Hyunchul Jung from Norma, the Seoul‑based quantum‑computing company, has validated the performance of its quantum AI algorithms on NVIDIA’s CUDA‑Q platform, achieving up to 73.32‑fold acceleration of forward propagation for an 18‑qubit circuit compared with classical CPU‑based quantum simulators. The validation, carried out in partnership with Kyung Hee University Hospital at Gangdong, also shows backward propagation running 33.69‑ to 41.56‑fold faster and demonstrates that the newer NVIDIA GH200 Grace Hopper superchip delivers forward and backward propagation times 22 % and 24 % shorter than the older H200 GPU, underscoring the practical applicability of quantum AI for drug discovery.
Norma Achieves 73‑Fold Speedup in Quantum AI Drug Discovery with NVIDIA CUDA Q
On 1 September 2025, Seoul‑based quantum‑computing firm Norma announced that its proprietary quantum AI algorithms—QLSTM, QGAN and QCBM—delivered up to 73‑fold acceleration when run on NVIDIA’s CUDA‑Q platform. The benchmark, carried out in collaboration with Kyung Hee University Hospital at Gangdong, measured the forward and backward passes of 18‑qubit circuits that underpin drug‑discovery models. Forward propagation ran 60.14‑to‑73.32 times faster on NVIDIA H200 GPUs and GH200 Grace Hopper superchips than on conventional CPU‑based quantum simulators, while the loss‑function‑based backward propagation achieved 33.69‑to‑41.56‑fold speed‑ups. When the two chip families were compared directly, the GH200 reduced forward‑propagation times by 22 % and backward‑propagation times by 24 %, demonstrating the practical advantage of CUDA‑Q for hybrid quantum‑classical workloads.
Norma’s validation showed that CUDA‑Q not only boosts computational throughput but also enables realistic verification of algorithms before deployment on physical quantum hardware, cutting development costs and time. CEO Hyunchul Jung highlighted the partnership as a model for collaboration between domestic and international quantum‑technology companies and medical institutions, and announced plans to extend performance testing of quantum AI algorithms to defence and finance sectors.
Expanding Quantum AI Validation Beyond Pharma Signals Broad Industry Potential
In a September 2025 announcement, Norma outlined a strategy to apply the same CUDA‑Q benchmarking that proved successful in the drug‑discovery partnership to high‑impact domains such as defence and finance. The validation, conducted at Kyung Hee University Hospital’s Gangdong campus, measured 18‑qubit circuits on NVIDIA H200 GPUs and GH200 Grace Hopper superchips, yielding forward‑propagation times 60.14‑to‑73.32 times faster and backward‑propagation times 33.69‑to‑41.56 times faster than CPU‑based simulators. The GH200’s 22 % and 24 % improvements over the H200 further underscore the incremental advantage of the newer superchip architecture.
These performance figures, achieved within the CUDA‑Q environment that seamlessly integrates GPUs and quantum processing units, demonstrate that quantum AI algorithms can deliver tangible speed‑ups across a spectrum of industries. In defence, rapid simulation of quantum‑enhanced optimisation problems could shorten development cycles for secure communication protocols, while in finance, the ability to evaluate complex risk models in a fraction of the time could transform portfolio optimisation. Norma’s work shows that the cost and time barriers traditionally associated with quantum‑classical development can be substantially lowered, providing a concrete data set that supports broader adoption of quantum hardware beyond the pharmaceutical arena.
Original Press Release
Source: View Original Source
