The Quantum Supply Chain: Where Every Component Comes From and What Breaks If It Stops

A procurement that depends on eight countries

In March 2026, the Q-PAC consortium assembled a quantum computer in Denver from components supplied by five companies in four countries: a Dutch QPU, Dutch control electronics, an American cryostat, Dutch cabling, and Australian calibration software. The supply chain spanned three continents. It worked, and the five-month timeline proved the Quantum Open Architecture model is production-viable.

Now consider what would have happened if one link in that chain had failed. Not a dramatic geopolitical rupture. Something smaller. An FPGA allocation slip at AMD/Xilinx in the United States delays the control electronics by four months. A Bluefors cryostat order from Finland hits a backlog because three national quantum programs placed orders the same quarter. A change in Dutch export-control interpretation delays QuantWare QPU shipment to a non-EU buyer for eight weeks while paperwork clears. Delft Circuits, the only commercial supplier of superconducting flex cabling, has a production batch fail quality inspection and the next available slot is three months out.

None of these scenarios require a war, a sanctions regime, or a supply chain “decoupling.” They require only the ordinary friction of a small, specialized industrial base operating at capacity. The quantum supply chain is not fragile because of geopolitics alone. It is fragile because it is narrow: a handful of companies, concentrated in a handful of countries, supplying the entire Western quantum hardware stack. For a buyer assembling a quantum computer, understanding where that narrowness sits is as important as understanding the physics. This article maps the supply chain layer by layer, identifies where concentration risk is highest, and explains how architecture and procurement strategy can mitigate it.

What concentration risk actually means for a quantum build

Concentration risk in classical IT procurement is an annoyance. If one server vendor has a backlog, you buy from another. The interfaces are standardized, the components are commoditized, and the switching costs are manageable. None of that is true for quantum computing in 2026.

The quantum hardware supply chain has three properties that make concentration risk acute. First, the supplier base at most layers is measured in single digits, not dozens. There are four production-grade Western control electronics vendors, three dilution refrigerator manufacturers, and exactly one commercial supplier of superconducting flex cabling. Second, the components are not interchangeable without engineering effort. A Qblox control system and a Zurich Instruments control system both drive superconducting qubits, but they use different synchronization architectures, different pulse languages, and different FPGA platforms. Switching from one to the other mid-build means re-engineering the integration, not swapping a rack module. Third, lead times are long and unpredictable. A dilution refrigerator takes four to twelve months depending on configuration. Control electronics depend on a global FPGA supply chain that does not prioritize quantum. Flex cabling is manufactured in small batches at a single facility.

The result is that a supply chain disruption at any layer does not produce a brief delay. It produces a schedule slip measured in quarters, because the alternative supplier requires re-engineering and has its own lead time. A buyer who does not map these dependencies before placing the first order is a buyer who discovers in month five that first qubit signal has slipped to month fourteen.

The supply chain map: who makes what, and where

The series has covered each of these components in technical depth. This section maps them geographically and identifies where concentration risk is highest.

Quantum processing units. The QPU is, paradoxically, the least concentrated layer for superconducting systems. QuantWare (Netherlands) is the highest-profile pure-play supplier, with over 50 customers across 20 countries and the most complete product line from 5-qubit Soprano through 64-qubit Tenor. Rigetti (United States) sells the 9-qubit Novera with four-to-six-week delivery. IQM (Finland) ships QPUs for European national programs, including the 54-qubit Radiance at LRZ. For other modalities, the QPU market is more concentrated: Pasqal (France) for neutral atom, Quantinuum (UK/US) and IonQ (US) for trapped ion, Diraq (Australia) for silicon spin. But within superconducting, a buyer has genuine alternatives across three countries. If one supplier faces a production issue or an export constraint, the others can fill the gap with manageable re-integration effort, because the QPU is the component the QOA model was designed to make swappable.

Cryogenic infrastructure. Three manufacturers serve the Western market: Bluefors (Finland), Maybell Quantum (United States), and Oxford Instruments (United Kingdom). Bluefors is dominant, with over 1,800 systems shipped and the widest product line. The concentration risk here is real but geographically diversified across three allied nations. The deeper risk is lead time, not geography: all three manufacturers are capacity-constrained, and a surge in national quantum program orders (the kind that the CHIPS Act quantum package, France’s €3.3 billion quantum investment, and the UK’s £2 billion ProQure program are now generating) can push delivery timelines past twelve months. The cryostat is the longest-lead item in a superconducting build. A buyer who waits to order the cryostat until the facility is ready adds six to twelve months to the program. The cryogenic infrastructure article covers the operational implications; the supply chain point is simpler: order early, confirm the delivery slot in writing, and have a second-choice vendor qualified in case the primary cannot deliver.

Helium-3. This is the structural dependency that has no short-term mitigation. The working fluid in every dilution refrigerator is derived almost entirely from the radioactive decay of tritium in nuclear weapons stockpiles. Terrestrial production runs at roughly 22,000 to 30,000 liters per year. Demand from quantum computing, medical imaging, neutron detection, and fusion research exceeds supply and is rising. The United States and Russia hold the largest stockpiles. Market prices range from $1,900 to $2,600 per liter; a single Bluefors XLD1000sl holds approximately 40 liters. The 2026 Gulf region helium supply disruptions primarily affect helium-4, but they illustrate how geopolitical events cascade into cryogenic supply chains.

Bluefors and Maybell have both signed forward-purchase agreements with Interlune for lunar-sourced helium-3, but Interlune’s CEO has acknowledged that commercial-scale extraction is “going to be in the early 2030s, no earlier than that.” Until then, helium-3 is a single point of failure in the supply chain that no procurement strategy can fully eliminate for superconducting and silicon-spin builds. The playbook’s recommendation is to negotiate multi-year helium-3 offtake as part of the cryostat procurement and to commission helium recovery infrastructure from day one so that the initial charge is recycled rather than replaced. Every warm-up event that vents helium-3 is a supply chain event, not just an operational one.

Control electronics. Three Western vendors: Qblox (Netherlands), Quantum Machines (Israel), and Zurich Instruments (Switzerland, Rohde & Schwarz group). The control system build guide covers the technical differentiation. From a supply chain perspective, the Netherlands carries outsized concentration risk across multiple layers: QuantWare (QPU), Qblox (control), and Delft Circuits (cabling) are all Dutch companies. A single Dutch export-control decision, a policy change, or even a natural disaster concentrated in the Delft-Leiden corridor would affect three layers of a superconducting build simultaneously. The mitigation is to qualify at least one non-Dutch alternative at each layer: Rigetti or IQM for the QPU, Zurich Instruments or Keysight for control, and conventional coaxial cabling (at the cost of reduced channel density) for the wiring.

Every control platform depends on AMD/Xilinx or Intel/Altera FPGAs, both manufactured in the United States. This is a hard US dependency that applies across all three control vendors, across all modalities that use commercial control electronics. As the control system article discusses in detail, FPGA allocation is frequently the binding constraint on control electronics delivery. The quantum industry is a small customer competing against defense, telecom, and AI for the same chips. A buyer cannot mitigate this dependency by switching control vendors, because all three share it. The only mitigation is early ordering, written allocation confirmation, and (for sovereign programs) a discussion with the FPGA vendor about strategic allocation priority.

Cryogenic cabling. Delft Circuits (Netherlands) is the sole commercial supplier of superconducting flex cabling. Its Cri/oFlex technology (NbTi superconducting stripline on polyimide, 0.3 mm thick) pushes channel density to 256 per cryostat loader today, with a roadmap to 4,096 by 2029. The alternative is conventional coaxial cabling, which tops out at roughly 168 channels per loader and is too thick and thermally conductive to scale past a few hundred qubits. For any superconducting system targeting more than 100 qubits, Delft Circuits is a dependency without a commercial substitute. This is the narrowest chokepoint in the Western quantum supply chain. A buyer planning a 200-qubit-class system in 2027 should be in conversation with Delft Circuits about production capacity now, not when the QPU order lands.

Calibration software. Three platforms: Q-CTRL Boulder Opal Scale-Up (Australia), QuantrolOx Quantum EDGE (United Kingdom), and Quantum Machines QUAlibrate (Israel, bundled with OPX hardware). The geographic diversification here is reasonable. The risk is functional rather than geographic: QUAlibrate is tied to Quantum Machines hardware, so a buyer who chooses Qblox or Zurich Instruments for control is choosing between Q-CTRL and QuantrolOx for calibration. The OS and orchestration article covers why this choice matters operationally. From a supply chain perspective, the calibration layer has adequate diversification and low switching cost relative to other layers.

QEC decoders. Riverlane (United Kingdom) is the most prominent commercially available real-time decoder vendor, with Deltaflow 2 deployed at OQC and Oak Ridge National Laboratory. GPU-hosted decoders running on NVIDIA hardware via NVQLink are the alternative path, using open-source software (PyMatching) on commercially available GPU nodes. The decoder layer has lower concentration risk than cabling or cryogenics because the GPU path provides a genuine, independently sourced alternative. The HPC integration article covers the decoder decision in detail.

HPC integration. NVIDIA dominates this layer through NVQLink, CUDA-Q, and the Grace Hopper / GB200 GPU platform. This is a single-vendor dependency for the low-latency QPU-HPC coupling that real-time error correction requires. The dependency is deep (NVQLink is the only production-validated Western interconnect for microsecond-class GPU-QPU round trips), but it is a dependency on the world’s largest GPU company, which reduces the supply chain risk relative to a dependency on a five-person startup. QRMI, the scheduling abstraction, is an open standard and not NVIDIA-locked. The classical HPC hardware (servers, Ethernet switches, RDMA fabric) is commodity infrastructure with multiple suppliers.

Export controls as procurement constraints

The full treatment of quantum export controls is in my dedicated analysis. For a builder, the actionable summary is this: the 2024 Wassenaar Arrangement update added quantum computing entries, the U.S. Export Administration Regulations control certain quantum technologies, FPGAs, and cryo-CMOS components, and the EU Dual-Use Regulation 2021/821 includes quantum items on its control list. Origin Quantum and QuantumCTek are on the U.S. Entity List. For government, defense, and regulated-sector programs, the playbook specifies an end-to-end Western supply chain with no PRC-origin hardware or software.

What this means in practice: a buyer outside the United States, the EU, or a close ally should verify component-by-component eligibility before issuing purchase orders. A non-EU buyer sourcing QuantWare QPUs from the Netherlands needs to confirm that the Dutch export-control authority will approve the shipment to their jurisdiction. A buyer in a country without an FPGA supply agreement needs to confirm AMD/Xilinx allocation will not be restricted. These are not theoretical concerns. They are the procurement equivalent of the visa check before booking the flight.

Quantum Open Architecture as supply chain resilience

This is the argument the supply chain geography map makes visually, and it deserves to be stated plainly: Quantum Open Architecture is not only a technical architecture for assembling quantum computers from modular components. It is a supply chain resilience strategy.

In the vertically integrated model (IBM, Google, Quantinuum selling complete systems), the buyer has no supply chain risk to manage because they have no supply chain choices to make. They also have no alternatives if the vendor has a production problem, a policy change, an export restriction, or a strategic pivot. The buyer is locked in, and lockout is the other side of lock-in.

In the QOA model, every layer has at least two qualified Western suppliers in different jurisdictions. If QuantWare cannot deliver a QPU, Rigetti or IQM can. If Qblox has FPGA allocation issues, Zurich Instruments or Keysight are alternatives with different FPGA platforms. If Bluefors has a twelve-month backlog, Maybell or Oxford Instruments offer comparable dilution refrigerators. The switching cost is real (no two vendors are plug-compatible, and re-integration requires engineering effort), but it is bounded effort measured in weeks, not the unbounded dependency of a single-vendor lock-in where the alternative is to start the entire program over.

The QUB reference architecture, the QuantWare-Q-CTRL-Qblox joint design validated by Q-PAC, takes this further by pre-validating specific multi-vendor combinations. A buyer who specifies QUB components gets a tested integration path. A buyer who needs to substitute one vendor for sovereignty reasons can use the QUB as a baseline and re-validate the specific interface that changed, rather than re-integrating the entire stack.

For sovereign quantum programs, QOA also provides something the vertically integrated model cannot: transparency into the supply chain. When you buy an IBM Quantum System Two, you do not know which cryostat manufacturer, which FPGA, which cabling technology, or which calibration firmware is inside. When you assemble a QOA system, you know exactly what is in every layer, where it came from, and whether it meets your supply chain requirements. That transparency is why the playbook specifies QOA as the preferred model for defense and intelligence customers.

Modality as a supply chain decision

One dimension of the supply chain that no other article in this series frames explicitly: the choice of qubit modality is partly a supply chain choice.

A superconducting build carries the deepest supply chain. It requires a dilution refrigerator, helium-3, precision microwave cabling, commercial control electronics, and a GPU node for real-time decoding. The supply chain spans cryogenics vendors in Finland, the US, and the UK; QPU vendors in the Netherlands, the US, and Finland; control vendors in the Netherlands, Israel, and Switzerland; a sole-source cabling vendor in the Netherlands; and FPGA vendors in the US. The QOA ecosystem is most mature here, but so is the dependency graph.

A neutral-atom build sidesteps the entire cryogenic supply chain. No dilution refrigerator, no helium-3, no millikelvin cabling. Pasqal’s Orion runs at room temperature in a standard server rack. The supply chain shifts to lasers (Toptica, M-Squared, both European) and ultra-high vacuum components (a broader, less concentrated supplier base). The trade-off is that neutral-atom systems ship as largely integrated appliances from Pasqal, so the QOA model is less developed here and the buyer trades cryogenic supply chain risk for single-vendor integration risk.

A silicon-spin build eliminates helium-3 dependency. The higher operating temperature (approximately 1 K versus 10 to 20 mK) enables helium-4-only cryostats. The CMOS fabrication thesis means QPUs can be manufactured at existing semiconductor fabs, eventually at volume. The supply chain risk shifts to the semiconductor industry’s constraints (fab access, cleanroom time, lithography tools) rather than quantum-specific ones. The QPU vendor base is currently narrow (Diraq in Australia, SemiQon in Finland, Intel in the US), but the manufacturing path is the one most likely to commoditize over time.

A sovereign program choosing its primary modality should consider this supply chain dimension alongside the physics. The program that picks superconducting gets the most mature QOA ecosystem and the deepest dependency graph. The program that picks neutral atom gets the simplest supply chain and the least QOA flexibility. The program that picks silicon spin bets on a supply chain that barely exists today but has the clearest path to industrial scale. None of these is the “right” answer. All of them are supply chain decisions disguised as physics decisions.

What a buyer should take away

Map your supply chain before you place your first order. Identify which components are sole-sourced, which have qualified alternatives, and which are subject to export controls that apply to your program. The playbook’s end-to-end Western supply chain specification for defense customers is a starting point, not an endpoint. The specific list of qualified vendors changes by jurisdiction, by modality, and by program classification.

Qualify at least two vendors at the layers where concentration risk is highest: QPU, cryogenics, and control electronics. The engineering effort to qualify a backup vendor is an insurance premium, not a waste. The cost of not having one is a program that stops when a single supplier cannot deliver.

Order the long-lead items early and in parallel. The cryostat, the control electronics (with FPGA allocation confirmed in writing), and the flex cabling are the three components most likely to set the program schedule. Treating any of them as a late-stage procurement is the single most common cause of schedule overruns in quantum builds.

Treat modality selection as partly a supply chain decision. If your program has a sovereignty mandate that restricts sourcing to domestic or allied vendors, map the available vendor base per modality before committing to the physics. The modality that best fits your computational goals may not be the modality that best fits your supply chain constraints.

For the security hardening decisions that protect a deployment once it is built (facility security, network segmentation, firmware integrity, and personnel) see the security article in this series.

For the full geopolitical and policy analysis of these dependencies, see my Quantum Sovereignty series, particularly the chokepoints analysis and the export controls article. For how concentration risk connects to national quantum strategies and the new geopolitics of technology, see my book Quantum Sovereignty.