In my previous article, “Quantum Is Now a Procurement Problem, Not a Research Problem”, I argued that the quantum industry is moving beyond laboratory prestige. Buyers are increasingly asking whether quantum systems can be integrated, secured, supported and delivered against a real operational requirement and not merely demonstrated in controlled conditions. There is a deeper implication. If quantum is becoming a procurement category, it must first become a manufacturing category.
The manufacturing problem is not identical across the field. Superconducting qubits depend on nanoscale tunnel-junction control. photonic systems require specialised materials, switching structures and packaging; silicon-spin approaches rely on semiconductor fabrication. and trapped-ion, neutral-atom and diamond-based systems bring different challenges in optics, vacuum, materials and integration. Quantum therefore has no single manufacturing playbook. It has several parallel industrialisation problems.
One recent process analysis found that modest critical-current variation can translate into tens of megahertz of qubit-frequency spread in a typical superconducting design, large enough to complicate frequency allocation and calibration in dense devices. Exact values depend on the qubit architecture and fabrication process.
That is where the sector faces its most uncomfortable truth.
it can produce extraordinary devices, but it has not yet demonstrated that it can manufacture large numbers of materially identical, reliable and economically viable ones. A quantum processor is not a single invention. It is an assembly of delicate materials, nanofabricated elements, microwave wiring, packaging, cryogenics, control electronics, calibration software and test infrastructure.
A breakthrough at any one layer is not enough if the rest of the stack cannot be reproduced.
The next quantum breakthrough may not look like a record-breaking processor. It may look like a better tunnel barrier, a stable fabrication recipe, a high-yield packaging process, an automated test flow, or an unglamorous cleanroom that delivers the same result repeatedly. That is not a footnote to the quantum story. It is the story.
Table of content
The qubit-count distraction
The Josephson-junction reality check
A quantum processor is a factory product
The cleanroom is now strategic infrastructure
Repeatability is the real moat
Manufacturing changes the investment case
The next milestones should look boring
The qubit-count distraction
Quantum companies understandably announce qubit milestones. They are simple, legible and easy to compare. A 100-qubit machine appears more advanced than a 50-qubit machine in a way that an improvement in wafer-scale parameter variation does not.
But qubit count is an incomplete metric - and sometimes a misleading one.
A useful quantum processor must not only contain qubits. Those qubits need adequate coherence, controllable gates, stable frequencies, low crosstalk, tolerable calibration overhead and a workable error-correction path. As systems become larger, the challenge becomes less about creating one excellent qubit and more about ensuring that many qubits behave predictably enough to work together.
That is fundamentally a manufacturing question.
For superconducting systems, the Josephson junction sits at the heart of the problem. This tiny structure controls the nonlinear electrical behaviour that makes a transmon qubit possible.
Small variation in its critical current changes the qubit’s transition frequency.
One recent review ( arxiv - Advances in Josephson Junction Materials and Processes Toward Practical Quantum Computing ) notes that a 3 percent fluctuation in critical current can produce about a 1.5 percent frequency change approximately 75 MHz for a qubit designed to operate at 5 GHz. That is substantial when typical qubit frequency spacings are only tens to around 100 MHz.
In other words, a fabrication variation that sounds marginal in ordinary engineering terms can produce a major systems-level consequence. Qubits can collide in frequency, become harder to address individually, exhibit stronger unwanted interactions, or force designers to discard otherwise usable areas of a device.
This is why “we made more qubits” does not necessarily mean “we can build a better computer.”
The Josephson-junction reality check
The quantum industry’s manufacturing constraint is becoming visible in where serious money is going.
In August, Princeton University announced that it would lead a new US National Science Foundation Quantum Leap Challenge Institute called MARQUIS:
Manufacturable and Resilient superconducting Quantum Information Systems. The institute will receive $27.9 million over five years and is one of eight initiatives funded in an NSF package totaling $290 million. this title is actually more important than it first appears.
Not faster. Not bigger. Not more exotic. but -
Manufacturable and resilient.
MARQUIS is focused on materials and fabrication methods for superconducting processors, including the Josephson junction. It brings together universities, semiconductor-process organisations and industry advisers to address barriers between high-quality quantum devices in research settings and devices that can be made consistently at scale.
The implication is blunt: the sector has recognized that its core circuit element remains a bottleneck.
A prior study ( inspirehep) examining more than 3,600 Josephson junctions found average resistance variation of 3.7 percent across a wafer, with 2.7 percent average variation on individual chips. The authors inferred wafer-level qubit transition-frequency variation of 1.7 percent to 2.5 percent, with 60 percent to 70 percent attributed to junction-area fluctuations and the remainder to tunnel-junction inhomogeneity.
Those figures do not mean quantum fabrication is failing. They show why it is difficult. If a processor has only a handful of qubits, engineers can select good devices, customise frequencies and manually work around imperfections. At hundreds or thousands of physical qubits, that approach becomes economically and operationally absurd.
A large-scale quantum computer cannot depend on heroic intervention.
It must depend on process control.
A quantum processor is a factory product
The phrase “manufacturing quantum computers” can create the wrong mental image. It suggests a conventional factory line stamping out finished boxes. The reality is closer to an extreme version of advanced semiconductor production combined with precision microwave engineering, specialist materials science and cryogenic systems integration.
The chip is only the beginning.
Each quantum processor requires wafers, lithography, deposition, etching, cleaning, oxidation control, metrology, dicing and packaging. It may require microwave resonators, interconnects, filters, amplifiers and low-temperature wiring. It then has to be connected to a control stack that can generate and measure signals with exacting timing and stability. Finally, the system must be cooled, calibrated, benchmarked and continuously maintained.
The burden compounds as systems grow.
More qubits mean more control channels, more possible interference paths, more calibration parameters, more components that can drift and more opportunities for a small fabrication variation to become a large operational failure. The quantum computer is therefore not one machine. It is a tightly coordinated manufacturing-and-operations system.
This is why the field should stop treating fabrication as a subordinate engineering detail. It is the central constraint between a promising qubit and a commercially deployable processor.
The cleanroom is now strategic infrastructure
The UK offers a useful example of the shift from research capacity towards industrial capacity.
In May, Infleqtion announced a Quantum Innovation Centre at Oxford Technology Park, describing it as a facility that will triple the company’s UK research, production and systems-integration footprint. The company explicitly links the site to manufacturing, integration and deployment rather than positioning it only as an R&D expansion. That may sound mundane beside dramatic claims about neutral atoms, fault tolerance or future quantum advantage. but it is not.
The availability of the right facility determines whether a company can move from a one-off device to a repeatable production cycle. It affects process ownership, iteration speed, intellectual-property control, quality assurance, workforce development and the ability to deliver systems without becoming dependent on a fragile external supply chain.
It also reveals a wider strategic reality.
Quantum manufacturing requires specialised cleanrooms, measurement equipment, packaging facilities, cryogenic test capability and people who can operate all of it. These assets cannot be spun up overnight. They are expensive, capacity-constrained and often concentrated in a small number of regions or institutions.
The industry may talk about a race for error-corrected quantum computing. In practice, it is also racing for fabrication access, test capacity and manufacturing talent.
Repeatability is the real moat
Start-ups often seek differentiation through a distinctive qubit architecture, algorithm, material or control technique. Those matter. But as quantum moves closer to a product market, the defensible advantage may increasingly sit elsewhere: in the capacity to make the same device twice. and then a one hundred times. and then one thousand times - all with known yields, traceable performance variation, qualified suppliers and a maintenance model that does not require the inventors to remain permanently in the loop.
That is the difference between a scientific achievement and an industrial platform.
The semiconductor industry learned this lesson over decades.
The winning companies were not simply those that could demonstrate a transistor, a chip or a node shrink. They were the firms that built process-control systems, equipment ecosystems, design rules, packaging standards and vast manufacturing organisations around those achievements.
Quantum will not replicate the semiconductor industry exactly. Its systems involve radically different physics, different operating environments and, in many cases, more complex integration requirements. But the economic logic is the same. Exceptional performance without repeatability does not scale.
Manufacturing changes the investment case too.
because it has consequences for investors, governments and enterprise buyers.
For investors, a quantum company’s technical claims should be assessed alongside less glamorous questions:
What is the fabrication process, and who owns it?
What are the critical materials and single-source components?
What is the device yield, not simply the best observed performance?
How much calibration is automated?
What happens when a system must be reproduced at another site?
Can the company service, upgrade and qualify units after delivery?
For governments, manufacturing capacity is becoming part of quantum sovereignty. The competition is not just to host exceptional physicists, it is to retain critical process know-how, trusted suppliers, cleanroom access and the skilled workforce needed to turn devices into systems.
For buyers, it means due diligence must extend beyond benchmarks. A procurement team should ask whether a supplier can deliver a machine that meets a specification repeatedly, on a schedule, with a secure support model and an intelligible path to upgrades.
A remarkable prototype can win a conference. A repeatable product wins a contract.
The next milestones should look boring
The industry will continue to celebrate qubit-count records. It should. Technical progress deserves recognition. But the milestones that matter most may soon sound less exciting:
Lower device-to-device frequency variation
Improved wafer yield
Longer-term process stability
Faster and more automated calibration
Better packaging and interconnect reliability
Shorter build-to-test cycles
Standardized qualification and test methods
More suppliers capable of making critical components
These are not marketing-friendly headlines. They are, however, the conditions for an actual quantum industry.
There is an irony here. Quantum computing is routinely described as one of the most revolutionary technologies ever developed. Yet its progress towards commercial relevance may depend on an old industrial discipline:
make the same thing reliably, understand why it works, document the process, control the variation and improve the yield.
and none of it is glamorous. It is much harder than glamorous. but its where the quantum race will increasingly and likely be won.