A fixed-frequency superconducting chip can now entangle distant qubits directly, skipping the intermediate steps that kill circuit fidelity.
A master mason points to a direct circuit path, ignoring the traditional segmented arch design.
Researchers at Information Engineering University in Zhengzhou, China have demonstrated a 99.0% average gate fidelity for an “N3-CZ” gate that directly couples next-nearest-neighbor superconducting qubits. The implementation uses fixed-frequency and fixed-coupler qubits, a topology long dismissed as too rigid for complex circuits. The result is not an incremental fidelity gain. It is a direct challenge to the planar, nearest-neighbor architectures that Google and IBM have bet their fault-tolerant roadmaps on.
The Swap-Gate Bottleneck
A simpler-robed cartographer sketches a direct bridge while an ornate-robed colleague draws a fragile stepping-stone route.
Standard superconducting quantum processors arrange qubits in a two-dimensional grid where each qubit talks only to its immediate neighbors. To entangle two distant qubits, you execute a daisy-chain of SWAP gates, shuttling quantum information step by step across the chip. Each SWAP is a composite operation built from multiple two-qubit gates. Each gate introduces error. The accumulated infidelity across a chain of SWAPs can swamp the circuit before it completes.
The industry accepted this as a hardware constraint and chased higher individual gate fidelities as the fix. The logic was simple: make each gate so reliable that the chain works. The N3-CZ result shows that bypassing the chain entirely can be more effective than perfecting its links.
How the N3-CZ Gate Works
The N3-CZ gate is implemented via simultaneous cross-resonance drives applied to fixed-frequency, fixed-coupler superconducting qubits. Instead of routing through the intermediate neighbor, the gate creates a direct controlled-Z interaction between qubits separated by one intermediate node. Quantum process tomography confirmed 98.9% fidelity under realistic conditions, as reported by Quantum Zeitgeist.
The practical payoff is immediate and measurable. The research team demonstrated that the N3-CZ gate reduces circuit depth and gate count in quantum ripple-carry adders and graph state preparation. The team stated, “we demonstrate that the N3-CZ gate significantly reduces circuit depth and gate count in key quantum applications.”
A 99.0% gate that replaces three 99.9% SWAP gates delivers a higher-fidelity circuit. The math is straightforward. The industry’s focus on raw two-qubit gate fidelity misses the point. Chasing another decimal on a nearest-neighbor CNOT while ignoring topology is the wrong problem. The N3-CZ result exposes how much of the fault-tolerant roadmap is built on an artificially constrained, planar neighbor model.
The Topology War Has Started
This single result shifts the competitive moat from raw qubit count to connectivity graph efficiency. A fixed-frequency chip with next-nearest-neighbor gates can run shallow, error-corrected circuits with fewer physical qubits than a planar chip with higher individual gate fidelities. The manufacturing advantage compounds the threat. Fixed-frequency, fixed-coupler qubits are simpler to fabricate and calibrate than the tunable, frequency-crowded architectures that dominate Western efforts. They avoid the crosstalk and frequency-collision headaches that plague tunable designs.
The broader fidelity landscape supports this direction. Fluxonium qubits have demonstrated 99.94% CNOT gate fidelity stable above 99.9% for 24 days without recalibration, with a 60 ns gate duration and non-decoherence-related errors isolated to 2 × 10⁻⁴, according to a 2024 arXiv preprint. Neutral-atom processors hit 99.854% CZ gate fidelity, improving to 99.941% with loss postselection. Spin qubits reached 99.992% single-qubit fidelity in simultaneous operation, with Lawrie et al. noting in Nature Communications that “next-nearest neighbor pairs are highly robust to cross-talk errors.” A solid-state spin system at room temperature achieved 99.920% CNOT fidelity by noise filtering, per Physical Review Letters.
The pattern is clear. High-fidelity gates are appearing across multiple platforms. The differentiator now is not who can squeeze out another 0.01% on a single gate. The differentiator is whose connectivity graph lets them run useful circuits before decoherence wins.
Here is the prediction: within 18 months, a major cloud quantum provider will license or replicate a next-nearest-neighbor gate scheme. AWS, Azure, or Google itself will move to avoid being undercut by Chinese processors that compress circuits more efficiently. The alternative is watching a fixed-frequency chip with fewer total qubits outperform a planar flagship on algorithm-relevant benchmarks.
What This Means for the Field
For quantum algorithm designers, the message is clear. Start optimizing for connectivity graphs, not just gate counts. The compiler that maps logical circuits to physical qubits is now the leverage point, and architectures that simplify that mapping will win.
For investors, the calculus shifts. The company that solves the wiring problem without exotic hardware wins the noisy intermediate-scale quantum era. Fixed-frequency, fixed-coupler designs have lower calibration overhead and fewer failure modes. That translates to faster iteration cycles and sooner commercially useful results. A fidelity of 99.99% implies an average error of about 0.01% per gate under the measurement method used, but as upqubit.com notes, that does not mean every circuit will behave correctly 99.99% of the time. Circuit-level fidelity is what matters, and topology determines circuit-level fidelity.
For competitors, the simplicity is the threat. Tunable qubit architectures carry a complexity tax that fixed-frequency designs do not. The N3-CZ gate shows that simplicity does not mean limited capability. It means a faster path to circuits that work.
The fixed-frequency chip did not just hit a fidelity number. It jumped over the wall that Google and IBM have been trying to tunnel through with incremental gate improvements. The SWAP gate is becoming legacy logic. The race is now for the most efficient graph, not the most qubits. The Zhengzhou team just put a flag in that ground.
