Q.ANT today announced the availability of its second-generation Native Processing Unit, the NPU 2, designed to deliver substantial gains in energy efficiency and performance for artificial intelligence and high-performance computing workloads. Developed at Q.ANT’s facilities in Stuttgart and St. Louis, the NPU 2 utilizes native analog computation in light to perform nonlinear mathematics, achieving up to 30x lower energy consumption and 50x higher performance compared to conventional CMOS-based systems. This photonic processor enables new applications, including physical AI, advanced robotics, and physics-based simulations, and is available as a complete 19” server solution with x86 host processing and Linux.
Q.ANT NPU 2: Enhanced Photonic Processing Capabilities
Q.ANT has unveiled the NPU 2, a second-generation photonic processor designed to dramatically improve energy efficiency and performance for AI and high-performance computing (HPC). Unlike traditional silicon-based chips, the NPU 2 utilizes light to perform calculations, achieving up to 30x lower energy consumption and 50x higher performance for complex workloads. This leap is enabled by native analog computation in light, bypassing the limitations of transistor-based logic and opening doors to more powerful, sustainable AI solutions.
The NPU 2’s enhanced nonlinear processing core significantly reduces the need for extensive parameters and deep training—critical for image learning, classification, and physics simulations. Delivered as a fully integrated, 19-inch rack-mountable server, the system seamlessly integrates with existing HPC infrastructures via PCIe and standard APIs (C/C++/Python). A live demonstration at Supercomputing 2025 showcases the NPU 2 learning images with a nonlinear neural network, highlighting its rapid scaling capabilities.
Q.ANT reports achieving in one year what took a decade for digital computing—demonstrating the potential of photonic processing. The NPU 2 isn’t just about speed; it’s about enabling applications previously too compute-intensive for conventional systems. Applications like defect recognition in manufacturing, optimized logistics, and advanced AI models combining statistical reasoning with physics simulation are now more economically viable—paving the way for smarter, more efficient AI solutions.
Photonic Computing: Performance and Sustainability Benefits
Q.ANT has unveiled its second-generation Native Processing Unit (NPU 2), a photonic processor designed to dramatically improve energy efficiency and performance for AI and high-performance computing (HPC) workloads. Unlike traditional silicon-based chips, the NPU 2 utilizes light for computation, enabling up to 30x lower energy use and 50x higher performance. This is achieved by performing nonlinear mathematics natively in light, sidestepping the limitations of transistor-based logic and promising a pathway beyond the scaling limits of current digital hardware.
The NPU 2’s advancements extend beyond raw speed and efficiency. Q.ANT’s technology allows for significantly reduced parameter counts and training depth in AI models – specifically for image learning and physics simulations. This means achieving greater accuracy with less computational effort. Delivered as a standard 19” server, the NPU 2 integrates seamlessly with existing infrastructure using PCIe and common APIs, making photonic acceleration readily deployable in data centers.
Q.ANT reports impressive scaling rates; achieving image classification capabilities in just one year – a feat that took a decade in traditional computing. This rapid progress positions photonic computing as a viable solution for computationally intensive tasks like computer vision, defect detection, and advanced scientific modeling. Systems equipped with the NPU 2 are available for order now, with shipments expected in the first half of 2026, promising a future where sustainability and performance go hand-in-hand.
NPU 2 Integration & Availability for HPC Workloads
Q.ANT has announced immediate availability of its NPU 2, a second-generation photonic processor designed to dramatically improve energy efficiency and performance for AI and HPC workloads. Leveraging native analog computation in light, the NPU 2 achieves up to 30x lower energy use and 50x higher performance compared to traditional CMOS-based systems. This leap is enabled by performing nonlinear mathematics directly with light, bypassing the limitations of transistor-based logic, and addresses the growing energy demands of increasingly complex AI applications.
The NPU 2 is delivered as a fully integrated, server-class system – a 19” rack-mountable unit incorporating both the photonic processors and a host x86 CPU. This turnkey approach, coupled with PCIe and C/C++/Python API support, allows for seamless integration into existing HPC and data center infrastructures. Q.ANT’s Photonic Algorithm Library (Q.PAL) provides optimized, nonlinear algorithms, accelerating tasks like image learning and physics simulation, and demonstrating accuracy gains with fewer parameters.
Early demonstrations showcase the NPU 2’s rapid scaling – progressing from digit recognition to complex image classification within a single year. Customer shipments of the NPU 2-equipped servers are scheduled for the first half of 2026. This technology promises to unlock new potential in fields requiring both high computational power and energy efficiency – including advanced robotics, drug discovery, and materials science – by making previously compute-intensive tasks economically viable.
