Oracle and Classiq have demonstrated a quantum software engineering workflow capable of running a 36-qubit simulation, a significant leap from Classiq’s support for simulations with up to 29 qubits in its standard flow. The demanding simulation, focused on a portfolio optimization application, required 512 GiB of memory, highlighting the immense computational resources needed for even simulated quantum workloads. This was achieved by connecting Classiq’s quantum expert AI agent, which generated a Jupyter notebook from a natural language prompt, with Oracle’s high-performance NVIDIA DGX A100 GPUs on Oracle Cloud Infrastructure. The entire workflow, from initial prompt to a coherent prototype, took less than 15 minutes, demonstrating a pathway to accelerate quantum application development.
Classiq AI Agent Generates Portfolio Optimization Notebooks
The convergence of artificial intelligence and quantum computing recently advanced as Classiq demonstrated an AI agent capable of autonomously generating functional quantum notebooks for complex financial modeling. This capability, coupled with Oracle’s high-performance computing infrastructure, enabled a 36-qubit simulation, a substantial increase from Classiq’s platform supporting simulations with as many as 29 qubits in its standard flow, and showcased a streamlined quantum software engineering process. The demonstration centered around a portfolio optimization application, highlighting a practical path from initial concept to executable code. A key challenge in quantum application development is the sheer computational demand of simulation; a 36-qubit statevector alone requires approximately 512 GiB of GPU device memory, assuming single-precision floating-point numbers. This scale is critical, as it allows for validation of models and debugging of circuits before deployment on actual quantum hardware.
The workflow began with a natural-language prompt directed at Classiq’s AI agent, which then generated a complete Jupyter notebook for a discrete portfolio optimization problem. The entire process, from prompt to coherent prototype, took under 15 minutes. The application modeled 12 correlated assets, each with eight possible allocation levels, creating a vast search space of 68.7 billion possible raw allocation combinations before applying the budget requirement. Utilizing the Quantum Approximate Optimization Algorithm (QAOA), the 36-qubit circuit, with a depth of 730, was synthesized by Classiq’s engine, remaining readable and adaptable without requiring developers to work at the gate level. The resulting quantum-sampled portfolio allocation exhibited a 4.63% objective gap compared to a classical benchmark, but demonstrated a “strong modeled objective value,” proving the viability of the AI-driven workflow and scalable simulation infrastructure.
Qubit Simulation Achieved via Oracle Cloud Infrastructure
Quantum simulation is increasingly vital as developers seek to validate models and debug circuits before utilizing nascent quantum hardware; however, the computational demands rapidly escalate with circuit complexity. This achievement highlights the growing importance of robust classical infrastructure in supporting the quantum ecosystem. The scale of the simulation is particularly noteworthy, requiring 512 GiB of memory to represent the quantum state. A 36-qubit statevector, comprised of 68.7 billion possible raw allocation combinations before applying the budget requirement, necessitates substantial resources; assuming single-precision floating-point numbers, the raw statevector alone demands considerable GPU device memory even before accounting for simulator overhead. Oracle addressed this by utilizing eight NVIDIA DGX A100 GPUs, distributing the workload to accommodate the massive dataset. The simulation itself ran for approximately five hours, demonstrating a concrete timeframe for executing complex quantum tasks.
This notebook included problem setup, a hybrid classical-quantum workflow, a classical benchmark, and result analysis. According to the team, “The workflow produced a feasible portfolio allocation with a strong modeled objective value and a clear path for further refinement,” indicating the practical potential of this combined approach. They were able to create a coherent prototype in under 15 minutes from a simple prompt, significantly reducing development time. The researchers suggest that “This combination allows organizations to scale their quantum activity and achieve meaningful results today, maintaining model portability as hardware continues to scale,” suggesting a viable path toward practical quantum applications.
While there was a 4.63% objective gap relative to the classical reference, the more important result was the workflow itself: Oracle GPU infrastructure enabled a simulation beyond the scale available in standard quantum development environments, while Classiq simplified the path from natural-language intent to a synthesized, executable quantum program.
Quantum Approximate Optimization Algorithm Implementation Details
The successful execution of a 36-qubit simulation, facilitated by Oracle Cloud Infrastructure, marks a significant step forward in scaling quantum algorithm development, particularly for the Quantum Approximate Optimization Algorithm (QAOA). This requirement drove the use of eight NVIDIA DGX A100 GPUs, underlining the necessity of distributed, multi-GPU setups for complex quantum simulations. The application chosen for this proof of concept, a discrete portfolio optimization problem, modeled 12 correlated assets with a budget of 15 chunks, creating a substantial search space of 68.7 billion possible raw allocation combinations before applying the budget requirement. While classical optimizers can exploit structure within such problems, the scale proved useful for validating a quantum optimization workflow. The implemented 36-qubit circuit, generated using Classiq’s platform, utilized three layers of QAOA and a depth of 730 gates, yet remained “readable and easy to adapt” according to the researchers, due to the use of high-level algorithmic constructs.
The team leveraged Classiq’s synthesis engine to implement the circuit, demonstrating a capacity to translate complex problems into executable quantum programs. The entire five-hour simulation, running across the NVIDIA GPUs, employed 16,384 shots per sampling and 15 outer iterations of the COBYLA optimizer. Results showed a 4.63% objective gap between the quantum-sampled portfolio and a classical benchmark, but the primary achievement was the workflow itself.
Hybrid Workflow Yields Comparable Portfolio Allocation Results
The ability to rapidly prototype and test quantum algorithms is becoming increasingly vital, and a recent collaboration between Oracle and Classiq has demonstrated a hybrid approach capable of handling simulations previously out of reach for many developers. The core of the achievement lies in connecting Classiq’s AI-driven quantum software engineering platform with Oracle Cloud Infrastructure (OCI). This notebook, while AI-generated, underwent technical review by the Classiq team, who adjusted parameters and validated execution before initiating the large-scale simulation. The application modeled 68.7 billion possible raw allocation combinations before applying the budget requirement, and the results showed a 4.63% objective gap relative to a classical benchmark.
