Two papers published on 15 October tackle different bottlenecks on the road to useful quantum computers. In Science, Google Quantum AI reports producing the scarce “magic states” needed for universal quantum computation with about 40 times fewer errors. In Nature Physics, RIKEN researchers show that silicon spin qubits can keep very high accuracy while several are controlled at once. Both are incremental but meaningful advances.
Quantum computers promise to solve certain problems, in chemistry, materials science and cryptography, that are out of reach of conventional machines. The obstacle is error. Qubits are fragile, and useful algorithms need billions of operations with very low error rates. The field’s answer is quantum error correction, which spreads information across many physical qubits to create more reliable “logical” qubits. Two papers published on Thursday address different parts of that challenge.
Google: cultivating magic
Error-corrected quantum computers can carry out a set of operations called Clifford gates relatively cheaply. But Clifford gates alone cannot perform universal quantum computation; a classical computer can efficiently simulate them. The missing ingredient is usually supplied by special resource states known as magic states, such as the “T state”, which are consumed to implement non-Clifford operations.
Preparing high-quality magic states is expensive. The standard method, magic-state distillation, combines many noisy copies to produce a few good ones, and Google notes that some estimates suggest up to 95% of a future quantum computer’s capacity could be devoted to producing them. Because each state is used up once, they are a constant bottleneck.
In 2024, Google researchers Craig Gidney, Noah Shutty and Cody Jones proposed a cheaper alternative called magic state cultivation. Instead of distilling many states, a single state is prepared inside a small error-correcting code, repeatedly checked, and “grown” into a larger, better-protected code, with faulty attempts discarded early.
Willow, which Google unveiled in December 2024 as a 105-qubit superconducting processor, was used to show that logical-qubit errors fall as an error-correcting code is made larger, a key milestone known as operating “below threshold”. The code distance describes this size: a distance-5 code can correct more errors than a distance-3 code, at the cost of more physical qubits.
The new Science paper, “Magic state cultivation on a superconducting quantum processor”, reports the first experimental demonstration on Google’s Willow chip. According to Google, the team prepared logical T states with a record fidelity of 99.99%, about 40 times fewer errors than states prepared without cultivation. They also “grafted” a state from a distance-3 colour code onto a distance-5 surface code, moving it from one error-correcting scheme to another.
“We’ve shown that magic state cultivation works on Google Quantum AI hardware,” write Emma Rosenfeld and Craig Gidney. They add that “plenty of technical challenges remain” and that it must still be demonstrated at larger scale. Google’s blog does not state how many qubits were used or what fraction of attempts was discarded, which matters because cultivation trades discarded attempts for quality. The team also had to correct for laboratory noise from drifting electronics and temperature changes.
RIKEN: many qubits, one wire
The second paper addresses the hardware itself. Superconducting qubits, used by Google and IBM, are large and must be cooled to near absolute zero with extensive wiring. Silicon spin qubits, which store information in the spin of single electrons trapped in quantum dots, are far smaller and can in principle be made with the same manufacturing techniques as conventional chips. Their challenge is control: driving one qubit with microwaves can disturb its neighbours.
Writing in Nature Physics, a team from the RIKEN Center for Emergent Matter Science led by Yi-Hsien Wu and Leon Camenzind, in the group of Seigo Tarucha, reports a five-quantum-dot silicon array controlled through a single shared microwave line. Rather than calibrating every combination of qubits, they used calibrations between pairs of qubits to compensate for the unwanted phase shifts caused by simultaneous driving.
According to reports of the paper, individual qubits reached fidelities above 99.99%, with the best near 99.999%. With up to three qubits driven simultaneously, fidelity stayed above 99.99%, and with all five driven at once it remained at or above 99.9%. That is often cited as a practical threshold for error-correction schemes such as the surface code. The remaining loss was attributed to decoherence caused by the stronger microwave drive needed for multiple qubits.
“High-fidelity control can be maintained even when several qubits are operated at the same time,” Camenzind said. The approach matters because the number of control lines is one of the main barriers to scaling: a device that needs a separate wire for every qubit cannot easily grow to thousands. Industrial efforts, such as Intel’s 12-qubit “Tunnel Falls” silicon research chip released in 2023, face the same wiring problem. The results concern single-qubit operations; two-qubit gates, which are harder, were not the focus.
Putting it in perspective
Neither paper produces a useful quantum computer. Google’s result improves one essential ingredient on existing hardware, but it is company-reported, and NSH has not read the full paper. RIKEN’s result is a careful engineering advance on a five-qubit device, far from the millions of qubits that large-scale machines may need. No independent expert comment was available at the time of writing.
Together, however, they show how the field is now advancing: through error-correction and control techniques that reduce the cost of building fault-tolerant machines.
Why it matters for India
India’s National Quantum Mission, approved by the Union Cabinet in April 2023 with an outlay of about ₹6,003 crore to 2030–31, aims to develop intermediate-scale quantum computers with 50 to 1,000 physical qubits over eight years, using platforms that include superconducting and semiconductor qubits. Thematic hubs at IISc, IIT Madras, IIT Bombay and IIT Delhi lead its work on computing, communication, sensing and materials. Techniques that reduce the overhead of magic states, or allow dense control of silicon qubits, are directly relevant to those goals and to India’s parallel semiconductor ambitions.
