Fault-tolerant architecture developer Iceberg Quantum and silicon spin-qubit provider Diraq have compiled and validated Iceberg’s Pinnacle quantum low-density parity-check (qLDPC) architecture on Diraq’s silicon spin-qubit hardware platform. Enabled by early access to the logical orchestration layer of NVIDIA CUDA-Q Logical, the cross-stack integration demonstrates that high-rate qLDPC error correction codes can be compiled onto physical silicon spin hardware without requiring global non-local wiring lattices or unconstrained physical interconnects.
The compilation mapping leverages Pinnacle’s modular Processing Unit structure, restricting non-local qubit routing strictly to localized processing blocks rather than across the entire QPU die. To fulfill the intra-block non-local connectivity required by generalized bicycle (GB) qLDPC codes, the architecture utilizes electron shuttling—physically translating spin qubits across the silicon array. By optimizing intra-block shuttling schedules and code structures, the team constrained shuttling-induced decoherence to sit within standard physical gate error budgets, matching hardware-aware physical qubit count estimates to within 5% of theoretical paper predictions.
| [ Iceberg Pinnacle & Diraq Silicon Spin Compilation Parameters ] | ||
|---|---|---|
| System Layer | Hardware & Routing Implementation | CUDA-Q Logical & QEC Output |
| Physical Qubit Platform | • Diraq Silicon CMOS Spin Transistors • Physical Qubit Footprint: 150,000 Qubits | • Target Logical Qubit Output: 1,000 Logical Qubits • ~150 Physical Qubits per Logical Qubit |
| Interconnect & Routing | • Intra-Block Electron Shuttling Loops • Localized Modular qLDPC Connectivity | • Native Compilation of Logical Primitives (Adders) • Full-Stack Resource Estimation Generation |
| QEC Architecture | • Iceberg Pinnacle QLDPC (GB Codes) • Integrated Magic State Distillation & Injection | • Shuttling-Dependent Noise Budget Validation • 5% Variance vs. Theoretical Qubit Estimates |
The CUDA-Q Logical integration enabled end-to-end compilation of key fault-tolerant primitives, including logical adder circuits, mapped directly down to physical operations on Diraq hardware models. The hardware-aware compilation validates Diraq’s technical target outlined in The Case for Silicon, confirming that 1,000 logical qubits can be supported with 150,000 physical qubits (~150 physical qubits per logical qubit), reducing physical hardware requirements by nearly an order of magnitude compared to traditional surface code implementations.
Review the joint press release via Diraq Newsdesk here, inspect the technical briefing on Iceberg Quantum Blog here, explore our detailed analysis of Iceberg Quantum’s Pinnacle Architecture and QLDPC Benchmarks here, and read our previous coverage of Diraq’s Silicon QPU Roadmap and Capital Capitalization here.
September 14, 2026
