Researchers have built an 11-qubit quantum processor from individual phosphorus atoms embedded in isotopically purified silicon-28, achieving two-qubit gate fidelities of 99.9 percent. The chip uses a modular two-register architecture that pairs nuclear and electron spins as qubits, and its physical-level benchmarks exceed 99 percent across the board. Because the underlying material is silicon, the same element that powers every smartphone and laptop on Earth, the work raises a direct question: can quantum computers be mass-produced in the factories that already churn out classical chips?
Why silicon fidelity records change the quantum timeline
Most quantum computing platforms, including superconducting circuits and trapped-ion systems, rely on exotic fabrication environments that share little with conventional semiconductor manufacturing. A silicon-based approach that matches or exceeds their error rates would let builders tap a global network of chip foundries instead of constructing specialized facilities from scratch. That is the practical weight behind the new processor’s reported numbers.
The 11-qubit atom processor described in a recent Nature report achieves physical-level benchmarks above 99 percent, with two-qubit gate fidelities reaching 99.9 percent. Those figures sit among the best recorded for any qubit technology at this scale. A separate study demonstrated logical encoding and universal logical gate operations on five nuclear spins using a [[4,2,2]] error-detecting code, then ran a variational quantum eigensolver to compute the ground-state energy of a water molecule. That sequence, moving from raw gate quality to encoded logical operations to a chemistry application, compresses into silicon hardware a progression that took superconducting platforms years to achieve.
One hypothesis worth testing is that if each additional qubit pair continues to add roughly 0.1 percentage points of fidelity, the donor architecture could reach the 99.99 percent threshold often cited for distance-5 surface-code logical qubits by around 2027. The current data does not confirm a linear trend at that rate, and the published results cover only 11 physical qubits. Scaling from 11 to the dozens or hundreds required for a distance-5 code introduces new sources of crosstalk, control complexity, and thermal noise that could flatten or reverse the improvement curve. The hypothesis is plausible but unproven, and any realistic roadmap has to treat it as an optimistic upper bound rather than an expectation.
Gate fidelities, logical operations, and the water molecule test
Three peer-reviewed papers anchor the technical case. The Nature paper on the 11-qubit processor describes a modular design in which phosphorus donor atoms are placed in purified silicon-28, a substrate engineered to remove most of the nuclear-spin noise that ordinary silicon introduces. Each donor contributes both a nuclear spin and an electron spin, giving the architecture two types of qubit with different strengths: nuclear spins hold information for long periods, while electron spins enable fast interactions between distant qubits.
This design builds on earlier demonstrations of fast two-qubit interactions between donor electrons, showing that the coupling mechanism can be both rapid and precise. In the new device, the team scaled to 11 qubits and performed randomized benchmarking across multiple nuclear spins, reporting single-qubit Clifford fidelities alongside detailed two-qubit gate metrics for controlled-Z operations. Those measurements use widely adopted protocols, allowing direct comparison with competing platforms. The results exceeded 99 percent at the physical level and reached 99.9 percent for the best two-qubit gates, pushing silicon donors into the top tier of qubit technologies by raw gate quality.
The logical-operations study pushed the same hardware class further. Using five nuclear spins encoded in a [[4,2,2]] code, researchers demonstrated universal logical control and applied it to estimate the ground-state energy of a water molecule through a variational quantum eigensolver. That calculation is a standard benchmark in quantum chemistry, chosen because classical computers can verify the answer to high precision. Completing it on encoded qubits, rather than raw physical qubits, signals that the silicon donor platform can support error-protected computation, not just isolated high-fidelity gates.
The encoded experiment is particularly important for assessing scalability. Error-detecting codes like [[4,2,2]] do not yet provide full fault tolerance, but they force the hardware to implement sequences of gates, measurements, and feedforward operations that resemble those required for larger surface-code patches. Demonstrating universal logical gates on this code shows that the control stack can coordinate multiple qubits with sufficient timing precision and stability to keep errors from exploding during a modest algorithm. It is an early but concrete sign that the architecture can move beyond toy demonstrations toward structured error correction.
Separately, researchers affiliated with the same program reported that silicon quantum chips had been produced via standard semiconductor fabrication line processes, including CMOS-manufactured qubits. While that work involved a different device from the 11-qubit donor chip, it demonstrated that quantum-grade silicon components can survive fabrication steps similar to those used for conventional integrated circuits. Together with the high-fidelity donor results, this supports the broader claim that silicon-based quantum processors might eventually piggyback on existing foundry infrastructure.
Scaling gaps and what to watch before 2027
Several pieces of evidence are still missing from the public record. The primary Nature papers and associated preprints do not disclose the exact CMOS process node, layer stack, or yield statistics from any commercial fabrication run. Institutional reports confirm fab-line compatibility in general terms but do not specify defect rates or throughput for the donor-qubit devices specifically. Without yield data, any projection about mass-manufactured quantum chips remains speculative.
There is also limited information about how the control electronics will scale. The current 11-qubit device relies on carefully engineered microwave and radio-frequency control lines, with cryogenic wiring that is feasible at small scales but becomes challenging as qubit counts climb into the hundreds or thousands. Integrating more of the control stack into cryo-CMOS circuits is an obvious path, yet this introduces additional heat loads and potential noise sources that could degrade the very fidelities that make the donor platform attractive.
Another open question is how robust the record fidelities are to device-to-device variation. The reported benchmarks focus on a specific chip, characterized in detail. To enable industrial-scale production, the architecture must tolerate fabrication imperfections in donor placement, local strain, and interface quality while still delivering high-fidelity gates across many dies. That, in turn, requires statistical studies over large device batches, which have not yet been reported.
On the algorithmic side, the water-molecule calculation is a meaningful milestone but only a small step toward applications that would justify large-scale hardware. The variational quantum eigensolver used in the experiment is relatively shallow and resilient to certain noise patterns. More demanding workloads, such as larger molecular systems or error-corrected implementations of phase estimation, will require deeper circuits and tighter error budgets. Demonstrations of such algorithms on encoded donor qubits would provide a clearer picture of how the platform performs under stress.
Between now and 2027, several indicators will show whether the optimistic fidelity projections are on track. First, independent replication of 99.9 percent two-qubit gates on different donor chips would validate that the performance is not a one-off result. Second, scaling to tens of qubits with comparable fidelities, while maintaining low crosstalk, would suggest that the architecture can handle increased complexity. Third, publishing even coarse-grained yield and variability data from fabrication runs would clarify how close the technology is to true foundry readiness.
If those milestones materialize, the silicon donor approach could move from laboratory curiosity to a leading candidate for fault-tolerant quantum processors. If they do not, the current achievements will still stand as a striking demonstration of what is possible when atomic-scale precision meets the material that already underpins modern computing. In either case, the 11-qubit processor marks a turning point: it shows that the path to scalable quantum computing may run not only through exotic new materials and bespoke fabrication lines, but also through the familiar silicon wafers that have been refined over decades of classical chipmaking.
More from Morning Overview
*This article was researched with the help of AI, with human editors creating the final content.
