China's Origin Wukong Demonstrates Coherent QRAM Routing

On a 72-qubit superconducting quantum computer in Hefei, a team from Origin Quantum, the University of Science and Technology of China, and the Hefei Comprehensive National Science Center has demonstrated something the quantum computing field has struggled with for years: a working quantum router that can direct information through a bucket-brigade quantum random access memory architecture without the exponential circuit depth that has made QRAM impractical to scale.

The work, published in Physical Review X on 25 August 2026, represents the first experimental realization of coherent quantum routing on a real superconducting processor using auxiliary energy levels of transmon qubits rather than the standard two-level computational basis. By exploiting these higher excited states, the team converted what would normally require dozens of decomposed logic gates into a handful of two-level transition gates arranged in a shallow circuit — drastically cutting circuit depth and the error accumulation that comes with it.

Why QRAM matters

Quantum random access memory is a critical missing component for several landmark quantum algorithms. Classical RAM reads one address at a time. QRAM, by contrast, can query multiple addresses in superposition simultaneously — a capability that Shor's algorithm, Grover's search, and quantum machine learning models all assume exists. Without it, these algorithms lose their theoretical speedup in practice.

The bucket-brigade architecture, proposed theoretically by Giovannetti, Lloyd, and Maccone in 2008, arranges quantum routers in a binary tree. A query passes through each level like a bucket in a fire brigade, and only nodes along the active path produce errors. The architecture's operation count scales logarithmically with memory size, and its native noise resilience is attractive. But building the routers themselves has been the sticking block: each router must coherently and reversibly direct data based on an address qubit in superposition, and conventional gate-based decompositions cause circuit depth to balloon as the tree grows.

The auxiliary-level trick

The Chinese team's innovation centers on what they call a transition composite gate, or TCG. Instead of using only the ground and first excited states of a superconducting qubit — the standard computational basis — they harness higher energy levels as transient channels for information routing. Microwave pulses precisely drive information through these auxiliary levels to complete directional routing, then return the qubit to the computational basis.

This approach yields two structural advantages. First, the routing operation requires far fewer gates, producing a shallow circuit that keeps error accumulation low. Second, the encoding scheme uses non-adjacent three-level states, which means control errors automatically generate detectable markers — a built-in eraser-detection mechanism that filters out erroneous results through post-selection, without requiring additional ancilla qubits for error correction.

Experimental results

The team fabricated three independent quantum routers and one two-layer routing network on the Origin Wukong processor. The results, reported in the PRX paper:

• A single quantum router achieved 98% information transmission efficiency, meaning the quantum state arrived at the correct output port in 98% of trials.
• The two-layer routing network achieved 93% overall transmission efficiency.
• For random-access fidelity — a stricter metric measuring how faithfully the output quantum state matches the ideal — the single router averaged 94.8%, with a peak of 95.74%.
• The two-layer network's average fidelity was 82.40%.

Transmission efficiency and fidelity are distinct but related metrics. Efficiency measures whether information reaches the correct destination; fidelity measures how closely the received state matches the intended quantum state. Both must be high for QRAM to function in real algorithms.

Complementary work at Zhejiang University

Nearly simultaneously, a separate team at Zhejiang University published a related result in Nature Physics (vol. 22, pp. 745–750, 2026), demonstrating a complete bucket-brigade QRAM prototype addressable at 4-bit and 8-bit scales. That work validated the architecture's native noise resilience end-to-end and reported routing operation fidelity of 94.5%, with 4-bit QRAM query fidelity of 80.9% and 8-bit fidelity of 60.4%.

The two efforts are complementary rather than competitive. The Origin Wukong work focuses on the router component itself — maximizing its efficiency, demonstrating scalability through layering, and introducing the auxiliary-level TCG scheme. The Zhejiang University work demonstrates a complete end-to-end QRAM prototype and validates the bucket-brigade noise-resilience principle experimentally. Together, they signal that Chinese groups are producing multiple, converging advances in a subfield long dominated by theory.

From qubit counts to system function

The broader significance lies less in any single metric and more in what it signals about the trajectory of China's quantum computing program. Origin Quantum, founded in 2017 by academician Guo Guangcan and professor Guo Guoping, has pursued a full-stack strategy: chips, control systems, operating software, and cloud access. The third-generation Origin Wukong processor, launched in January 2024, carries 72 computational qubits. A fourth-generation 180-qubit system, Origin Wukong-180, came online in May 2026. By June 2026, the platform had completed more than one million quantum computing tasks for users in 163 countries.

The QRAM result fits into a deliberate shift. Rather than racing purely for higher qubit counts — the metric that has dominated the field's headlines — the team is tackling system-level functional components that quantum computers need to run real algorithms. QRAM is one such component; quantum error correction, interconnects, and compiler toolchains are others. The distinction matters because a quantum computer's practical utility depends on the quality of these supporting systems, not just the raw number of qubits.

What it does not yet mean

The demonstration is real but bounded. Three routers and a two-layer network validate the concept and its scalability direction, but a practically useful QRAM would need many more layers — and at greater depths, problems such as crosstalk, cryogenic stability, and control calibration become progressively harder. All performance figures come from the research team's own measurements and have not yet been independently reproduced. The 98% transmission efficiency represents a best-case result under specific experimental conditions, not a general-purpose performance guarantee.

Guo Guoping, Origin Quantum's chief scientist, has said he expects quantum computers to serve as accelerators working alongside classical machines within three to five years, reducing computation time and cost. The QRAM router is a step toward that vision, but the distance between a two-layer network and a system that accelerates real workloads remains substantial.

Analysis

The most meaningful takeaway is architectural. For over a decade, QRAM has been a theoretical construct whose practical implementation was blocked by circuit depth scaling. The auxiliary-level routing approach does not solve the problem outright — it will still need to prove itself at scale — but it offers a genuinely different design path, one that trades gate count for hardware-native physics. Whether that trade pays off as the network grows is the question the next round of experiments will answer.