NTT’s Radical Vision: Slashing AI Power Consumption with Optical Technology

The Future is Light: A Game-Changer for AI Infrastructure

At the GDS2025 World Digital Summit, Akira Shimada, President of NTT (Nippon Telegraph and Telephone Corporation), unveiled a vision that could redefine the infrastructure of artificial intelligence (AI) globally. As the world races toward AI dominance, a new bottleneck emerges—not computing power, but energy. Shimada’s answer? A revolutionary shift from traditional electronic communication to optical (light-based) computing, which promises unprecedented energy efficiency and speed.

NTT’s ambitious move centers around a futuristic communications framework known as IOWN (Innovative Optical and Wireless Network). This architecture leverages photonics—communication through light—to drastically cut energy use, allowing data centers and AI systems to operate at a fraction of today’s energy demands. The implications go far beyond just cost savings; this innovation could be the key to scaling AI sustainably.

Original

During a keynote at GDS2025, NTT President Akira Shimada emphasized that optical technology is superior to electricity for high-volume AI computing. He announced that NTT has developed computers that use just one-eighth the power of conventional models by leveraging light-based transmission. The breakthrough is part of the broader IOWN initiative, NTT’s next-generation communication infrastructure, designed to support the massive data and energy requirements of AI.

Shimada pointed out that light transmission maintains low power consumption even over long distances—a vital advantage as AI networks grow increasingly distributed. A central piece of this transformation is “photonic-electronic convergence”, which expands light usage from long-distance networking into device-level and intra-device communication.

Currently, a prototype of this energy-efficient computer is already running at NTT’s pavilion at the Osaka–Kansai Expo. More advanced models, with increased capacity, are expected to be released by 2026.

Looking ahead, NTT is developing miniature optical devices to directly connect CPUs and GPUs using light, eliminating unnecessary components and dynamically shutting off unused modules to conserve energy. Shimada set an ambitious target: by 2032, even on-chip communication will be entirely optical, potentially reducing power usage to 1/100th of current levels.

NTT, celebrating the 40th anniversary of its privatization, is rebranding itself—from a telecom giant to a pioneer of next-gen computing. With its deep expertise in optical communications, the company aims to lead the AI era by building its physical foundation.

What Undercode Say:

The message from NTT is bold, but also well-timed. As AI models grow exponentially in both size and deployment scale, power consumption has become a serious bottleneck. Traditional electronic architectures, no matter how advanced, cannot keep pace with the demand for energy-efficient computing. NTT’s pivot to photonic computing is not just clever—it might be necessary.

What makes this especially promising is IOWN’s modularity. The vision goes beyond raw speed; it proposes an intelligent, dynamic system that adapts energy usage in real-time. Only the required components are powered, while the rest are deactivated—almost like the brain’s energy economy. That’s not just innovation; it’s survival engineering for the AI age.

Another layer to consider is geopolitical. Japan, through NTT, is attempting to stake its claim in AI infrastructure leadership, an area currently dominated by American and Chinese companies. This isn’t just about technology—it’s about strategic positioning in the global AI race.

The fact that NTT’s prototype is already operational at Expo 2025 offers more than just a PR win—it shows tangible proof of concept. If the company can follow through on its roadmap and achieve 100x energy savings by 2032, it could redefine how data centers, edge computing, and AI factories operate worldwide.

One challenge, however, lies in integration. Shifting from an electronic to a photonic paradigm is not a plug-and-play process. It demands a complete ecosystem overhaul—hardware, software, supply chains, and developer retraining. But if any company is positioned to do it, it’s NTT, with decades of optical networking expertise under its belt.

It’s also worth noting that energy savings have global environmental benefits. As AI becomes the backbone of finance, healthcare, logistics, and security, the carbon footprint of data centers threatens to rival that of entire countries. Technologies like IOWN could shift that trajectory.

Finally, NTT’s name change is symbolic. Shedding its long-form legacy title (“Nippon Telegraph and Telephone”) in favor of the streamlined “NTT” suggests more than a rebranding—it marks a generational transformation from legacy telecom to digital infrastructure powerhouse.

🔍 Fact Checker Results

✅ Power Reduction Claims: Verified. NTT has demonstrated prototype systems reducing energy use to 1/8th through optical computing.
✅ IOWN Roadmap: Confirmed. Public plans exist to scale by 2026 and achieve full on-chip photonics by 2032.

❌ Global Market Share Lead: Not yet a reality.

📊 Prediction

By 2028, NTT’s IOWN-powered systems will begin replacing legacy AI server clusters in Japan and select global markets.
By 2030, power-saving optical computing will become a standard for AI workloads in both private data centers and national infrastructure.
By 2032, NTT will be seen as a foundational AI infrastructure provider, collaborating with international chipmakers to deploy fully photonic chips in commercial-grade AI systems.