Silicon MOS spin qubits just crossed a more useful threshold than another headline qubit count. imec and Diraq reported coherent operation and readout of an eight-qubit silicon MOS spin-qubit linear array built on imec’s 300 mm, CMOS-compatible spin-qubit platform.

That number is still small by the metrics that dominate quantum news. The engineering signal is larger: the devices were designed and fabricated in a process language semiconductor manufacturing already understands—wafer size, process control, and foundry-style discipline. The question shifts from “can a qubit work?” to “which manufacturing and control bottlenecks still block useful scaling?”

This is not evidence that a production-scale quantum processor is imminent. It is evidence that a silicon spin-qubit platform has moved beyond an isolated pair while keeping coherent operation and readout.

From a two-qubit block to an eight-qubit array

imec and Diraq describe the result as an extension of prior two-qubit building-block work to a larger array. The key word is array. A scalable processor needs more than a strong individual device; it needs a repeatable way to place, tune, read, and couple many devices without control burden exploding faster than the processor.

imec says the larger-array readout architecture did not require a significant increase in sensor count, wiring density, or thermal load. That is a promising reported scaling property—not a guarantee of performance at hundreds or millions of qubits.

All About Circuits’ July 31 coverage adds device context: the demonstrator uses silicon MOS quantum dots in an eight-dot arrangement, with a 90 nm geometry and pitch in the reported test setup. Those dimensions describe an experimental platform, not a product specification.

Why a 300 mm CMOS-compatible flow matters

A CMOS-compatible flow does not turn a quantum chip into a conventional SoC. Cryogenic behavior, charge noise, control fidelity, and error correction remain specialized constraints. It can still change the development path in three practical ways.

  1. Process learning becomes transferable. Wafer-scale metrology, contamination control, gate-stack integration, lithography discipline, and statistical process control already exist in semiconductor manufacturing. A quantum process that can use them inherits a deeper knowledge base than a one-off laboratory line.
  2. Variation becomes a design target. Scaling depends on distributions, not a record device. Engineers will need wafer- and lot-level data for threshold behavior, charge noise, qubit frequency, gate yield, and calibration burden. A strong median result fails if a wide tail makes large arrays unmanageable.
  3. The ecosystem can engage earlier. A credible 300 mm route gives materials suppliers, equipment vendors, design-tool teams, test companies, and foundries a clearer interface for standardization. Quantum expertise still matters; it gains a manufacturable substrate on which to operate.

Control-system co-design is the next bottleneck

The eight-qubit result does not erase the wiring and control challenge—it clarifies it. Each additional qubit adds operating points, readout paths, crosstalk risk, and calibration work. At cryogenic temperatures, every line, amplifier, and dissipated milliwatt counts.

Device architecture and control architecture therefore have to be designed together. A platform that limits sensor growth but demands excessive analog lines, long recalibration cycles, or weak fault isolation may not scale economically. A slightly less dense device can win if it supports simpler multiplexing, better test coverage, and predictable thermal budgets.

Standard semiconductor practices help but do not finish the job. Design-for-test must evolve into design-for-calibration. Yield learning must include quantum performance distributions. Process changes must be tracked against coherence and gate-fidelity shifts—not only classical electrical parameters.

A practical scorecard for foundry-compatible quantum claims

For the next silicon spin-qubit announcements, engineers and investors should look past qubit count and ask:

  • Was the device made on a production-relevant wafer size and process flow?
  • How wide is device-to-device variation across the wafer and across lots?
  • What are the single- and two-qubit fidelity distributions, not just the best values?
  • How do sensor count, wiring, heat load, and calibration time scale with the array?
  • Which process modules remain bespoke, and which are qualified for repeatable foundry use?
  • Can the device be tested and repaired at meaningful points in the manufacturing flow?

Those questions mark the bridge between a convincing quantum-device experiment and a semiconductor platform that can be iterated, qualified, and supplied at scale.

The engineering signal to watch

The imec–Diraq result is a material step because it pairs a larger coherent silicon MOS array with a 300 mm CMOS-compatible process. Its value is not that eight qubits settle the race to useful quantum computing. Its value is that process reproducibility, control overhead, and manufacturing yield now sit in the same conversation as coherence.

That is the durable semiconductor lesson: a quantum roadmap becomes credible when advances in qubit quality are matched by evidence that the array, the controls, and the manufacturing flow can improve together.

Sources