Short answer: Silicon spin qubits have moved from two-qubit gate fidelities of about 98% to reports above 99% in several laboratories. One Nature paper reports 99.5% for the two-qubit gate with fast electrical control, a 2024 Nature Physics paper reports repeatable gates above 99% and ties errors to physical sources, and a 2025 Nature paper reports foundry-made devices with control fidelities above 99%. These are small devices with two qubits, and each paper reports its own benchmark.
Why the 99% mark matters
The Nature paper reporting 99.5% explains the target: fault-tolerant quantum computers rely on quantum error correction, and the surface code requires universal gate fidelities above an error correction threshold of 99 per cent, according to the paper. It says only superconducting circuits, trapped ions and nitrogen-vacancy centres in diamond had delivered that, while electron spin qubits in silicon, promising for large-scale machines because of their nanofabrication capability, had been limited to 98 per cent for the two-qubit gate because of slow operation.
99.5% with fast electrical control
The authors report a two-qubit gate fidelity of 99.5 per cent and single-qubit gate fidelities of 99.8 per cent, using fast electrical control with a micromagnet-induced gradient field and a tunable two-qubit coupling. They identify the qubit rotation speed and coupling strength at which high-fidelity gates are robustly achieved, and run the Deutsch-Jozsa and Grover search algorithms with high success rates using their universal gate set. They describe the result as universal gate fidelity beyond the fault-tolerance threshold that may enable scalable silicon quantum computers.
Consistency and error sources, 2024
A 2024 Nature Physics paper focuses on consistency, since solid-state platforms are exposed to materials-induced variability between qubits. The authors study errors in a spin qubit processor and tie them to physical origins, and demonstrate consistent and repeatable two-qubit gates above 99% fidelity in the silicon metal-oxide-semiconductor quantum dot platform. They analyse errors and fidelities across multiple devices over extended periods. Error sources include slow nuclear and electrical noise on single qubits and contextual noise that depends on the control sequence. They also investigate the effect of qubit design, feedback systems and robust gate design, and say their results show both the capabilities and the challenges in scaling up silicon spin qubits.
Foundry-made devices, 2025
A 2025 Nature paper asks whether the high fidelities seen in academic settings can be reproduced when qubits are made in a semiconductor foundry. It reports silicon two-qubit devices made with standard semiconductor tooling in a 300-mm foundry environment. Single- and two-qubit control fidelities exceed 99% for all four devices, and state preparation and measurement fidelities reach up to 99.9%, as shown by gate set tomography. They report T1 up to 9.5 s and a Hahn-echo coherence time of 1.9 ms. They find that residual nuclear-spin-carrying isotopes contribute substantially to operational errors, and identify further isotopic purification as a pathway to higher performance.
Limits
All three papers use two-qubit devices. The fidelity figures come from different benchmarking methods, and none of them demonstrates a many-qubit processor or error correction on these devices.
FAQ
Why is 99% important?
The 2021 Nature paper says the surface code needs gate fidelities above a 99 per cent threshold.
What was the highest two-qubit fidelity here?
99.5 per cent, in the paper that used fast electrical control with a micromagnet.
Can foundries make these qubits?
The 2025 paper reports four foundry-made two-qubit devices, each above 99% control fidelity.
What limits fidelity now?
The 2024 paper names slow nuclear and electrical noise and contextual noise. The 2025 paper names residual nuclear-spin isotopes.
Does this give a large processor?
No. These are two-qubit devices.
All field notes