Marin Ivezic
A superconducting quantum processor spends its working life at about 10 millikelvins, roughly a hundredth of a degree above absolute zero. The chip is the part that gets photographed. The machine holding it at that temperature, a dilution refrigerator, comes from a handful of manufacturers worldwide, and their production schedules do more to set a customer’s delivery date than any qubit-count roadmap does.
That gap turns up constantly in the procurement reviews our participants bring into class. A vendor supplies a roadmap and a physics briefing. Neither document names the suppliers whose capacity decides whether the roadmap holds, and neither is dishonest for the omission. The missing information belongs to the suppliers, who have never been asked to summarize it.
Marin Ivezic’s supply chain series on PostQuantum.com maps that layer approach by approach. Four questions carry most of the analytical work, and you can run all four against a single vendor in an afternoon.
One term first. A modality is the physical system a machine uses to hold a qubit. Superconducting circuits, trapped ions, photons, neutral atoms, and electron spins in silicon are the five choices in serious commercial development today. Five physics decisions, and five different industrial bases behind them.
The bill of materials
The first question is the plainest. What does this machine physically require? We ask participants to start with the bill of materials rather than the specification sheet, because each modality depends on a distinct set of enabling hardware.
- Superconducting. Dilution refrigerators, helium-3 as the working fluid, microwave control electronics, and Josephson junction fabrication, the nanoscale element every superconducting qubit is built from.
- Trapped ions. Precision laser systems, ultra-high vacuum chambers, and increasingly, trap chips made on semiconductor production lines.
- Photonic. Photonic integrated circuit foundry capacity, single-photon sources and detectors, low-loss waveguides and fiber, and classical electronics fast enough to act on a measurement within nanoseconds.
- Neutral atoms. Optical tweezer arrays built from spatial light modulators, devices that split one laser beam into a grid of traps, plus high-power lasers to excite the atoms into the states that let them interact, vacuum chambers, and precision optics.
- Silicon spin. Isotopically purified silicon-28, cryogenic control chips, and lithography at leading-edge nodes.
Those five lists barely overlap. A vendor comparison written at the qubit level treats them as five versions of one purchase, and they are five separate purchases with five separate failure modes.
Count the suppliers, then place them on a map
Two answers, two different kinds of risk, and this is the distinction we press hardest on.
Dilution refrigerators come from a short list dominated by Finnish, British and Dutch firms. Helium-3 is scarcer still – it is produced almost entirely as a decay product of tritium held in national stockpiles, and governments allocate the supply. Isotopically purified silicon-28 requires an enrichment step offered by very few providers anywhere. Leading-edge lithography means ASML, one company in one country, operating under an export licence regime that both the Netherlands and the United States have tightened since 2023.
A layer with three suppliers in three countries is a commercial risk, and you manage it with dual sourcing and inventory. A layer with three suppliers in one jurisdiction is a policy risk, and it becomes your risk the first time that jurisdiction revises its control list.
National programs meet the same distinction from the other side. Sovereign capability in a given modality follows the industrial base a country already has, so a nation with a semiconductor sector and a nation with a precision-optics sector are not equally placed to build the same machine, whatever their research funding says.
What the processor companies are buying
The third question turns the analysis around: watch what the hardware companies build, buy, and bring in-house. IBM fabricates its own Josephson junctions. PsiQuantum builds through a commercial semiconductor foundry rather than a laboratory cleanroom. Trapped-ion builders have been moving trap fabrication onto standard semiconductor production lines rather than making traps by hand in a laboratory.
Each decision is a statement about a constraint. Companies integrate the layer they cannot afford to wait for, so a move of this kind tells you which supplier the buyer had decided it could no longer schedule around. These are bottleneck assessments made by the people with the best information, published in the form of a press release.
There’s a second reading for anyone assessing where value settles. Every layer a processor company absorbs stops being an independent business, and an independent supplier is a great deal easier to buy from than a competitor’s subsidiary.
Shared dependencies and correlated risk
The fourth question is the one worth the most time. What else draws on the same supply?
Photonic quantum computing is described as room-temperature, and that description holds for the processor. It doesn’t hold for the detectors. Superconducting nanowire single-photon detectors, used in many photonic designs, need cryogenic cooling of their own, which puts a room-temperature machine back in the queue for cryogenic hardware.
Trapped-ion and neutral-atom builders compete for capacity from the same precision laser and optics suppliers. Every modality eventually needs a real-time decoder, the classical hardware and software that reads error-correction measurements and works out which corrections to apply fast enough to keep pace with the machine. That last one is a conventional computing problem, and all five approaches have to solve it.
The consequence is a portfolio consequence. If you hold three quantum positions across three modalities and all three depend on one optics supplier, you hold one position with extra paperwork.
Running the four questions together
Take them in order, because the physical requirement determines everything below it. The output is a tiered map. At the top, suppliers with high concentration and no near-term substitute. Below them, suppliers with genuine alternatives and long switching times. At the bottom, infrastructure and distribution, where the exposure is cost and schedule rather than availability.
Two cautions we give participants. The map describes exposure, not likelihood, and it won’t tell you which modality wins. And it dates quickly. An acquisition, a new fab, or a revised control list can redraw a layer inside a quarter, so we redraw ours on that cycle.
Where we teach this
Supplier concentration, jurisdiction, and correlated dependency are procurement questions before they are physics questions. We teach them that way, inside the technology strategy and vendor assessment material rather than as an appendix to the hardware sessions, because the participants who need this are usually the ones signing the contract rather than operating the machine.
For the modality-by-modality detail behind these four questions, including the tiered supplier maps for each approach, Marin Ivezic’s quantum computing supply chain series is the long-form treatment.
The technology strategy and vendor assessment program at quantumacademy.com/ covers the four questions in full, with current enrollment and access windows listed there.