The German startup Q.ANT has launched the world's first industrial production line for a dedicated artificial intelligence processing unit (NPU) that operates using light (photons) instead of electricity (electrons).

This first commercial photonic processor performs complex calculations using light, ensuring at least 30 times greater energy efficiency and significant improvements in computing speed compared to traditional electronics.

This processor is compatible with existing computing environments and comes equipped with an industry-standard PCIe interface, enabling its direct use in practical applications, just like NVIDIA's graphics processing units.

The Q.ANT Neural Processing Unit (NPU) is designed for applications requiring intensive calculations, such as training and running artificial intelligence models and physical simulations, especially those involving matrix multiplication, which is a fundamental operation in artificial intelligence. Implementing this operation using light makes the models significantly faster and more energy-efficient.

Q.ANT has developed photonic Thin-Film Lithium Niobate (TFLN), known for its strong electro-optic properties, extremely low optical loss, and high microwave bandwidth. TFLN is used in the fabrication of photonic components to control light properties such as waveguides, modulators, frequency converters, and light splitters and combiners, enabling efficient processing of optical signals.

The Neural Processing Unit (NPU) has already been made available on the Q.ANT cloud platform, and several artificial intelligence models have been tested on it. It has proven its ability to maintain the accuracy of calculations while consuming less power, making it a more effective option for running AI models. Now that the production line for these units has been launched, it is expected to become widely available to anyone who wants to use it.