Q.ANT's Photonic NPU 2 Revolutionizes AI with Light-Based Computing

Q.ANT, a photonic deep-tech scale-up headquartered in Stuttgart, Germany, has announced the groundbreaking availability of its second-generation Native Processing Unit (NPU 2), marking a revolutionary leap in computing technology. This next-generation processor leverages enhanced nonlinear processing capabilities that perform mathematics natively in light, delivering orders-of-magnitude improvements in both energy efficiency and performance for AI and high-performance computing workloads. The company, founded by CEO Dr. Michael Förtsch in 2018, is offering these NPUs as integrated 19-inch server solutions complete with x86 host processors and Linux operating systems, making photonic acceleration immediately deployable in existing data center environments.

The Q.ANT NPU 2 represents a fundamental shift from traditional transistor-based computing to photonic processing, addressing the critical energy constraints that have limited AI advancement. Unlike conventional GPUs that consume massive amounts of power and require extensive cooling systems accounting for up to 40% of data center energy, Q.ANT's architecture delivers up to 30x lower energy use and 50x higher performance for complex workloads. This breakthrough enables entirely new classes of applications including physical AI, advanced robotics, next-generation computer vision, industrial intelligence, and physics-based simulation. The technology will be demonstrated live at Supercomputing 2025 in St. Louis, where visitors can experience how the NPU learns images within seconds using nonlinear neural networks through the Q.ANT Photonic Algorithm Library (Q.PAL).

What makes this announcement particularly significant is the rapid scaling of photonic computing compared to traditional CMOS technology. As Dr. Förtsch emphasized, "Photonic computing is scaling much faster than CMOS. What took ten years for digital computing, we've just achieved in one year with photonics." The second-generation NPU features enhanced analog units optimized for nonlinear network models that reduce parameter counts and training depth while improving accuracy. In practical applications, this technology makes computer vision systems economically viable for manufacturing, logistics, and inspection tasks that were previously too compute-intensive. The turnkey Native Processing Server NPS systems are available for order now with customer shipments scheduled for the first half of 2026, representing a major milestone in the transition toward sustainable, high-performance computing.