Conductor Quantum and SemiQon want AI-controlled silicon qubits to make quantum look like normal chips

YC-backed Conductor Quantum has partnered with Finnish startup SemiQon to push a very specific vision of quantum computing: silicon spin qubits fabricated in something closer to standard CMOS lines, controlled by AI-driven software that automates calibration, tuning, and readout. Rather than lab-scale experiments on exotic materials, the pitch is a quantum stack that looks and feels more like conventional semiconductor design and operation.

What SemiQon brings

SemiQon builds silicon-based quantum processors using customised CMOS processes. Their hardware integrates quantum dots (hosting spin qubits) with cryogenic CMOS on the same chip, aiming for:

Higher qubit density: Smaller physical qubits than many current superconducting implementations, enabling more qubits per unit area.
Monolithic integration: On-chip control logic reduces latency and wiring complexity versus setups that route every qubit to bulky external electronics.
Manufacturing compatibility: Using CMOS-like flows improves the odds of scaling production across existing fabs, rather than building entirely new process lines.

What Conductor Quantum adds

Conductor Quantum’s focus is AI-driven control software. Tuning quantum devices is currently human-intensive: experimentalists spend huge amounts of time calibrating gates, finding the right operating points, and compensating for drift. Conductor’s software stack aims to:

Automate qubit creation and tune-up: Use machine learning models that understand quantum transport behaviour in silicon devices to find and stabilise usable qubits.
Handle calibration at scale: As qubit counts rise towards thousands or more, manual tuning does not scale. Automated control loops become mandatory.
Provide a “control OS”: Offer APIs (like their Stanza and Models platforms) that hide low-level control complexity from higher-level quantum software stacks.

Why silicon and AI matter together

There are two bottlenecks in current quantum efforts: hardware that can scale to large numbers of qubits, and the control software that can keep those qubits usable. SemiQon is attacking the first with silicon-first hardware; Conductor is attacking the second with AI-based control. The partnership is about aligning those two tracks.

If you can fabricate qubit arrays on silicon and automatically tune them with AI, you have a plausible path to quantum processors that look more like many-core chips than lab experiments: etched, packaged, and driven by control software that runs predictable routines rather than bespoke scripts.

How realistic is “million-qubit” marketing?

SemiQon’s public messaging talks about building quantum processors for a “million-qubit era.” That is aspirational, but the basic logic is sound: smaller qubits + CMOS tooling + monolithic integration give you a better scaling story than macroscopic devices wired to racks of off-the-shelf electronics. The hard parts remain:

Maintaining coherence and gate fidelities as qubit counts grow.
Managing error correction overhead, which multiplies physical qubit requirements.
Ensuring cryogenic power and heat loads stay within realistic cooling budgets.

AI control does not remove those physics limits, but it should reduce the human bottlenecks that currently govern how many qubits you can keep tuned at once.

Competitive landscape

Superconducting qubits: Players like Google and IBM are further along in deployed qubit counts, but their hardware looks less like “normal” CMOS from a fab perspective.
Trapped ions and neutral atoms: Excellent coherence, but very different scaling stories; they do not map as neatly onto semiconductor manufacturing.
Other silicon-spin efforts: SemiQon is not alone here; academic and industrial teams are also pursuing spin qubits in silicon with different process trade-offs.

What to watch next

Joint demos showing SemiQon hardware controlled by Conductor’s software at non-trivial qubit counts.
Evidence that automated tune-up improves both stability and uptime versus manually tuned devices.
Any announcements on fab partnerships or pilot lines that put SemiQon hardware into more mainstream manufacturing flows.

Editor’s take

Quantum is full of PowerPoint roadmaps to “millions of qubits.” The interesting thing here is not the slogan but the alignment: pick a qubit technology that likes CMOS, and pair it with control software that assumes you will have to manage thousands of devices without humans in the loop. If quantum ever becomes a normal part of the compute stack, it will look more like this than like a lab bench full of coax cables.