the scene for the next border in AI on the edge-where intelligence immediately meets immediately.