Ep 9: Can an Ordinary Chip Factory Build a Quantum Computer?

Today on Quickly Quantum: can an ordinary chip factory — the kind that already makes the processor in your phone — actually build a quantum computer? Diraq and the Belgian chip-research hub imec say yes, and they've got eight coherent qubits on a real 300 millimeter production line to prove it. Before that, in the headlines: a bosonic-qubit startup out of Quebec just knocked its error rate down a hundredfold, the EU launched a fifty million euro pilot line for neutral-atom chips, a new theoretical framework might solve trapped-ion computing's wiring problem, and quantum stocks got dragged into an oil-price shock that had absolutely nothing to do with quantum. Welcome back to Quickly Quantum, your daily brief on the quantum frontier. It's Tuesday, July 14, 2026. Let's get into it.

First up: Nord Quantique, the bosonic qubit startup out of Sherbrooke, Quebec, just published a peer-reviewed result that quietly moves the goalposts on qubit quality. They've pushed SPAM errors — that's state-preparation-and-measurement error, basically how often a qubit gets loaded into the wrong state or misread when you check it — below zero point one percent on a single GKP grid-state qubit. GKP stands for Gottesman-Kitaev-Preskill, a way of encoding a qubit inside the wobble of a microwave photon rather than a single physical particle, which in theory makes it naturally resistant to certain kinds of noise. And here's the number that matters: that's roughly a hundredfold improvement over where this platform sat before, bringing bosonic qubits up to the same error 'bookends' — the best-case, well-behaved error rates — that leading superconducting transmon qubits already hit. The kicker is they did it without bolting on new hardware; same architecture, just tuned better. Now, why should you care if you've never heard of a GKP qubit before today? Because the whole pitch behind bosonic qubits is that you need fewer of them to build one reliable 'logical' qubit — the error-corrected unit that actually does useful computation — since each one is already partially self-correcting. If SPAM errors this low hold up at larger scale, that's a real argument for needing less hardware overhead than rivals building error correction the brute-force way, stacking dozens of ordinary qubits together to get the same protection. It's one experiment on one qubit, so treat it as a promising data point rather than a verdict — but it's exactly the kind of quiet, technical result the error-correction crowd will be citing for the next year.

Next, a caution flag worth flagging loudly. An analysis of hardware from IQM found that zero-noise extrapolation — a widely used error-mitigation technique where you deliberately run a calculation at several noise levels and then extrapolate backward to guess what a noiseless answer would look like — can report improvements of up to twenty-one percent that turn out to be artifacts of the method itself, not real reductions in hardware noise. Let that sit for a second, because zero-noise extrapolation isn't some fringe technique — it's one of the most common tools labs and vendors reach for when they want a benchmark number to look better than the raw hardware actually performs. This one comes from Quantum Zeitgeist, and it's single-source reporting — we haven't independently confirmed it yet — so hold it loosely, and we'll be watching for independent replication before treating it as settled. But the implication, if it holds up, is uncomfortable for the whole industry: some fraction of the 'we improved performance by X percent' claims you've read this year, across multiple companies, may be measuring the mitigation math instead of the machine. That doesn't make zero-noise extrapolation worthless — it's a legitimate technique with real published theoretical grounding — but it means any benchmark leaning on it needs to show its work, ideally with a cross-check against a mitigation-free baseline. This is exactly the kind of story that won't make many headlines but should change how skeptically you read the next 'record performance' press release, quantum or otherwise. Consider it your standing reminder that in this field, the gap between a genuine result and a flattering methodology can be razor thin.

Third story: trapped-ion computing's scaling bottleneck might have a mathematical way around it. Trapped-ion qubits — individual charged atoms held in place by electromagnetic fields and linked with laser pulses — are prized for extremely clean, high-fidelity operations, but stringing together the entangling gates that connect many ions at once has always gotten harder, and slower, as the chain grows longer. A new numerical framework, detailed in an arXiv preprint, synthesizes multi-tone laser control fields — carefully shaped combinations of laser frequencies — that can drive entangling gates across ion chains scaled all the way up to a thousand ions, using only global control, meaning you're not adding a dedicated laser or control channel for every single ion as the chain grows. The preprint's headline finding is that the control resources needed stay modest even as ion count scales up dramatically, which, if it holds in real hardware, chips away at one of the most persistent objections to trapped-ion architectures: that control overhead explodes as you add ions, even though the qubits themselves are excellent. Now, this is a numerical framework — theory and simulation, not a demonstrated thousand-ion chip sitting on a lab bench — so the real test comes when someone tries implementing it on actual hardware, where laser noise, ion heating, and vacuum-system engineering all get a vote. But for a platform whose biggest knock has always been 'great qubits, hard to wire up at scale,' a credible theoretical path that keeps control complexity in check is worth logging. Watch for a hardware group to try validating pieces of this over the next year.

Fourth: a reminder that the quantum stock market and quantum science live in almost entirely separate universes. IonQ, D-Wave, Rigetti, and Quantum Computing Inc. all sold off sharply, with IonQ tumbling eight percent and the others down around six percent apiece. And the cause had nothing to do with qubits, error rates, or Nature papers — it was a broader risk-off wave tied to an oil-price spike out of the Strait of Hormuz, the kind of macro shock that sends investors fleeing speculative, pre-profit names first and asks questions later. These four fit that profile exactly: high-beta, richly valued relative to current revenue, and popular with retail traders who treat them like momentum plays rather than infrastructure bets. 24/7 Wall Street, which reported the move, noted there was no company-specific news behind any of the four drops — which is itself the tell. Here's my read: none of this tells you anything about whether silicon spin qubits, trapped ions, or superconducting chips are actually getting better, because that work happens in labs and fabs on completely different timescales than a trading session. The stock swings are a market-structure story, not a technology story, and it's worth separating the two every time a quantum stock headline crosses your feed — the temptation to read technical meaning into a price move is strong and almost always wrong. If you're tracking this sector for the science rather than the ticker, today's lesson is simple: watch the papers, not the price chart, on days like this.

Fifth headline: an early but notable industrial handshake. Quantinuum, Rolls-Royce, error-correction specialist Riverlane, and the University of Edinburgh signed an agreement to explore applying quantum computing to industrial design and simulation — the kind of workload that shows up in jet-engine engineering, where you're modeling airflow, materials, and combustion under conditions too complex for classical computers to fully capture. Now, 'agreement to explore' is doing a lot of work in that sentence — this is not a deployed application, not a benchmark, not even a pilot with a timeline attached. It's four organizations formally saying they're going to look at where quantum might actually help Rolls-Royce's engineering problems. But don't dismiss it outright, because who's in the room matters: Rolls-Royce is a jet-engine maker with real, expensive simulation problems and no obvious incentive to sign a press-release partnership for the sake of headlines; Riverlane brings the error-correction expertise that turns noisy qubits into something trustworthy enough for engineering-grade answers; Edinburgh brings academic firepower to keep the science honest. This is single-source reporting out of The Quantum Insider, so we haven't independently verified every detail — but the shape of it fits a pattern worth flagging: industrial giants aren't waiting for fault-tolerant quantum computers to exist before they start building the relationships and workflows they'll need when it does. Whether this particular pact produces anything Rolls-Royce actually uses is a multi-year question. Whether aerospace is quietly becoming one of quantum's most serious buyer industries? That trend just got another data point.

And last in the headlines: Europe just put real money behind manufacturing at scale, in a different qubit flavor. Pasqal, the French neutral-atom quantum computing company, and the EU's Chips Joint Undertaking formally launched Q-PLANET, a fifty million euro, multi-year program bringing together dozens of European partners to build a standardized fabrication line for neutral-atom quantum chips — including open Process and Assembly Design Kits, essentially shared blueprints and rulebooks that let multiple companies design chips for the same production line instead of everyone reinventing their own fab process from scratch. Neutral-atom qubits work by trapping individual atoms in tight grids of laser light and manipulating them with more laser light, and Pasqal is one of the platform's biggest commercial bets. Q-PLANET is one of six parallel EU pilot lines, each targeting a different qubit technology, which tells you something about where European policy has landed: rather than betting the continent's quantum future on one horse, the EU is trying to industrialize several manufacturing pathways at once and let the results sort out which scales best. Fifty million euros doesn't buy you a fault-tolerant quantum computer, but it buys the unglamorous, expensive plumbing — standardized processes, shared tooling, design kits — that turns a lab demo into something a hundred companies could build on. And that word, 'manufacturing,' is about to come up again, because our main story today is about exactly that question, for a completely different qubit technology, on a foundry line that's already up and running.

Our main story today, and I'm calling this one The Foundry Bet: can silicon spin qubits ride the entire semiconductor industry's existing manufacturing muscle straight to the scale quantum computing actually needs, instead of building an entirely new industrial supply chain from scratch? Here's what happened. Diraq, an Australian silicon-qubit company, and imec, the Belgian nanoelectronics research hub that's basically Europe's chip-fabrication brain trust, published a peer-reviewed paper in Nature Communications showing coherent operation of an eight-qubit silicon spin-qubit array — built entirely on imec's standard 300 millimeter CMOS manufacturing line, the exact kind of production line that makes ordinary computer chips by the billions. Now, what's a silicon spin qubit, and why does 'CMOS' matter so much here? A spin qubit encodes information in the quantum spin — think of it as a tiny built-in magnetic orientation — of a single electron trapped inside a structure that looks, physically, almost exactly like a normal transistor. CMOS is the standard manufacturing process behind essentially every chip in your phone, laptop, and car. Diraq's whole bet, as a company, is that because their qubits look like transistors, they can piggyback on decades of chip-fab infrastructure, expertise, and yield engineering instead of inventing an entirely new supply chain the way superconducting qubits, or trapped-ion and neutral-atom systems, largely have to. This new result is a fourfold jump from where the same team stood in 2025, when they demonstrated a two-qubit unit cell on the same platform. Now it's eight qubits, in a linear array, with — and this is the detail that actually matters technically — no loss of coherence or control fidelity as they scaled up. Coherence, for anyone hearing the term for the first time, is basically how long a qubit can hold onto its quantum state before noise scrambles it; losing coherence as you add qubits is one of the most common ways scaling attempts fail. There's a second, quieter detail buried in the paper that I think matters even more: scaling the readout architecture — the wiring and sensors you need to actually measure each qubit's state — didn't require a significant jump in sensor count, wiring density, or thermal load. That matters because readout wiring is one of the classic ways qubit architectures hit a wall; add more qubits, and normally you need proportionally more wires, more sensors, more heat management, until the wiring itself becomes the bottleneck. If this readout architecture keeps scaling favorably, that's a structural advantage, not just a one-time achievement. Diraq's CEO, Andrew Dzurak, isn't shy about what he thinks this proves. 'This is what an industrial pathway to quantum computing looks like,' he said, framing the result as evidence the company can hit its stated roadmap: thousands of qubits by 2029, more than one million qubits by 2031.

So how does this shape up when you dig into the actual voices behind it? Let's stack them. Kristiaan De Greve, fellow and program director for quantum computing at imec, framed it as a manufacturing story first: 'The future of quantum computing depends not only on qubit quality but also on the ability to manufacture increasingly complex quantum processors with the reproducibility, yield and scale of the semiconductor industry,' he said, adding that the result 'demonstrates that industrial 300mm CMOS-compatible manufacturing can support quantum systems beyond isolated qubit pairs.' His point, in plain terms: it's not enough to make one great qubit in a lab — you need to make thousands of them, identically, on a production line, the way chipmakers already do for ordinary processors. Dzurak went further, leaning into the cadence of it all. 'Nine months ago, we showed the world that silicon MOS qubits could be fabricated reliably using imec's 300 millimeter CMOS platform technology,' he said. 'Today, imec has scaled and Diraq has tested the size of the array using exactly the same process, with no compromise in coherence. This is the cadence we need to reach utility scale, and it is the type of cadence we expect to keep.' And on X, the Diraq company account, @diraqQC, posted the same framing to its followers: 'New paper in Nature Communications! With imec, we've demonstrated an 8-qubit silicon spin-qubit array made in a 300mm CMOS foundry. This proves we can scale with zero loss in coherence.' Now here's where I want to slow down, because 'proves we can scale' is exactly the kind of sentence this show exists to poke at. Eight qubits is still tiny — you could fit dozens of these arrays inside the smallest superconducting or trapped-ion processors already running today. This is a manufacturing-process proof point, not a working quantum processor, and going from eight qubits to 'thousands by 2029, over a million by 2031' is an aggressive extrapolation from one fourfold scale-up, achieved, by the company's own telling, in under a year. A fourfold jump is real and worth celebrating. Repeating that jump five or six more times in a row, on schedule, without new engineering walls showing up, is a very different claim than demonstrating it once. There's also a performance gap worth naming plainly: silicon spin qubits still lag superconducting and trapped-ion platforms on raw two-qubit gate fidelities in head-to-head comparisons today. Manufacturability at industrial scale doesn't automatically buy you performance parity — those are two separate races, and Diraq is currently ahead in exactly one of them. Here's the honest version for you, the listener: the manufacturing story here is genuinely exciting, and the readout-scaling detail is the kind of thing experts in the field will be citing for a while. But 'we did it on a real chip foundry line' and 'we're on track for a million qubits' are two very different sentences, and only one of them is backed by data in this paper. Time for the Hype Check. I'm putting this one at a six. The 300 millimeter CMOS result and the flat readout scaling are real, peer-reviewed, and matter — that's the substance. But the million-qubit-by-2031 framing is doing more work than eight qubits earns it, and until we see a second and third scale-up on the same cadence, treat the roadmap as a bet, not a schedule.

These manufacturing threads — Diraq's foundry line, Europe's six pilot programs, Nord Quantique's error numbers — are exactly what we'll keep pulling on this week, because 'can this actually scale' is quietly becoming the only question that matters across this industry. If the show's earning your time, hit follow wherever you're listening so you don't miss where this goes next. That's Quickly Quantum for today. New episodes every day. This is an AI-voiced podcast, created and built by a real person using today's cutting-edge technology. And remember: nothing on this show is financial advice. I'm Brian Lampert — see you tomorrow.