Norma, a South Korean quantum computing company, has validated the performance of its quantum AI algorithms on the NVIDIA CUDA-Q platform. By running these algorithms for drug development on NVIDIA GH200 Grace Hopper Superchips, the company observed computational speeds up to 73 times faster compared to traditional CPU-based quantum simulators. This project was launched as part of a joint research effort with Kyung Hee University Hospital at Gangdong, aimed at discovering drug candidates.
Norma’s quantum AI algorithms, such as QLSTM, QGAN, and QCBM, were tested on the NVIDIA CUDA-Q platform, which is designed to integrate GPUs and QPUs for hybrid operations. The test utilized NVIDIA H200 GPUs and GH200 Superchips. The results showed that the execution and measurement (forward propagation) of an 18-qubit quantum circuit was between 60.14 and 73.32 times faster than CPU-based methods, while the loss function-based correction process (backward propagation) was between 33.69 and 41.56 times faster. The NVIDIA GH200 demonstrated a 22% shorter forward propagation time and a 24% shorter backward propagation time compared to the H200.
The project is positioned as having reduced development costs and time by enabling verification of algorithms prior to deployment on quantum hardware. This work demonstrates the practical applicability of quantum AI technology in drug discovery. Norma plans to continue expanding performance testing of its quantum AI algorithms across sectors such as healthcare through technological cooperation with NVIDIA.
Read the full announcement here.
September 1, 2025
