Could NTT’s Low-Power Chip Be a Game Changer for Drones?

A new AI chip from NTT Research might soon help drones think smarter without guzzling power like a college student downing energy drinks. EETimes reported on it and showed us that the company says its prototype can perform real-time 4K object detection while using under 20 watts of power, a feat that could transform how aerial robots see and react to the world.

4K Detection Without the Trade-Off

Until now, AI vision systems on drones have always faced a painful choice: high accuracy or long battery life. Most settle for less detail, shrinking 4K video frames down to around 600×600 pixels before running object detection models like YOLO, short for “You Only Look Once.”

NTT decided to skip the compromise. Instead of squishing images, the chip splits each 4K frame into smaller 600×600 blocks, runs detection on each in parallel, and then stitches the results together. The full resolution stays intact, yet it still runs smoothly at 30 frames per second.

CEO Kazu Gomi explained, “We can maintain high-resolution accuracy and keep power under 20 watts because we reuse motion vector information.”

In other words, it remembers what’s moving and what’s not, just like an experienced pilot who doesn’t panic at every passing bird (just like me when I started flying).

A Boost for Drone Vision

For drones, this could be a huge leap forward. Real-time, uncompressed 4K processing means spotting people, vehicles, or obstacles from greater distances and identifying them more precisely.

Even better, the chip lets drones process data onboard rather than sending massive video streams to the ground. That means less lag, fewer dropped frames, and less radio traffic clogging up the skies. In swarm operations where dozens of drones need to cooperate, that could make all the difference.

Ian O’Connor, professor at École Centrale de Lyon, noted that real-time 4K inference on the edge is still a rarity. “That translates into more robust detection, especially for small or distant objects,” he said. He also praised NTT’s use of dynamic precision control, calling it “a technical advantage worth watching.”

Competing in a Crowded Field

NTT’s not the only one chasing smarter edge AI. Qualcomm’s Snapdragon Flight RB5 and Nvidia’s Jetson line already handle onboard intelligence, but they usually process compressed video or consume more power. If NTT can really pull off uncompressed 4K inference under 20 watts, it could outfly the competition.

The chip still lacks a name, but commercial release is planned for early 2026. NTT is already inviting drone manufacturers to collaborate, helping to adapt the design for rugged environments and manage the thermal load.

To support developers, an SDK will be released to let integrators tune inference models or go beyond the built-in YOLO algorithms. In short, it’s meant to be both powerful and flexible, like a Swiss Army knife for onboard AI.

Looking Ahead

NTT’s ambitions go far beyond this chip. The company is also researching biodegradable electronics and optical processors that could one day replace power-hungry GPUs. Gomi even hinted that within five or six years, optical chips could enable drones to run generative AI. “It may be possible to mount a GenAI platform on a flying object using our optical chips,” he said.

It’s a bold claim, but then again, NTT has a history of surprising the world. Back in 1999, it launched i-mode, bringing the internet to mobile phones long before smartphones were a thing. Now it hopes to repeat history by leading the edge AI revolution in drones and robotics.

If the chip works as advertised, it could become a standard in both civilian and defense UAVs, helping drones think faster and see farther without constantly begging for more battery life.

DroneXL’s Take

If NTT’s engineers are right, this chip could finally give drones the brains they deserve. Imagine a small quadcopter that can identify a person, car, or tree in full 4K detail, all while sipping just 20 watts.

Of course, talk is cheap until it flies. But if the results match the hype, NTT might just prove that less power can actually mean more intelligence. After all, the smartest drones might soon be the ones that know when to stop wasting our valuable energy.

Photo credit: NTT Research