Cryogenic hardware accelerator for Quantum State Discrimination at 4K

A quantum computing unit (QPU) requires tight integration with supporting classical electronic circuits (CEC). Current systems use room temperature electronics, such as CPU hosts and field programmable field arrays (FPGAs), for the CEC. Unfortunately, this approach faces scalability issues due to spatial, mechanical, and thermal constraints. Implementing the CEC with superconducting computing offers a potential solution, as superconducting circuits can operate at over 50GHz with very low switching energy inside the cryogenic refrigerator. This work introduces a novel superconducting, machine learning based, hardware accelerator for qubit state discrimination that operates natively at 4K. Our accelerator addresses one of the major superconducting challenges, in particular the low area density that had hindered practical superconducting prototypes, by using a biologically inspired data representation that enables minimum footprint computing units. Our proposed superconducting ML accelerator for qubit state discrimination is 12X faster and requires 20X less area than state-of-the-art, requiring less than 17k JJs for the computing elements.