ITRI Launches 3.2T Optical Engine to Boost AI Data Speed
New silicon photonics tech doubles bandwidth for AI data centers, aiming to solve copper wire limits and reduce heat.
Key points
- ITRI's new 3.2T optical engine doubles data speed to 400GB/s using light instead of copper.
- The technology reduces heat and signal loss, addressing bottlenecks in AI data centers.
- DiaTrack monitors dialysis toxins in real-time, allowing doctors to adjust treatment instantly.
The Industrial Technology Research Institute (ITRI) has unveiled a new 3.2 Terabit-per-second optical engine designed to handle the exploding data demands of artificial intelligence. Announced via PR Newswire at the Taiwan Innotech Expo, this silicon photonics technology moves data using light rather than electrical signals, effectively doubling the capacity of the previous generation.
As AI workloads grow, traditional copper connections are reaching their physical limits for bandwidth and heat management. This new system allows data centers to transmit roughly 400 gigabytes of data per second, a volume comparable to twenty 4K movies, while generating less waste heat than conventional copper interconnects.
Light replaces copper for faster transmission
The core innovation lies in shifting from electrical to optical transmission. By using multiple integrated high-speed channels, the engine achieves 3.2 Tbps of total capacity. This approach reduces signal loss and thermal load, which are critical bottlenecks in modern server racks.
The trade-off involves a complex supply chain. ITRI has assembled over 20 partners across chip design and packaging to commercialize the technology. While this strengthens Taiwan's position in global AI infrastructure, it requires deep coordination to ensure consistent manufacturing quality at scale.
Real-time monitoring improves dialysis care
Beyond data centers, ITRI is deploying a similar focus on real-time data in healthcare. Their DiaTrack technology connects directly to dialysis machines to analyze toxin levels every 15 minutes. This provides doctors with immediate insights during treatment, rather than waiting for lab results hours later.
This shift allows for personalized adjustments to dialysis settings while the patient is still connected. The system has completed proof-of-concept testing with National Cheng Kung University Hospital and is scheduled for a pilot deployment in late 2026, aiming to reduce the risk of chronic inflammation.
Scaling production requires global coordination
Commercializing these technologies is not just an engineering challenge but a logistical one. ITRI is working with international firms to standardize 3.2T specifications, ensuring compatibility across different hardware ecosystems. This standardization is crucial for widespread adoption.
However, the reliance on a tightly knit network of over twenty local and international partners creates a potential single point of failure. Any disruption in this supply chain could delay the rollout of both the optical engines and the medical devices, highlighting the complexity of bringing advanced silicon photonics to market.
