NTT unveiled the world’s first low-power AI inference chip capable of processing 4K video in real time for autonomous vehicles and surveillance drones. NTT unveiled the world’s first low-power AI inference chip capable of processing 4K video in real time for autonomous vehicles and surveillance drones. NTT unveiled the world’s first low-power AI inference chip capable of processing 4K video in real time for autonomous vehicles and surveillance drones. NTT unveiled the world’s first low-power AI inference chip capable of processing 4K video in real time for autonomous vehicles and surveillance drones. Surveillance systems and autonomous vehicles are hitting a wall, literally and figuratively. This is especially true regarding real-time decision-making in power- and bandwidth-constrained environments. Today’s autonomous AI solutions are largely reliant on energy-intensive GPUs and cloud connectivity, which often leads to struggles in processing ultra-high-definition video at the edge. At the Upgrade 2025 conference, Japanese telecoms giant NTT unveiled a potential fix: the world’s first low-power AI inference chip capable of processing 4K video in real time on edge devices—bringing sharper vision, faster response, and responsible intelligence and navigation to drones and autonomous systems such as self-driving cars. NTT NTT NTT The chip is capable of real-time 4K video inference at just under 20 watts to deliver effective performance without the energy burden typical of GPU-based solutions. "With our new AI inference chip, we’re enabling autonomous systems to see farther, react faster, and operate independently. This is a critical step toward safer, more sustainable deployments of AI in the real world," Kazu Gomi, President and CEO of NTT Research, told me. "We are currently testing the chip on drones for surveillance applications, but in the future, it can also be integrated into autonomous vehicles, where real-time inference with low power consumption is critical. By running advanced object detection AI models like YOLOv3 directly on-device, the chip eliminates the need for video compression or cloud-based processing. This leads to a boost in both efficiency and operational independence. For instance, it can allow drones to identify people or infrastructure from Japan’s legal flight ceiling of 150 meters—five times higher than current solutions allow. “A high level of inference at the edge is a disruptive force, and potentially a true game changer for a range of industrial and commercial use cases,” Carlos Moreno Alvarez, senior cloud architect at cloud infrastructure platform Escala24x7, told me. “Its impact lies in enabling data that’s traditionally processed centrally to be handled at the edge instead, producing more actionable insights for sectors that rely on near-real-time analytics and manage massive volumes of data.” The Race for Smarter, Sustainable AI at the Edge The AI chip’s implications go far beyond technology demos and might change the way AI is deployed for on-field use cases, from smart surveillance to real-time inspections of bridges and pipelines in remote areas. In disaster zones, where seconds count and power sources are scarce, a drone equipped with NTT’s chip could help identify survivors, monitor environmental threats, or assess structural damage, without relying on a high-bandwidth connection or draining precious battery life. “Advances in edge technology, particularly in model distillation—are allowing smaller models to perform comparably to larger ones, making them viable to run almost anywhere, even on consumer-grade devices,” JD Raimondi, head of data science at IT services platform Making Sense, told me. “The ability for AI models to operate offline makes them especially valuable in scenarios where quick, local decisions are critical. While edge AI models aren’t yet reliable enough to fully automate high-stakes tasks, they can help respond swiftly to emerging situations.” NTT’s chip leverages an in-house AI inference engine to dramatically boost computational efficiency. Its key differentiators are proprietary machine learning techniques, such as interframe correlation and dynamic bit-precision control, which enhance computer vision and image processing while maintaining high levels of accuracy. Gomi further emphasized, saying, "AI infrastructure is growing so massive and energy-intensive that it risks becoming unsustainable. With our new AI inference chip, NTT Research is helping to close that gap. By enabling real-time 4K video processing directly at the edge, we are paving the way for faster, more efficient AI while easing the burden on both devices and cloud data centers." However, NTT is not alone in the race to deploy AI chips for surveillance and autonomous vehicle applications. Xpeng, a leading Chinese EV manufacturer, has developed its autonomous driving chip, the Turing chip, which is set for mass production in Q2 2025. Designed to support up to 30 billion AI model parameters, it integrates two specialized neural network processors and delivers around 20% greater efficiency compared to typical automotive chips. Xpeng plans to deploy the Turing chip across its vehicles, robotics, and future flying cars. Likewise, Nvidia remains a dominant force in the sector with its Thor chip, designed specifically for autonomous vehicles with 750 TOPS (trillions of operations per second). Turing chip Turing chip Thor chip Thor chip AI, Photonics and the Future of Networking NTT plans to commercialize the chip within fiscal year 2025 through its subsidiary, NTT Innovative Devices Corporation. But this chip is just one piece of a broader vision. It’s also being explored as part of NTT’s Innovative Optical and Wireless Network (IOWN) initiative, a next-gen data infrastructure that uses photonics to boost speed and energy efficiency. “Traditional GPUs are often idle nearly 50% of the time, with processors and memory bottlenecked by transmission delays,” Gomi noted. “Photonics can bridge that inefficiency. With analog-style optical circuits, we could eventually replace GPUs in some scenarios and ease the pressure on data centers.” He also underscored that the rise of powerful edge AI does not eliminate the need for human oversight. “Before AI or AGI can replace skilled, human-level decision-making, it must be trained and tested with extraordinary care,” he said. “Technology can only go so far. Human judgment remains essential, especially when safety is on the line.” As autonomous systems evolve, Gomi believes the conversation must expand beyond capabilities to include responsibility. Who is accountable when AI makes a mistake? How do we build trust in systems that operate beyond our immediate control? NTT’s approach to these challenges includes collaboration across disciplines, ranging from encryption and secure data sharing with NTT DATA to exploring photonics as a long-term solution to today's data center power and latency bottlenecks. With its AI chip, NTT isn’t just enabling smarter surveillance or safer driving, the company is showcasing that intelligence at the edge can be both powerful and principled.

The Great Office Debate: Are RTO Mandates Really About Collaboration or Control?

AI Agents in Finance: Game-Changer or a Risky Gamble?

Read My Stories

Too Long; Didn't Read

Get Free Unlimited IP Geo Data With IPinfo Lite

Edge AI is Getting Smarter: NTT’s Chip Powers 4K Video in Drones and Cars

About Author

Comments

TOPICS

THIS ARTICLE WAS FEATURED IN

Related Stories

Untitled Story

AI Agents in Finance: Game-Changer or a Risky Gamble?

Edge AI: the Innards, Use Cases, Benefits, and Barriers

Putting the Blinders on AI

This AI Model Gives Edge Devices Eyes on the Back of Their Heads

How 3D Object Detection Helps AI Avoid Obstacles

AI Agents in Finance: Game-Changer or a Risky Gamble?

Edge AI: the Innards, Use Cases, Benefits, and Barriers

Putting the Blinders on AI

This AI Model Gives Edge Devices Eyes on the Back of Their Heads

How 3D Object Detection Helps AI Avoid Obstacles

Light-Mode

Classic

Newspaper

Dark-Mode

Neon Noir

Minty

HN StartUps