Quantum cloud platform qBraid has announced a series of infrastructure expansions and algorithmic breakthroughs aimed at consolidating its hybrid quantum-classical development pipeline. The updates establish qBraid as a remote cloud target within the NVIDIA CUDA-Q framework, expand qBraid Lab‘s on-demand graphics processing unit (GPU) hardware fleet, and deploy Google Cloud’s AlphaEvolve automated coding agent to resolve resource bottlenecks in fault-tolerant quantum chemistry simulations.

Unified Remote Compilation and Multi-Vendor Hardware Access

Through its integration as a remote cloud target within NVIDIA CUDA-Q, developers can compile and dispatch quantum kernels directly to qBraid-supported physical hardware using the native nvq++ compiler toolchain. The architecture enables users to target hardware backends from vendors such as Rigetti, IonQ, IQM, and QuEra through a single qBraid API key by adjusting a machine flag within the execution statement. The pipeline includes access to qBraid’s free Quantum Intermediate Representation (QIR) state-vector simulator (qbraid:qbraid:sim:qir-sv), supporting asynchronous submission and on-disk future persistence for workloads scaling up to 30 qubits and 2,000 shots.

[ @cudaq.kernel ] ──► [ nvq++ --target qbraid ] ──► [ qBraid API Target ] ──► Multi-Vendor QPUs (Rigetti, IonQ, IQM)

On-Demand Fleet Expansion Across Advanced GPU Architectures

To support intensive hybrid workloads such as tensor network simulation, variational optimization, and neural-network error-correction decoding, qBraid Lab has expanded its infrastructure to offer on-demand access to over 20 GPU instance types. Orchestrated by qBraid CTO Ryan Hill, the pay-as-you-go fleet eliminates reserved capacity friction by allowing users to launch configurations directly within a browser-based JupyterLab or VS Code environment. The available compute tiers span multiple hardware generations, including:

The instance profiles support the native execution of specialized quantum calibration models, such as the NVIDIA Ising open AI family, which run pre-configured alongside the CUDA-Q compilation stack.

AI-Driven Fermion-to-Qubit Encoding via Google Cloud AlphaEvolve

Addressing the foundational mathematical layer of quantum chemistry, qBraid’s research team—including Dr. Kenny Heitritter, James Brown, and Tarini Hardikar—partnered with Google Cloud’s AlphaEvolve early-access program to optimize fermion-to-qubit encodings. Translating molecular electron structures into qubit operators presents a significant design challenge due to an exponential search space (exceeding 1050 possible configurations for an 8-orbital molecule). Leveraging Gemini models within an evolutionary loop, the AlphaEvolve agent iteratively modified a seed python structure based on qBraid’s proprietary Generalized Superfast Encoding family, evaluating approximately 1,500 program variants against a strict, un-gameable exact verifier scoreboard.

The resulting AI-generated encoding rules successfully bypassed traditional manual design constraints, achieving an exact quantum error correction code distance of 5 on dense molecular Hamiltonians, where human engineering had previously topped out at distance 3. When validated against held-out chemical systems like beryllium hydride (BeH2​) and water (H2​O) that the model never encountered during training, the generated code sustained its distance-5 protection. The discovered structure delivered a 3.4 to 7.9 times lower logical error rate under exact decoding while requiring 4.2 to 5.0 times fewer data qubits than standard fault-tolerant compilation paths, reducing the physical hardware overhead required for deep molecular simulations.

The software documentation, hardware configuration dashboards, and open-source SDK components can be reviewed via the qBraid GPU Fleet Expansion Details here, the qBraid NVIDIA CUDA-Q Target Guide here, the qBraid AlphaEvolve Quantum Chemistry Report here, and the qBraid GitHub Repository here.

June 24, 2026

Print Friendly, PDF & Email