QANT Opens Pilot Line for Optical Processor Production

Startup Q.ANT GmbH, based in Stuttgart, Germany, recently celebrated the commencement of production of lithium niobate optical processors on 6-inch wafers with a launch event. The company has made a significant investment of €14 million in its own tools to establish a pilot production line for thin-film lithium niobate (TFLN) circuits, in addition to utilizing cleanrooms and equipment provided by the Institute of Microelectronics Stuttgart (IMS) that are traditionally used for CMOS production.

This development has been recognized as a strategic move to enhance chip sovereignty in Germany and Europe. Q.ANT claims that by employing optical processing instead of electrons, their technology offers a 30-fold increase in energy efficiency and a 50-fold enhancement in computing speed. Moreover, the circuits feature much larger geometry feature sizes compared to cutting-edge electronic circuits.

Central to Q.ANT’s offerings are thin-film lithium niobate (TFLN) on insulator die that incorporate configurable arrays of optical modulators. These die consist of multiple laser light inputs and outputs, forming what the company refers to as a Native Processing Unit (NPU). To facilitate seamless access to the NPU for users of conventional AI/ML servers, Q.ANT has developed a firmware/software stack for electro-optic control known as LENA (Light Empowered Native Arithmetics).

CEO Michael Förtsch revealed that while the manufacturing has commenced on 6-inch wafers, the production line has been designed for 200mm-diameter wafers. The current production capacity stands at 1,000 wafers per year, with the potential to scale up to around 10,000 wafer starts annually. Förtsch also mentioned the company's goal to expand by replicating the development in a different mature wafer fab to achieve further scaling.

Professor Jens Anders, the director of the IMS CHIPS foundry operation, commended the transformative potential of the pilot line at IMS CHIPS, stating that it showcases how innovative technologies can flourish using existing infrastructure, thereby setting a precedent for energy-efficient next-generation computing. Q.ANT’s Native Processing Servers have the capability to accelerate various tasks including AI model training and inference, scientific and engineering simulations, real-time processing of complex mathematical equations, and high-density tensor operations for machine learning.