Q.ANT's photonic processor has gone live at Leibniz Supercomputing Centre, cutting energy use by up to 90%.
www.allaboutcircuits.com, Jul. 30, 2025 –
What if the future of AI computing wasn’t electric at all? German startup Q.ANT has deployed the "world’s first commercial photonic processor" for AI workloads at the Leibniz Supercomputing Centre (LRZ), turning light into the newest building block for high-performance computing (HPC).
This marks the first time an analog photonic co-processor has been integrated into an operational high-performance computing (HPC) environment. By partnering with LRZ, one of Europe’s top data centers, Q.ANT is putting its light-powered processor to the test in real-world AI and scientific simulation applications. The deployment coincides with Q.ANT’s €62 million ($71.8 million) Series A funding round, one of Europe’s largest in the photonics space, which will drive global expansion and commercialization of its energy-efficient technology.
The Q.ANT Native Processing Server (NPS) is a major architectural shift for HPC systems. The company built NPS using a proprietary thin-film lithium niobate photonic chip known for fast switching and low optical loss. The NPS performs computations using light instead of electricity. This approach enables remarkable performance gains: up to 90 times lower energy consumption per workload and up to 100 times greater data center capacity, thanks to its high computational density and passive cooling requirements.
By eliminating heat generation at the chip level, the photonic processor avoids the need for active cooling systems that normally consume a significant portion of a data center’s energy budget. The system delivers 16-bit floating point operations with near-perfect accuracy and supports integration with mainstream AI tools like PyTorch, TensorFlow, and Keras through standard PCIe interfaces.
The NPS integrates this photonic core into a PCIe form factor compatible with standard x86 systems. Supporting both C++ and Python APIs, the NPS allows developers to run familiar AI frameworks like PyTorch and TensorFlow while tapping into a photonic backend for acceleration. With 16-bit floating-point precision and consistency close to 100%, the chip executes complex workloads, like image segmentation or speech recognition, at speeds far beyond conventional digital processors, all while consuming 45 W of power.
Photonic computing’s ultra-fast data handling makes it especially well-suited for modern AI applications. The NPS can process vast amounts of data using multiple wavelengths of light simultaneously, drastically improving throughput for tasks like image recognition, neural network inference, and scientific simulations.